Online Communities and Social Computing - OCSC 2011

Lecture Notes in Computer Science Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris...

Author: A. Ant Ozok | Panayiotis Zaphiris

1110 downloads 2020 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Lecture Notes in Computer Science Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board David Hutchison Lancaster University, UK Takeo Kanade Carnegie Mellon University, Pittsburgh, PA, USA Josef Kittler University of Surrey, Guildford, UK Jon M. Kleinberg Cornell University, Ithaca, NY, USA Alfred Kobsa University of California, Irvine, CA, USA Friedemann Mattern ETH Zurich, Switzerland John C. Mitchell Stanford University, CA, USA Moni Naor Weizmann Institute of Science, Rehovot, Israel Oscar Nierstrasz University of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen TU Dortmund University, Germany Madhu Sudan Microsoft Research, Cambridge, MA, USA Demetri Terzopoulos University of California, Los Angeles, CA, USA Doug Tygar University of California, Berkeley, CA, USA Gerhard Weikum Max Planck Institute for Informatics, Saarbruecken, Germany

6778

A. Ant Ozok Panayiotis Zaphiris (Eds.)

Online Communities and Social Computing 4th International Conference, OCSC 2011 Held as Part of HCI International 2011 Orlando, FL, USA, July 9-14, 2011 Proceedings

13

Volume Editors A. Ant Ozok UMBC, Information Systems Department 1000 Hilltop Circle Baltimore, MD 21250, USA E-mail: [email protected] Panayiotis Zaphiris Cyprus University of Technology Department of Multimedia and Graphic Arts 3036 Limassol, Cyprus E-mail: [email protected]

ISSN 0302-9743 e-ISSN 1611-3349 e-ISBN 978-3-642-21796-8 ISBN 978-3-642-21795-1 DOI 10.1007/978-3-642-21796-8 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2011929699 CR Subject Classiﬁcation (1998): K.4, K.6, H.3-5, C.2, J.1, J.3 LNCS Sublibrary: SL 3 – Information Systems and Application, incl. Internet/Web and HCI

© Springer-Verlag Berlin Heidelberg 2011 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, speciﬁcally the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microﬁlms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a speciﬁc statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Camera-ready by author, data conversion by Scientiﬁc Publishing Services, Chennai, India Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Foreword

The 14th International Conference on Human–Computer Interaction, HCI International 2011, was held in Orlando, Florida, USA, July 9–14, 2011, jointly with the Symposium on Human Interface (Japan) 2011, the 9th International Conference on Engineering Psychology and Cognitive Ergonomics, the 6th International Conference on Universal Access in Human–Computer Interaction, the 4th International Conference on Virtual and Mixed Reality, the 4th International Conference on Internationalization, Design and Global Development, the 4th International Conference on Online Communities and Social Computing, the 6th International Conference on Augmented Cognition, the Third International Conference on Digital Human Modeling, the Second International Conference on Human-Centered Design, and the First International Conference on Design, User Experience, and Usability. A total of 4,039 individuals from academia, research institutes, industry and governmental agencies from 67 countries submitted contributions, and 1,318 papers that were judged to be of high scientiﬁc quality were included in the program. These papers address the latest research and development eﬀorts and highlight the human aspects of design and use of computing systems. The papers accepted for presentation thoroughly cover the entire ﬁeld of human–computer interaction, addressing major advances in knowledge and eﬀective use of computers in a variety of application areas. This volume, edited by A. Ant Ozok and Panayiotis Zaphiris, contains papers in the thematic area of online communities and social computing (OCSC), addressing the following major topics: • • • • •

On-line communities and intelligent agents in education and research Blogs, wikis and twitters Social computing in business and the enterprise Social computing in everyday life Information management in social computing

The remaining volumes of the HCI International 2011 Proceedings are: • Volume 1, LNCS 6761, Human–Computer Interaction—Design and Development Approaches (Part I), edited by Julie A. Jacko • Volume 2, LNCS 6762, Human–Computer Interaction—Interaction Techniques and Environments (Part II), edited by Julie A. Jacko • Volume 3, LNCS 6763, Human–Computer Interaction—Towards Mobile and Intelligent Interaction Environments (Part III), edited by Julie A. Jacko • Volume 4, LNCS 6764, Human–Computer Interaction—Users and Applications (Part IV), edited by Julie A. Jacko • Volume 5, LNCS 6765, Universal Access in Human–Computer Interaction— Design for All and eInclusion (Part I), edited by Constantine Stephanidis

VI

Foreword

• Volume 6, LNCS 6766, Universal Access in Human–Computer Interaction— Users Diversity (Part II), edited by Constantine Stephanidis • Volume 7, LNCS 6767, Universal Access in Human–Computer Interaction— Context Diversity (Part III), edited by Constantine Stephanidis • Volume 8, LNCS 6768, Universal Access in Human–Computer Interaction— Applications and Services (Part IV), edited by Constantine Stephanidis • Volume 9, LNCS 6769, Design, User Experience, and Usability—Theory, Methods, Tools and Practice (Part I), edited by Aaron Marcus • Volume 10, LNCS 6770, Design, User Experience, and Usability— Understanding the User Experience (Part II), edited by Aaron Marcus • Volume 11, LNCS 6771, Human Interface and the Management of Information—Design and Interaction (Part I), edited by Michael J. Smith and Gavriel Salvendy • Volume 12, LNCS 6772, Human Interface and the Management of Information—Interacting with Information (Part II), edited by Gavriel Salvendy and Michael J. Smith • Volume 13, LNCS 6773, Virtual and Mixed Reality—New Trends (Part I), edited by Randall Shumaker • Volume 14, LNCS 6774, Virtual and Mixed Reality—Systems and Applications (Part II), edited by Randall Shumaker • Volume 15, LNCS 6775, Internationalization, Design and Global Development, edited by P.L. Patrick Rau • Volume 16, LNCS 6776, Human-Centered Design, edited by Masaaki Kurosu • Volume 17, LNCS 6777, Digital Human Modeling, edited by Vincent G. Duﬀy • Volume 19, LNCS 6779, Ergonomics and Health Aspects of Work with Computers, edited by Michelle M. Robertson • Volume 20, LNAI 6780, Foundations of Augmented Cognition: Directing the Future of Adaptive Systems, edited by Dylan D. Schmorrow and Cali M. Fidopiastis • Volume 21, LNAI 6781, Engineering Psychology and Cognitive Ergonomics, edited by Don Harris • Volume 22, CCIS 173, HCI International 2011 Posters Proceedings (Part I), edited by Constantine Stephanidis • Volume 23, CCIS 174, HCI International 2011 Posters Proceedings (Part II), edited by Constantine Stephanidis I would like to thank the Program Chairs and the members of the Program Boards of all Thematic Areas, listed herein, for their contribution to the highest scientiﬁc quality and the overall success of the HCI International 2011 Conference. In addition to the members of the Program Boards, I also wish to thank the following volunteer external reviewers: Roman Vilimek from Germany, Ramalingam Ponnusamy from India, Si Jung “Jun” Kim from the USA, and Ilia Adami, Iosif Klironomos, Vassilis Kouroumalis, George Margetis, and Stavroula Ntoa from Greece.

Foreword

VII

This conference would not have been possible without the continuous support and advice of the Conference Scientiﬁc Advisor, Gavriel Salvendy, as well as the dedicated work and outstanding eﬀorts of the Communications and Exhibition Chair and Editor of HCI International News, Abbas Moallem. I would also like to thank for their contribution toward the organization of the HCI International 2011 Conference the members of the Human–Computer Interaction Laboratory of ICS-FORTH, and in particular Margherita Antona, George Paparoulis, Maria Pitsoulaki, Stavroula Ntoa, Maria Bouhli and George Kapnas. July 2011

Constantine Stephanidis

Organization

Ergonomics and Health Aspects of Work with Computers Program Chair: Michelle M. Robertson Arne Aar˚ as, Norway Pascale Carayon, USA Jason Devereux, UK Wolfgang Friesdorf, Germany Martin Helander, Singapore Ed Israelski, USA Ben-Tzion Karsh, USA Waldemar Karwowski, USA Peter Kern, Germany Danuta Koradecka, Poland Nancy Larson, USA Kari Lindstr¨om, Finland

Brenda Lobb, New Zealand Holger Luczak, Germany William S. Marras, USA Aura C. Matias, Philippines Matthias R¨ otting, Germany Michelle L. Rogers, USA Dominique L. Scapin, France Lawrence M. Schleifer, USA Michael J. Smith, USA Naomi Swanson, USA Peter Vink, The Netherlands John Wilson, UK

Human Interface and the Management of Information Program Chair: Michael J. Smith Hans-J¨ org Bullinger, Germany Alan Chan, Hong Kong Shin’ichi Fukuzumi, Japan Jon R. Gunderson, USA Michitaka Hirose, Japan Jhilmil Jain, USA Yasufumi Kume, Japan Mark Lehto, USA Hirohiko Mori, Japan Fiona Fui-Hoon Nah, USA Shogo Nishida, Japan Robert Proctor, USA

Youngho Rhee, Korea Anxo Cereijo Roib´ as, UK Katsunori Shimohara, Japan Dieter Spath, Germany Tsutomu Tabe, Japan Alvaro D. Taveira, USA Kim-Phuong L. Vu, USA Tomio Watanabe, Japan Sakae Yamamoto, Japan Hidekazu Yoshikawa, Japan Li Zheng, P. R. China

X

Organization

Human–Computer Interaction Program Chair: Julie A. Jacko Sebastiano Bagnara, Italy Sherry Y. Chen, UK Marvin J. Dainoﬀ, USA Jianming Dong, USA John Eklund, Australia Xiaowen Fang, USA Ayse Gurses, USA Vicki L. Hanson, UK Sheue-Ling Hwang, Taiwan Wonil Hwang, Korea Yong Gu Ji, Korea Steven A. Landry, USA

Gitte Lindgaard, Canada Chen Ling, USA Yan Liu, USA Chang S. Nam, USA Celestine A. Ntuen, USA Philippe Palanque, France P.L. Patrick Rau, P.R. China Ling Rothrock, USA Guangfeng Song, USA Steﬀen Staab, Germany Wan Chul Yoon, Korea Wenli Zhu, P.R. China

Engineering Psychology and Cognitive Ergonomics Program Chair: Don Harris Guy A. Boy, USA Pietro Carlo Cacciabue, Italy John Huddlestone, UK Kenji Itoh, Japan Hung-Sying Jing, Taiwan Wen-Chin Li, Taiwan James T. Luxhøj, USA Nicolas Marmaras, Greece Sundaram Narayanan, USA Mark A. Neerincx, The Netherlands

Jan M. Noyes, UK Kjell Ohlsson, Sweden Axel Schulte, Germany Sarah C. Sharples, UK Neville A. Stanton, UK Xianghong Sun, P.R. China Andrew Thatcher, South Africa Matthew J.W. Thomas, Australia Mark Young, UK Rolf Zon, The Netherlands

Universal Access in Human–Computer Interaction Program Chair: Constantine Stephanidis Julio Abascal, Spain Ray Adams, UK Elisabeth Andr´e, Germany Margherita Antona, Greece Chieko Asakawa, Japan Christian B¨ uhler, Germany Jerzy Charytonowicz, Poland Pier Luigi Emiliani, Italy

Michael Fairhurst, UK Dimitris Grammenos, Greece Andreas Holzinger, Austria Simeon Keates, Denmark Georgios Kouroupetroglou, Greece Sri Kurniawan, USA Patrick M. Langdon, UK Seongil Lee, Korea

Organization

Zhengjie Liu, P.R. China Klaus Miesenberger, Austria Helen Petrie, UK Michael Pieper, Germany Anthony Savidis, Greece Andrew Sears, USA Christian Stary, Austria

Hirotada Ueda, Japan Jean Vanderdonckt, Belgium Gregg C. Vanderheiden, USA Gerhard Weber, Germany Harald Weber, Germany Panayiotis Zaphiris, Cyprus

Virtual and Mixed Reality Program Chair: Randall Shumaker Pat Banerjee, USA Mark Billinghurst, New Zealand Charles E. Hughes, USA Simon Julier, UK David Kaber, USA Hirokazu Kato, Japan Robert S. Kennedy, USA Young J. Kim, Korea Ben Lawson, USA Gordon McK Mair, UK

David Pratt, UK Albert “Skip” Rizzo, USA Lawrence Rosenblum, USA Jose San Martin, Spain Dieter Schmalstieg, Austria Dylan Schmorrow, USA Kay Stanney, USA Janet Weisenford, USA Mark Wiederhold, USA

Internationalization, Design and Global Development Program Chair: P.L. Patrick Rau Michael L. Best, USA Alan Chan, Hong Kong Lin-Lin Chen, Taiwan Andy M. Dearden, UK Susan M. Dray, USA Henry Been-Lirn Duh, Singapore Vanessa Evers, The Netherlands Paul Fu, USA Emilie Gould, USA Sung H. Han, Korea Veikko Ikonen, Finland Toshikazu Kato, Japan Esin Kiris, USA Apala Lahiri Chavan, India

James R. Lewis, USA James J.W. Lin, USA Rungtai Lin, Taiwan Zhengjie Liu, P.R. China Aaron Marcus, USA Allen E. Milewski, USA Katsuhiko Ogawa, Japan Oguzhan Ozcan, Turkey Girish Prabhu, India Kerstin R¨ ose, Germany Supriya Singh, Australia Alvin W. Yeo, Malaysia Hsiu-Ping Yueh, Taiwan

XI

XII

Organization

Online Communities and Social Computing Program Chairs: A. Ant Ozok, Panayiotis Zaphiris Chadia N. Abras, USA Chee Siang Ang, UK Peter Day, UK Fiorella De Cindio, Italy Heidi Feng, USA Anita Komlodi, USA Piet A.M. Kommers, The Netherlands Andrew Laghos, Cyprus Stefanie Lindstaedt, Austria Gabriele Meiselwitz, USA Hideyuki Nakanishi, Japan

Anthony F. Norcio, USA Ulrike Pfeil, UK Elaine M. Raybourn, USA Douglas Schuler, USA Gilson Schwartz, Brazil Laura Slaughter, Norway Sergei Stafeev, Russia Asimina Vasalou, UK June Wei, USA Haibin Zhu, Canada

Augmented Cognition Program Chairs: Dylan D. Schmorrow, Cali M. Fidopiastis Monique Beaudoin, USA Chris Berka, USA Joseph Cohn, USA Martha E. Crosby, USA Julie Drexler, USA Ivy Estabrooke, USA Chris Forsythe, USA Wai Tat Fu, USA Marc Grootjen, The Netherlands Jeﬀerson Grubb, USA Santosh Mathan, USA

Rob Matthews, Australia Dennis McBride, USA Eric Muth, USA Mark A. Neerincx, The Netherlands Denise Nicholson, USA Banu Onaral, USA Kay Stanney, USA Roy Stripling, USA Rob Taylor, UK Karl van Orden, USA

Digital Human Modeling Program Chair: Vincent G. Duﬀy Karim Abdel-Malek, USA Giuseppe Andreoni, Italy Thomas J. Armstrong, USA Norman I. Badler, USA Fethi Calisir, Turkey Daniel Carruth, USA Keith Case, UK Julie Charland, Canada

Yaobin Chen, USA Kathryn Cormican, Ireland Daniel A. DeLaurentis, USA Yingzi Du, USA Okan Ersoy, USA Enda Fallon, Ireland Yan Fu, P.R. China Afzal Godil, USA

Organization

Ravindra Goonetilleke, Hong Kong Anand Gramopadhye, USA Lars Hanson, Sweden Pheng Ann Heng, Hong Kong Bo Hoege, Germany Hongwei Hsiao, USA Tianzi Jiang, P.R. China Nan Kong, USA Steven A. Landry, USA Kang Li, USA Zhizhong Li, P.R. China Tim Marler, USA

XIII

Ahmet F. Ozok, Turkey Srinivas Peeta, USA Sudhakar Rajulu, USA Matthias R¨ otting, Germany Matthew Reed, USA Johan Stahre, Sweden Mao-Jiun Wang, Taiwan Xuguang Wang, France Jingzhou (James) Yang, USA Gulcin Yucel, Turkey Tingshao Zhu, P.R. China

Human-Centered Design Program Chair: Masaaki Kurosu Julio Abascal, Spain Simone Barbosa, Brazil Tomas Berns, Sweden Nigel Bevan, UK Torkil Clemmensen, Denmark Susan M. Dray, USA Vanessa Evers, The Netherlands Xiaolan Fu, P.R. China Yasuhiro Horibe, Japan Jason Huang, P.R. China Minna Isomursu, Finland Timo Jokela, Finland Mitsuhiko Karashima, Japan Tadashi Kobayashi, Japan Seongil Lee, Korea Kee Yong Lim, Singapore

Zhengjie Liu, P.R. China Lo¨ıc Mart´ınez-Normand, Spain Monique Noirhomme-Fraiture, Belgium Philippe Palanque, France Annelise Mark Pejtersen, Denmark Kerstin R¨ ose, Germany Dominique L. Scapin, France Haruhiko Urokohara, Japan Gerrit C. van der Veer, The Netherlands Janet Wesson, South Africa Toshiki Yamaoka, Japan Kazuhiko Yamazaki, Japan Silvia Zimmermann, Switzerland

Design, User Experience, and Usability Program Chair: Aaron Marcus Ronald Baecker, Canada Barbara Ballard, USA Konrad Baumann, Austria Arne Berger, Germany Randolph Bias, USA Jamie Blustein, Canada

Ana Boa-Ventura, USA Lorenzo Cantoni, Switzerland Sameer Chavan, Korea Wei Ding, USA Maximilian Eibl, Germany Zelda Harrison, USA

XIV

Organization

R¨ udiger Heimg¨artner, Germany Brigitte Herrmann, Germany Sabine Kabel-Eckes, USA Kaleem Khan, Canada Jonathan Kies, USA Jon Kolko, USA Helga Letowt-Vorbek, South Africa James Lin, USA Frazer McKimm, Ireland Michael Renner, Switzerland

Christine Ronnewinkel, Germany Elizabeth Rosenzweig, USA Paul Sherman, USA Ben Shneiderman, USA Christian Sturm, Germany Brian Sullivan, USA Jaakko Villa, Finland Michele Visciola, Italy Susan Weinschenk, USA

HCI International 2013

The 15th International Conference on Human–Computer Interaction, HCI International 2013, will be held jointly with the aﬃliated conferences in the summer of 2013. It will cover a broad spectrum of themes related to human–computer interaction (HCI), including theoretical issues, methods, tools, processes and case studies in HCI design, as well as novel interaction techniques, interfaces and applications. The proceedings will be published by Springer. More information about the topics, as well as the venue and dates of the conference, will be announced through the HCI International Conference series website: http://www.hci-international.org/ General Chair Professor Constantine Stephanidis University of Crete and ICS-FORTH Heraklion, Crete, Greece Email: [email protected]

Table of Contents

Part I: On-Line Communities and Intelligent Agents in Education and Research Promoting Reﬂective Learning: The Role of Blogs in the Classroom . . . . Rahayu Ahmad and Wayne G. Lutters

3

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sung-Wook Baek, Jong-Hoon Seo, and Tack-Don Han

12

Gaining Insight into the Application of Second Life in a Computing Course: Students’ Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . James Braman, Alfreda Dudley, Kidd Colt, Giovanni Vincenti, and Yuanqiong Wang

20

Third Agers and Social Networking in Higher Education . . . . . . . . . . . . . . Christopher M. Connolly and Gabriele Meiselwitz

30

A Real-time Disaster Situation Mapping System for University Campuses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kayo Iizuka, Yasuki Iizuka, and Kyoko Yoshida

40

Recent Trends in Software Support for Online Communities for Teaching and Research Projects in Higher Education . . . . . . . . . . . . . . . . . Daniel Kadenbach and Carsten Kleiner

50

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments: Recommendations for Educators and Wiki Designers . . . . Portia Pusey and Gabriele Meiselwitz

60

Teacher Agents: The Current State, Future Trends, and Many Roles of Intelligent Agents in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kevin Reed and Gabriele Meiselwitz

69

Part II: Blogs, Wikis and Twitters Interpreting User-Generated Content: What Makes a Blog Believeable? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rahayu Ahmad and Wayne G. Lutters Extracting Events from Wikipedia as RDF Triples Linked to Widespread Semantic Web Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Carlo Aliprandi, Francesco Ronzano, Andrea Marchetti, Maurizio Tesconi, and Salvatore Minutoli

81

90

XVIII

Table of Contents

Collaborative Sensemaking during Admin Permission Granting in Wikipedia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Katie Derthick, Patrick Tsao, Travis Kriplean, Alan Borning, Mark Zachry, and David W. McDonald Mining Social Relationships in Micro-blogging Systems . . . . . . . . . . . . . . . Qin Gao, Qu Qu, and Xuhui Zhang Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nirupama Dharmavaram Sreenivasan, Chei Sian Lee, and Dion Hoe-Lian Goh

100

110

120

Impact of Blog Design Features on Blogging Satisfaction: An Impression Management Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wee-Kek Tan and Hock-Hai Teo

130

An Exploratory Study of Navigating Wikipedia Semantically: Model and Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I.-Chin Wu, Yi-Sheng Lin, and Che-Hung Liu

140

Part III: Social Computing in Business and the Enterprise Productivity Improvement by Using Social-Annotations about Design Intent in CAD Modelling Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gerardo Alducin-Quintero, Manuel Contero, Jorge Mart´ın-Guti´errez, David A. Guerra-Zubiaga, and Michael D. Johnson Handshake: A Case Study for Exploring Business Networking for the Enterprise, Inside and Out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laurie E. Damianos, Donna L. Cuomo, and Stan Drozdetski “Your Team Cohesion is Low”: A Systematic Study of the Eﬀects of Social Network Feedback on Mediated Activity . . . . . . . . . . . . . . . . . . . . . . Luciano Gamberini, Francesco Martino, Anna Spagnolli, Roberto Ba` u, and Michela Ferron

153

162

172

E-Business Solutions in the Cable TV Industry . . . . . . . . . . . . . . . . . . . . . . Viorica Harrison and June Wei

182

Measuring the Success of On-line Communities in an Enterprise Networking Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lester J. Holtzblatt and Laurie E. Damianos

188

Business-to-Business Solutions for the Cosmetic Industry . . . . . . . . . . . . . Manami Koblentz and June Wei

197

Table of Contents

XIX

Online Crowdsourcing in the Public Sector: How to Design Open Government Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Giordano Koch, Johann F¨ uller, and Sabine Brunswicker

203

An M-Pill Framework in the Electronic Healthcare . . . . . . . . . . . . . . . . . . . Nien-Chieh Lee, Hi Tran, Albert Yin, and June Wei

213

Online Design Discussion Sites: Emerging Resource for Creative Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moushumi Sharmin and Brian P. Bailey

219

Social Computing for Home Energy Eﬃciency: Technological and Stakeholder Ecosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marc Torrent-Moreno, Rafael Gim´enez, Regina Enrich, Jos´e Javier Garc´ıa, and Mar´ıa P´erez

229

Part IV: Social Computing in Everyday Life Conversational Lives: Visualizing Interpersonal Online Social Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Heng Chen and Aisling Kelliher

241

CICRO: An Interactive Visual Interface for Crowd Communication Online . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Masao Ohira, Hitoshi Masaki, and Ken-ichi Matsumoto

251

Features to Support Persistent Chat Conversation . . . . . . . . . . . . . . . . . . . . Mohsen Rahimifar and Siti Salwah Salim

261

Design and Evaluation of an Online Social Support Application for Family Caregivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthieu Tixier and Myriam Lewkowicz

267

An Investigation into the Social Network between Three Generations in a Household: Bridging the Interrogational Gaps between the Senior and the Youth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tsai-Hsuan Tsai, Yi-Lun Ho, and Kevin C. Tseng

277

Preschoolers as Video Gamers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexander Voiskounsky

287

GAZO GAZO KUN: Photo-Sharing System Using an Anthropomorphic Photo Frame for Communication Support . . . . . . . . . . . . . . . . . . . . . . . . . . . Takashi Yoshino and Tomoya Matsuo

297

XX

Table of Contents

Part V: Information Management in Social Computing Is Your Social Networking Privacy Reliant on Intuitive Interfaces? . . . . . Deborah S. Carstens and Veronica Giguere

309

A Study on Social Network Services Visualization Based on User Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Young Suk Han, Jong Kyu Choi, and Yong Gu Ji

319

A Virtual World Prototype for Interacting with a Music Collection . . . . . Jukka Holm and Arto Lehtiniemi

326

A Supervised Machine Learning Link Prediction Approach for Tag Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Manisha Pujari and Rushed Kanawati

336

Beyond the Social Search: Personalizing the Semantic Search in Social Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J´ ulio Cesar dos Reis, Rodrigo Bonacin, and M. Cec´ılia C. Baranauskas

345

Factors Inﬂuencing Online Social Interactions . . . . . . . . . . . . . . . . . . . . . . . . Shankar Ravi

355

Eﬀects of a Synchronized Scoring Interface on Musical Quality . . . . . . . . . Yuji Takai, Masao Ohira, and Ken-ichi Matsumoto

363

Measurement of Tagging Behavior Diﬀerences . . . . . . . . . . . . . . . . . . . . . . . Li-Chen Tsai, Sheue-Ling Hwang, and Kuo-Hao Tang

373

Community Search: A Collaborative Searching Web Application with a User Ranking System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Papon Yongpisanpop, Masao Ohira, and Ken-ichi Matsumoto

378

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

387

Promoting Reflective Learning: The Role of Blogs in the Classroom Rahayu Ahmad and Wayne G. Lutters Department of Information Systems, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA {arahayu1,lutters}@umbc.edu

Abstract. The enthusiasm for adopting social media technologies should be tempered by a critical, empirical understanding of how they facilitate an effective learning envionrment. This study analyzed blog use in two graduatelevel university courses, specifically identifying reflective learning markers in 279 blog entries. This analysis was deepened with follow-up interviews of several top bloggers. The reflective bloggers were characterized as being comfortable with their audience, having a deeper thought process, enjoying the interaction, and progressing well throughout the semester. Our results suggest guidlines for effective use of blogging in the classroom to support reflective learning.

1 Introduction Web 2.0 technologies like web blogs (blogs) and wikis have received great interest as a means of innovation in the classroom. The enthusiasm for adopting these technologies should be tempered by a critical, empirical understanding of how they facilitate a conducive learning environment. Ideally, students should take active roles in constructing new knowledge and have skills to critically examine their assumptions and beliefs. These are known as reflective learning skills, which have been recognized as a essential component of life long learning [6]. Some researchers have claimed that blogs are “effective at supporting reflective learning more so than other technologies” [26, p. 1656]. Despite the attractiveness of these arguments, many of the studies claiming the appropriateness of blogs for reflective learning fail to provide evidence from student’s blog entries to support their assertions [23][27]. There is a dearth of empirical work that can thoughtfully guide educators on the correct use of blogs to support reflective learning. For example, in a study involving 56 students in a teacher education program, only five expressed any interest in incorporating blogs in their own classrooms after experimenting with blogs. The primary reason cited was that they were unable to see how the blogs could facilitate reflective learning [11]. Contributing to our understanding of the impact blogging can have on reflectice learning, this study examines actual students’ blogs entrees for identifying markers of reflective thinking, and does not rely merely on self-report perceptions of blogging. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 3–11, 2011. © Springer-Verlag Berlin Heidelberg 2011

4

R. Ahmad and W.G. Lutters

2 Background 2.1 Reflective Learning Reflective learning has been recognized as one of the most important lifelong learnin skills as it empowers students to wrestle effectively with complex ideas [6]. Reflective learning can be summarized as a process of conscious thought as to what is being learned, which means that the learner is not simply a passive recipient but rather an active contributor within the learning process [22]. Through reflective thinking, the learner becomes intellectually responsible and can make informed, accountable decisions [9]. One means of fostering reflective learning among students is through writing reflective journals. Journaling has been recognized as effective in developing students’ meta cognitive and reflective skills [18]. This approach has been widely adopted in nursing and teacher education as it has been found to be helpful in promoting deeper understanding [15]. Through journaling students become aware of their thought process and the consequences of their experiences. However, students need to be encouraged to be critical in their reflection, in which mere description of events and experiences in journals is not sufficient; students should be able to synthesize the new experience or knowledge and able to relate with previous knowledge, forming a coherent perspective towards an issue or phenomenon [10]. 2.2 Blogs as Tool for Reflective Learning Traditionally hard-copy learning logs have been used to document experiences and reflective thoughts. The rise of blog publishing gives new aspiration for reflection learning. Unlike traditional journals, which only serve the instructor, blogs have the ability to engage a larger audience, forming a close knit community of learners. The public nature of blogs within the class encourages the students to view each other work, hence promoting collaborative writing and reflection among students [1]. Researchers have emphasized the needs for collaborative reflection in enhancing the development of professional knowledge [17]. Furthermore, a blog is interactive in the sense that readers can respond with comments which will be beneficial in promoting a collaborative learning environment [7]. Initial studies showed that blogging can support reflective thinking, critical thinking, and collaboration, allowing students to take ownership of their learning [2][20]. However, there is very limited evidence showing the levels of reflection that are demonstrated in the students’ blogs. Some papers made substantial contributions providing arguments of the appropriateness of blog features for reflective learning, but did not demonstrate the real implementation in a class [12][25]. The majority of these studies relied on students’ self-report data on their perceptions of blog usefulness, not their actual blogging activity or content of their entries [27]. Although there have been a few papers demonstrating initial indicators of reflective learning in blogs, these studies were mainly conducted in a very small classes, ranging from five to twenty students [19][21]. Thus, we still do not know whether similar findings can be reached in a larger population of learners. This study attempts to address call for more in-field studies of the informal nature of blogging to foster reflective learning [1][16][19].

Promoting Reflective Learning: The Role of Blogs in the Classroom

5

3 Study Design This study is based on the blog implementation in two sections of a graduate course within the Information Systems Department at UMBC. This course, called Online Communities, employed a hybrid approach in which there were a face-to-face academic lectures and discussions every week for 3 hours supplemented with significant online activities. Prior to every week’s class, students were given an assignment to be explored and they were required to blog about their experiences. These personal blogs were managed using the Blackboard course content management software’s blogging tool. These posts were private within a “neighborhood,” and accessible only by participants and instructional staff for the two classes (and researchers after the conclusion of the semester). Basic sets of reflective guideline questions were included in each week’s activity Guided by these questions students were free to explore the assignments and be reflective in their blogs. These blogs were graded and contributed to student’s overall course grade. In total there were 279 distinctive blog entries posted by 31 students across the two classes. The majority of the students had not created a blog before, and none had ever been asked to produce a blog for a course assessment.

4 Analysis We conducted a content analysis on the weekly blog posts to classify them into reflective and non-reflective posts. We used a deductive approach in analyzng the data, guided by an established reflective coding framework by Kember et al. [13]. We used the following definition of reflective learning to guide our coding process: Reflective learning is the process of internally examining and exploring an issue of concern, triggered by an experience, which creates and clarifies meaning in terms of self, and which results in a changed conceptual perspective [3, pp.100]. We also examined the interaction among the students through the commenting capabilities supported by the Blackboard blog system. Following this analysis, we conducted contextual interviews with seven individuals, talking through their blog entries from the prior semester.

5 Results and Discussion We start off this section by describing the characteristics of the reflective bloggers. This is followed by an examination of their reflective posts. 5.1 The Reflective Bloggers For classifying the reflective and non-reflective bloggers, we used the total number of entrees classified as reflective, using the method desribed above, as our indicator. Students who had more than 6 reflective posts (out of 91) were classified as reflective 1

For this study, we excluded the blog entries from weeks 1 and 13 as they were largely introductory. Posts from weeks 10, 11 and 12 were also excluded as the focused on structured debates.

6

R. Ahmad and W.G. Lutters

bloggers. In total, there were 8 (26%) students classified as reflective bloggers. Among the non reflective bloggers, 12 of them had 4 to 5 entrees classified as reflective. The rest of 11 students had lower than 3 entriess classified as reflective posts. Figure 1 illustrates the total coded reflective posts for each student. Comfortable with Peer Audience. In our interviews with the top reflective bloggers, they explained how they were comfortable in blogging their thoughts and experiences. They enjoyed blogging for known audience, in this case, their peers. Blogging in a close knit community that shared the same purpose motivated them to be more reflective: “I am comfortable blogging in that environment because it was a small community and I knew the people there.” Josh2 “The audience was interested in this stuff from the getgo, so that makes it easier.” Jay

Fig. 1. Total Reflective Posts by Student

Deeper Thoughts. The students appreciated the process of blogging which motivates them to critically review their learning experience [8]. Being required to blog encouraged them to systematically think about their experiences: “It does help my thinking process. In thinking about posting, it triggered some thoughts that I wouldn't have thought about or looked over if I was just thinking it and not writing it down.” Josh “I think you always learn more and get more out of an experience if you have to explain it or share it with others.” Sara

2

Our stylistic convention is to identify quotations from the interviews in italics and direct excerpts from student blog entries in Courier font.

Promoting Reflective Learning: The Role of Blogs in the Classroom

7

“Dr. Z would give us certain things to do each week and since we had to blog about it, it allowed us to really think about our experiences and what we were observing rather than just going through the motions.” Tom Interaction is Important. In reference to Stahl’s social theory of computer-supported collaborative learning [22], learning is argued as more a process of knowledge creation process through interactions or conversations. In blogs, conversations can take place through commenting features which allows for personalized feedback and allow subsequent referencing and revisions. All our reflective bloggers were enthusiastic in receiving comments and feedback from their peers especially constructive feedback that allow extended discussions in their blogs as described: “Yeah. I recall really looking forward to when people would ask engaging questions within the comments. I always enjoy a compliment, but it was fun to "extend" discussion into the comments” Jay “If someone just posted "I agree" that's not as much value. I prefer feedback that contains content and adds value to the topic.” Josh Positive Evolution. Students thought their blogs were progressing well when they learned to understand more about the audience and the dynamics of the class mainly from the interactions within blogs. The topic of the assignment plays significant role, in which the latter assignments allow them to be more personalized in writing something they can relate more and having real experiences: “It's gotten longer and I'm a bit more comfortable voicing more opinions after spending time with the other classmates.” Sara “There was a feeling of "improvement," for sure. At the start I didn't really bother with a lot of style (as I can recall), but I like to think I was a lot more free. I got more comfortable asking research questions, adding formatted text, pictures. I also like to think I got better with the research questions.” Jay 5.2 Reflective Posts As demonstrated by Figure 2 there was no uniform distribution of reflective posts although there was a slight positive trend toward more reflective posts over the course of the semester. Most students were reflective in their week 2 entrees followed by week 10 entrees. We will futher describe the potential factors that influenced their reflective process. Guiding Questions. We argued that blogs alone might not be sufficient for promoting reflective thinking. We echoed Chen’s [5] recommendation that students should be aware that reflection should be demonstrated in their blog posts. Students need to be provided with a common ground topic or issue to initiate their reflective writing process [16]. Anecdotal evidence in previous studies suggested that students faced difficulty writing an online diary whose inputs they were (relatively) free to determine [8][21]. Among the problems faced by the students were difficulty in finding the topics to talk about and attracting others to talk on the particular topic. Our students shared the importance of the guided questions as quoted:

8

R. Ahmad and W.G. Lutters

Fig. 2. Percentage of Reflective Posts by Week

I think one of my main issues with blogging is having something to blog about. So when we had assignments and topics that my fellow classmates might be interested in reading about, it was fairly easy to blog. Tom The exercises were very helpful since it gave us a topic to discuss and for people like me who haven't blogged, it was a starting point. Sara Most of the time, especially in second week, the guiding questions stimulated the students to be more reflective in their blog posts. The instructor would supplement the assignments with guiding reflective questions for example: We have all had experiences with online communities and social computing over the years. Think back over your history and provide a retrospective on your blog.In which communities have you participated? What tools have you used? Think critically.What has changed over time? What has remained the same? What general questions arise from these specific experiences? In respond to these guiding reflective questions, students were able to reflect on own their experience and formed perspective on the evolution of soial computing. Following is one exceprt that demonstrates the kind of reflective response: It is interesting to look at how social computing has evolved over time. The core method of communication (text line chat) hasn’t changed much, the environment in which the communication occurs has. Individual representation has evolved from a screen name, to a buddy icon, to 3D avatars. It also appears that there is a wider age gap in the social computing scene. Ben Similarly in another activity from week 8, the guiding questions required users to explore behind the scenes of the Wikipedia editing process and provoked them to think on the stability and reliability of Wikipedia. This resulted numerous reflective posts such as:

Promoting Reflective Learning: The Role of Blogs in the Classroom

9

I was skeptical of Wikipedia.org at first because of the validity of information, because Wikipedia.org was not a credible source and if you think logically it is not a credible source until information in it is cited with a credible source.I got over my paranoia and became more of an avid user. Overall the user base is what Wikipedia.org's strength is people; who have passion for information and not only information but correct information. User's put some good quality time in their posts and are passionate for sharing information. There is definitely an overly more positive user base then negative which helps Wikipedia.org to remain accurate and free of ambiguity. Matthew Immersion and Learning. We proposed that the type of learning activities may influence the reflective learning process. In week 10, for example, the students were required to blog on their experiences in massive multiplayer online role playing games. Students were able to articulate and analyze the components for an engaging virtual world as demonstrated by this excerpt: While both Secondlife and MMORPGs have social interaction, they are very different. The MMORPG games give users a common starting point for communication. They are all in the game trying to complete a set of goals. These games also have guilds which are tiny communities within the game that players belong. I believe these guilds give the user a sense of belonging and therefore foster more in game communication/interaction. In my experience, Second Life is too open and doesn't give players a sense of belonging Ben The immersion in virtual worlds triggered some of the deepest reflective moments of the course which led to individualized knowledge construction demonstrated by most of the students.

6 Recommendations This study examined the role of blogs in promoting reflective learning. Some practical guidelines emerged that would be useful for anyone seeking to incorporate blogs for this purpose or to fine-tune their existing implementations. • Define and bound the audience. The greatest utility came from having the blogs open to more than just the instructor, but limited to the students enrolled in the class. By knowing their audience students were able to tailor their blog entries and felt more comfortable than posting for world-wide readers on the Internet. • Provide thoughtful prompts. The exercises and guiding questions were essential to provide a common ground among the students and make reflection a clear goal of the learning experience. • Encourage active peer feedback. While the blog entries were evaluated by the instructor, the most valuable feedback came from the peer discussions that emerged in the comments surrounding student posts. • Make blogging a routine, ongoing activity. It is critical for students to be able to see development in their own thinking and that of their peers. This is not possible with a one-off blogging activity. Further research is required to understand the effective design of reflective prompts and peer commenting processes.

10

R. Ahmad and W.G. Lutters

References 1. Armstrong, L., Berry, M., Lamshed, R.: Reflective Learning and Blogs. In: International Conference on Computers in Education (2004) 2. Bouldin, A.S., Holmes, E.R., Fortenberry, M.L.: Blogging About Course Concepts: Using Technology For Reflective Journaling In A Communications Class. American Journal of Pharmaceutical Education 70(4) (2006) 3. Boyd, E.M., Fales, A.W.: Reflecctive Learning: Key To Learning From Experience. Journal of Humanistic Psychology 23(2), 99–117 (1983) 4. Chen, H.L., Cannon, D., Gabrio, J., Leifer, L., Toye, G., Bailey, T.: Using wikis and weblogs to support reflective learning in an introductory engineering design course. In: Proceedings: American Society for Engineering Education Annual Conference & Exposition, Portland, Oregon (2005) 5. Costa, A.L., Kallik, B.: Getting Into The Habit Of Reflection. Educational Leadership 57(7) (2000) 6. Dickey, M.D.: The Impact Of Web-Logs (Blogs) On Student Perceptions Of Isolation And Alienation In A Web-Based Distance-Learning Environment. Open Learning 19(3), 279– 291 (2004) 7. Du, H.S., Wagner, C.: Learning With Weblogs: An Empirical Investigation. In: Proceedings: 38th Hawaii International Conference on System Science, Hawaii (2005) 8. Dewey, J.: How we think. DC Heath, New York (1993) 9. Fulwiler, T.: The Journal Book. Boynton/Cook, Portsmouth (1987) 10. Hernández-Ramos, P.: Web Logs And Online Discussions As Tools To Promote Reflective Practice. The Journal of Interactive Online Learning 3(1) (2004) 11. Ke, F., Xie, Y.: Blogging for Reflective Learning in an Introductory Political Science Course. In: 29th Annual Proceeding of the Accociations for Educational Communications and Technology (2006) 12. Kember, D., Jones, A., Loke, A., McKay, J., Sinclair, K., Tse, H., et al.: Determining The Level Of Reflective Thinking From Students’ Written Journals Using A Coding Scheme Based On The Work Of Mezirow. International Journal of Lifelong Education 18(1), 18– 30 (1999) 13. Knapp, C.E.: Lasting lessons: A Teacher’s Guide To Reflecting On Experience. ERIC Clearinghouse on Rural Education and Small Schools, Charleston (1993) 14. Loo, R., Thorpe, K.: Using Reflective Learning Journals to Improve Individual and Team Performance. Team Performance Management: An International Journal 8(5/6), 134–139 (2002) 15. Macduff, I.: Using Blogs as a Teaching Tool. Negotiation Journal, 107–124 (2009) 16. Manouchehri, A.: Developing Teaching Knowledge Through Peer Discourse. Teaching and Teacher Education 18, 715–737 (2002) 17. Morrison, K.: Developing Reflective Practice In Higher Degree Students Through A Learning Journal. Studies in Higher Education 21(3), 317–332 (1996) 18. O’Keeffe, M., Arnedillo-Sánchez, I., Flanagan, A.: An Investigation into the Use Of Weblogs For Reflection In Learning. International Conference e-Society (2006) 19. Oravec, J.: Blending By Blogging: Weblogs In Blended Learning Environments. Journal of Educational Media 28, 225–233 (2003) 20. Paulus, T.M., Payne, R., Jahns, L.: Am I Making Sense Here?: What Blogging Reveals About Undergraduate student Understanding. Journal of Interactive Online Learning 8(1) (2009)

Promoting Reflective Learning: The Role of Blogs in the Classroom

11

21. Stahl, G.: Building Collaborative Knowing: Elements Of A Social Theory Of Learning. In: Strijbos, J.W., Kirschner, P., Martens, R. (eds.) What we know about CSCL in higher education. Kluwer, Amsterdam (2003) 22. Stiler, G., Philleo, T.: Blogging And Blogspots: An Alternative Format For Encouraging Reflective Practice Among Preservice Teachers. Education 123(4), 789 (2003) 23. Thorpe, K.: Reflective Learning Journals: From Concept To Practice. Reflective Practice 5(93), 327–343 (2004) 24. Wagner, C.: Put Another (B)Log On The Wire: Publishing Learning Logs As Weblogs. Journal of Information Systems Education 14(2), 131–132 (2003) 25. West, R.E., Wright, G.A., Graham, C.R.: Blogs, Wikis, And Aggregators: A New Vocabulary For Promoting Reflection And Collaboration In A Preservice Technology Integration Course. In: Crawford, C., et al. (eds.) Proceedings of the Society for Information Technology and Teacher Education, pp. 1653–1658. Association for the Advancement of Computing in Education, Chesapeake (2005) 26. Williams, J.B., Jacobs, J.: Exploring the use of blogs as learning spaces in the higher education sector. Australasian Journal of Educational Technology 20(2), 232–247 (2004)

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher Sung-Wook Baek, Jong-Hoon Seo, and Tack-Don Han Dept. of Computer Science, Yonsei University, 134, Seodaemun-Gu, Seoul, 120-749, Republic of Korea {yoshiboarder,jonghoon.seo,hantack}@msl.yonsei.ac.kr

Abstract. In this paper, we focused on a research on a Location-Based Social Network Service to increase social interactions among people who are engaging in research and development. Since existing many of location-based social network services have only provided information of regions and social relations of users have targeted an unspecific majority of groups, they could not have active interactions, compared to users of online communities who have particular use purposes. ConAR what we designed to get over these weaknesses, aims at researches to academically exchange with others and it enables them to have social interactions with people of the same interests by helping them select discussion topics and decide assignment places simply and on impulse. Keywords: Social Computing, Social Interaction, Location-Based Service, Context-Aware Service, Communities, Human Computer Interaction.

1 Introduction Social Networking Service is useful to not only keep friendships with acquaintances, but also make new relationships. Because it is an online service, it is also possible to keep friendships with friends who live far away using this. This service is gradually developing into a variety of forms. For example, Twitter is now used to get news or information useful for life and Academia or Research Gate provides services like theses or projects of researchers for users for an academic exchange. In addition, Location-Based Social Networking Services using location information have been developed with an invention of Smart-phone. In particular, Location-Based Social Networking Service provides not only information of regions like restaurants and transportation, but social networking functions among people who live in the same regions. In spite of these services, however, it is not easy to make social relations with people who have common interests and social networking doesn't work as actively as online community services, because it still aims at an unspecific majority of people. The following scenarios suggest problems of the current Social Networking Services: Sung-Wook in a graduate school has a lot of interests in HCI. He opens research achievements and projects of named researchers in the HCI field and exchanges messages with them using an online Social Networking Service provided A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 12–19, 2011. © Springer-Verlag Berlin Heidelberg 2011

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

13

on Academia. One day, Sung-Wook participated in an academic conference held in COEX, Seoul to make a presentation of his thesis. He wanted to discuss with people who are interested in SNS like him, so tried to find them through his Location-Based Social Networking Service, but it was not easy for him to go forward and speak to them, because only a minority of people had interests in it and he had no idea of their research achievements and participation purposes in the conference. Here, we can extract two main problems from the above examples. Firstly, many of the current social networking services including Academia cannot be practically used for discussion in a conference or office or laboratory. Secondly, they don't provide social networking services for an academic exchange of users. ConAR has three characteristics as following. Firstly, it shows profiles and research performances of researchers around users based on their locations. Here, we designed it to help users search information what they really want by extracting information required for cooperation of researchers through a survey targeting researchers of companies and schools. Secondly, users can do online or offline social interactions, because it provides them with Contextual Information of participation possibility in offline meetings or meeting places where they prefer. Lastly, using 'ConAR', users can choose discussion topics by themselves and decide assignment places simply and promptly.

2 Methods To induce necessary information and functions for ConAR service, we performed a survey regarding 17 professional R&D researchers in graduate school and company. Based on such result, we induced necessary personal information and service to have scholarly exchange, and by using rapid prototyping tool, we designed UX of ConAR. Finally, to estimate the usefulness and usability of ConAR, we realized using Android SDK. For the estimation, we invited 12 researchers, and they belong to graduate school and company R&D center. The reason for limiting job like this is because the purpose of using ConAR is scholarly exchange in academy, school, or company, and actually, as a result of analyzing the job of academy participants, it was classified mainly as student, professor, and company researcher. 12 researchers were divided into 2 groups in 6 people, and those with no daily friendly relationship were chosen to estimate whether social interaction using ConAR is held well. 2.1 Personal Information Regarding personal information related to users, survey was made based on the information opened in Korea's graduate school R&D room, and 17 professionals participated in it. Table 1 is a survey investigation result regarding personal information of researchers needed for scholarly exchange. Through Table 1 we could largely check two kinds of information. Firstly, the answer that Social Networking Service is needed in scholarly exchange was 3.7 in average, so it could be checked that it was positive. Secondly, regarding the most important thing in scholarly exchange, F. Consider the area of research was 4.2 in average and it was the highest, and I. Consider the project was 3.9 in average, and J. Consider the experience of field was 3.8 in average in the greatest order.

14

S.-W. Baek, J.-H. Seo, and T.-D. Han Table 1. The result of survey about researcher's information for academical interaction

Number 1 2

Attribute

Evaluation Tool

Do you consider using SNS for the academic interaction? What kind of researcher's information do you consider when you do the academic interaction?

Negative 1 2 3 4 5 Positive

A

Consider the degree

B

Consider the personality

C

Consider the advisor

D

Consider the paper (Journal)

E

Consider the sex

F

Consider the area of research

G

Consider the social position

H

Consider the lab

I

Consider the project

J

Consider the experience of field

Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive

Fig. 1. Personal Information of ConAR

Average

Response Rate

3.7

93.7%

3.1

100%

3.5

100%

3.5

100%

3.5

100%

1.8

100%

4.2

100%

2.5

100%

3.0

100%

3.9

100%

3.8

100%

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

15

Contrarily, regarding scholarly exchange, what is not important was E. Consider the sex with 1.8 in average which is the lowest, and G. Consider the social position was 2.5 in average which the second lowest. We, through this result, made the Personal Information which will be used in ConAR as in Fig 1. Looking at Fig 1, not only personal information of investigated researcher, but also location of researcher, current status information, and meeting places they liked were also included. Researchers' personal information is data which is very much needed when searching for the people one wants have scholarly exchange through ConAR, and meeting place they like or current status of the self (in the paper, we will write this information as Contextual Information) and location information are data which is needed for meeting to be held in the real world. In the information above, information like Contextual Information can be a problem related to privacy invasion. So, we provide Privacy protection function which can protect location of the self and status information for researchers. In case this function is activated, Contextual Information of the self can be opened, and in case of not activated status, no such Contextual Information is opened. 2.2 Service To analyze necessary service for scholarly exchange, we also held survey regarding service during the survey in the above. We classified the items regarding service into 5 kinds, and we made the survey so that necessary 3 kinds in the greatest order could be answered. Table 2 is the following result. Table 2. The result of survey about service for academical interaction Number

Attribute Researcher management

Preference Service Researcher search

Researcher registration

4

Communication ability

Sending a note

Conversation request

5

Individual information protection function

6 7

3

Conversation rejection

Profile management

Research material management function

Paper search

File sending

Location Based Service function

Share the location

Map service

Researcher management Opening the Community Share my status Cooperation service Recommendation for place of appointment

As shown in Table 2, service which is needed for scholarly exchange is investigated by dividing it into 5 items. For example, the necessary function for researcher management service was preferred in the order of searching researcher, registering researcher, and managing researcher. Based on such result, we composed the service which will be provided in ConAR and realized necessary functions for these services.

16

S.-W. Baek, J.-H. Seo, and T.-D. Han

2.3 Prototyping Based on the investigated survey result above, we designed UX of ConAR in the method of Paper Prototyping. To use ConAR at the first time, user registration course should be held, and the course is expressed in Fig 2.

Fig. 2. User Registration Course Scenario

Because ConAR is a service based on location, personal privacy problem can occur. So, to previously prevent the problem, we composed a scenario as follows. Firstly, in case of using it for the first time, after user registration is held, Privacy function can be checked, and if it is Off, it will show an alert message so that no one would subconsciously open location information and status information. Even if a user is already registered, before starting ConAR service, the Privacy function is always checked and alert message is shown. Next is the scenario regarding searching researcher, MAP View, registering meeting place which is preferred, researcher management, personal profile management.

Fig. 3. ConAR's Scenario regarding Service

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

17

To express each 5 kind of ConAR service in mobile environment, we designed a scenario as in Fig 3. Using tab controller, we designed User Interface, and explanation regarding each tab as follows. Firstly, in case of Research Tab, it provides searching for researcher to have scholarly exchange. For example, it shows related researcher by searching Paper, Project, or studying field. In case of Map Tab, it shows researchers around using ConAR service through Google Map. If Privacy mode in the Off state, it is not signed in the Map. In case of Place Tab, it searches preferred meeting place and provides registration function. Meeting place such as library, coffe shop, vending machine, and etc can be chosen, and based on the current location of the self, it searches the location by searching close place. In case of Coworker Tab, it shows the list of people who went through scholarly exchange (for example, when giving and taking message was held or conversation was held by real meeting) through ConAR. Also, if researcher is chosen, functions which can make scholarly exchange are provided. Finally, in case of Profile Tab, personal information of the self can be changed, and also, On/Off setting regarding Privacy can be done. 2.4 Procedure In order to assess whether ConAR is helpful for an academic exchange of researchers, we conducted an experiment by dividing 12 participants into Group A and B consisting of six members separately. First of all, A group was organized with 4 of introvert people and 2 of active people, and it was observed if there was a social interaction among researchers by setting them to join a meeting actually in Contextual Information. On the contrary, B group was composed of 2 of introvert people and 4 of active people, and it was observed as well if there was a social interaction among them by setting them to be busy in Contextual Information.

Fig. 4. Experimenter estimating usefulness and usability of ConAR service

2.5 Results Fig 5 is a task process of ConAR that participants carried out. Although, Group A and B had equal processes in contacting researchers who they wanted to work with, using the Matching service of ConAR, the places for their social interaction were different between Group A and B depending on Contextual Information of researchers. For A group that was set to join the actual meeting in Contextual Information, participants

18

S.-W. Baek, J.-H. Seo, and T.-D. Han

Virtual World

Virtual World B Group Searching the researcher

Get recommendation of places for the appointment

Real World A Group

Check the users current status

Fig. 5. Social interactions in the virtual world and the real world

had 5 total meetings actually out of 8 attempts, but for B group that was set to be busy in Contextual Information had 9 attempts, they had a total of 8 social interactions using a message delivery function online.

3 Conclusion and Future Work We made participants have an academic exchange by carrying out a research on a location-based social networking service. ConAR connected researchers to other people who have the same interests to them through the Matching Function and made their academic exchange work smoothly in a conference or school or office by simplifying the process of the actual meeting. We also devised a social interaction considering the current conditions of researchers using Contextual Information. Lastly, we provided a location-based social networking service which is very useful for researchers based on online community, instant message delivery and User Management functions to keep their academic exchange continual. In addition to this experiment, we will carry out a study on improvement of cooperation of researchers in the actual meeting. Although, existing Matching services have been developed to enable researchers that are motivated to meet the opponents, to cooperate with each other in the actual meeting, other services should be developed for an online social interaction. Therefore, we will develop services providing functions like brain storming, voice recording or video filming and white board used in discussion places.

References 1. Wang, F.-Y., Zeng, D., Carley, K.M., Mao, W.: Social Computing: From Social Informatics to Social Intelligence. IEEE Computer Society, Los Alamitos (2007) 2. Yamamoto, D., Takumi, I., Matsuo, H.: Location-Based Social Network Services Employing Student Cards for University. ACMGIS (2009)

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

19

3. You, C.-W., Chen, Y.-L., Cheng, W.-H., Chen, M.-S., Sinica, A., Tsai, S.-A.: SocialCRC: A Social-and Context-Aware Rendezvous Coordination System. In: SIGCHI 2010 (2010) 4. Cummings, J.N., Butler, B., Kraut, R.: The Quality Of Online Social Relationships. Communications of the ACM (2002) 5. Zhao, D., Rosson, M.B.: How and Why People Twitter: The Role that Micro-blogging Plays in Informal Communication at Work. In: ACM 2009 International Conference on Supporting Group Work (2009) 6. Karimi, H.A., Zimmerman, B., Ozcelik, A., Roongpiboonsopit, D.: SoNavNet: A Framework for Social Navigation Networks. In: ACMGIS 2009 (2009)

Gaining Insight into the Application of Second Life in a Computing Course: Students’ Perspectives James Braman1, Alfreda Dudley1, Kidd Colt2, Giovanni Vincenti1, and Yuanqiong Wang1 1 Towson University Dept. of Computer and Information Sciences 8000 York Rd. Towson, MD, USA {jbraman,adudley,gvincenti,ywang}@towson.edu 2 Independent Consultant Sarasota, Florida, USA [email protected]

Abstract. As educational delivery methods have changed over the years, often through enhancement by technology, new and innovative strategies have been introduced to enhance learning. Virtual worlds such as Second Life are being used as an educational tool in many domains and for a variety of purposes. To gain a further understanding of the use of virtual worlds in the classroom from a student’s perspective, the authors discuss the feedback gained from students about their participation utilizing Second Life in a computer course. Pre and post surveys were distributed to the class to get student’s feedback. The responses from these surveys are reported in this paper. Keywords: Virtual Worlds, Second Life, Education, Perception, Educational Tools.

1 Introduction As virtual worlds and their technological counterparts (i.e., video games, interactive multimedia and other forms of virtual and augmented realities) become more common place, they can be used for much more than entertainment purposes alone. These technologies can be used to extend the capabilities of the traditional classroom to create a more immersive and interactive educational environment. As we begin to increase the use of virtual worlds as part of instruction, it is important to assess the perception of effectiveness as a learning tool by the students. Perception of a tool such as virtual worlds can often affect its use and adoption in an educational context [1]. The application of this particular technology in the classroom can have an impact on how students and educators react to it, and their willingness to use it for courserelated activities. In this paper, the authors outline their preliminary work and discuss feedback collected from students regarding their participation in Second Life© in a computer course. Second Life (SL), like many other virtual worlds, is a 3-dimensional internetbased interactive space where users can communicate in real-time through visual A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 20–29, 2011. © Springer-Verlag Berlin Heidelberg 2011

Gaining Insight into the Application of Second Life in a Computing Course

21

representations of themselves, called avatars. It is through these avatars that users can build objects, socialize, interact within the environment and “view” the world. Avatars can be created in many different forms and representations, allowing for a diverse variety of visual representations [2]. Unlike many variants of virtual reality, Second Life does not require expensive gear to interact within its environment; one can use the platform by using traditional hardware (i.e. keyboard, mouse, monitor, headsets) increasing access to average users. Students can use Second life in the classroom or on their home computer for class activities or at-home projects. The content within Second Life is almost entirely created by the users, where user creativity is encouraged. Built-in modeling tools can be used to create detailed objects and shapes enhancing many ideas and concepts. One can even use many different types of external programs like Blender, Plopp or 3D Studio Max, to create textures and elaborate graphics to enhance designs. Virtual goods and services can also be bought and sold using Second Life’s virtual currency called the Linden dollar which can be exchanged for real US currency on the Lindex Market exchange, enabling users to link real accounts to their virtual accounts. Second Life and similar platforms have often been referred to as a “3D-Web” and pointed out as a possibility of the future of online interaction. Teaching inside a virtual world not only provides students the means to be immersed in real-time scenarios that can be customized, but also allocates to them opportunities to interact with an online society that emerged from the technology they study in classes. Virtual classrooms are engaging to students as “Students engaged in educational games and simulations are interpreting, analyzing, discovering, evaluating, acting, and problem solving” [3, p. 116]. These virtual spaces provide a means to model complex systems and facilitate role-play simulations at a low cost [4]. Second Life has been used in many educational contexts across various disciplines in order to support and enhance learning [5]. In previous studies, students generally agreed that they did see potential educational benefits of using Second Life to enhance learning [6]. This led the authors to conduct further studies on the impact of virtual worlds/reality in educational contexts. In a related study, the professors used Second Life as a tool in several computer courses to teach students about decision making through virtual role playing activities with encouraging results [7]. In addition, Second Life was used to demonstrate the use of virtual worlds in real life case scenarios [8]. The authors wanted to gain insight on students’ responses on virtual worlds to begin investigations revealing concerns students have about using these technologies in the classroom. In the next section the surveys are discussed.

2 Methods In order to begin to assess students’ perceptions related to potential concerns of course related use of Second Life, two surveys were administered to students enrolled in a course titled Computers and Creativity during the Fall of 2009. The main goal of this course is to teach students basic computer skills while emphasizing multimedia and the creative nature of technology (i.e. flash animation, web design and basics of multimedia applications). The general structure of the class contains both lab and lecture components. The course content and creative goals correlate with the

22

J. Braman et al.

expressive power and creative nature of Second Life. Activities for this class included in-world student projects, virtual field trips, basic building and scripting. Second Life is also the basis for many class discussions. Second Life based Machinima (filming in 3D worlds) has also been incorporated into class discussion topics to illustrate filming techniques and video editing [9]. The typical audience of this course includes first and second year students from all fields of study. Three sections of the Computers and Creativity course were given the pre and post surveys, with a total enrollment of 92 students. The purpose of pre-Second Life survey was to assess students’ preconceived ideas about virtual worlds. At this point in the class, no information, discussions or labs had been related to virtual worlds or Second life. After this initial survey, the class spent two sessions discussing various aspects of virtual worlds and their impact on the real world. In addition, the class spent two lab sessions in Second Life learning to build objects, visiting areas of the environment and learning the features of the platform. After these Second Life class activities, the post-SL survey was administered to the same class sections to see if there was an overall change in student perception. The questions were the same as in the pre-SL survey, except for a few additional questions that were asked related to avatar appearance and their overall thoughts on Second Life. The pre-Second Life survey consisted of five questions and the post-Second Life survey consisted of nine questions.

3 Pre-Second Life (SL) Survey From the enrolled students, 67 students responded to the survey for a 72.8% response rate. The average age for those responding was 19.8, with a male to female ratio of 28 male (41.8%) and 39 female (58.2%). Students had a variety of backgrounds when it came to declared majors and areas of study. Question #1 asked “Do you have any prior experience with any 3D virtual worlds? (For Instance, Second Life, Sims online, WoW...etc) If so, which ones.” From the responses, 9% answered “Yes” (6) and 91% (61) reported “No”. Those that did note some experience, students noted that they had experience in virtual worlds such as: The Sims, Guildwars, World of WarCraft, Maplestory, IMVU, Gates of Andaron, and through Xbox Live. Question #2 asked “Have you ever taken a class that has used a virtual world as part of classroom instruction or as a learning tool?”. The majority responded “No”, 97% (65) and 3% (2) responded “Yes”. Of the two students who responded yes; one reported that they had participated in a class exercises for a business class in high school and the other noted they had been involved in an activity for an accounting class, but did not mention the actual environment. Question #3 asked “List any ethical concerns you have about Second Life or other virtual worlds”. In order to assess initial thoughts on the subject, no information was given up to this point on these types of issues or anything related to computer ethics. From the respondents, 38 students (56.7%) gave no answer or wrote “no” or “none” on the survey. Table 1 lists student concerns and its respective frequency. Some students responded to more than one item. A post-SL frequency column has also been included to compare pre and post SL results. More detail about the post-SL questionnaire is documented in the later section.

Gaining Insight into the Application of Second Life in a Computing Course

23

Table 1. Concerns about Second Life or other Virtual Worlds Concern

Privacy Concerns Detriment to Real Life Social Skills Identity Theft Social Issues/ Social problems (General) Virtual Crime Security Addiction Copyright infringement Not being taken seriously Online Stalkers Encountering Adult content “It scares me” Encountering Users with Malicious Intent Speaking with Strangers False Appearances General Safety Online

Frequency (Pre) 29 respondents 43.3% 13 7 4 2 2 2 2 2 1 1 1 1 0 0 0 0

Frequency (Post) 22 respondents 31.4% 8 3 2 3 0 1 3 1 1 0 0 0 1 1 1 1

Table 2. General Student Comments on Online Behavior General Student Comments on Online Behavior “Anonymity allows people to do and say anything” “Might be for those shy in RL” [Real Life] “Yes, it’s more informal” “Yes, you can be what you want online” “Maybe because no one knows me” “No, being online is just another way to be yourself” “Possibly I would be more social” “Not real, so I may act out differently” “Maybe because I could do anything without people knowing me”

We asked students in Question # 4 “Do you think your behavior online would be different than your behavior in Real Life? Why”. From the respondents 34 students (50.7%) answered “Yes” while the 23 students (34.3%) said “No” and the remaining 10 (15%) answered “Maybe”. Table 2 above highlights the general comments associated with answers for this question from those who gave a brief explanation. As a last question (Question #5), in order to understand what types of issues the students generally thought were important, five general concerns were listed randomly, which included: “Social Issues”, “Privacy Concerns”, “Network and Security Issues”, “Intellectual Property/ Copyright Issues” , and “Computer Crime Issues”. From the list of choices, students were asked to order these concerns from 1 (highest) to 5 (lowest). Table 3 lists the concerns and what the average ranking score was for all students. All respondents answered this question.

24

J. Braman et al. Table 3. Average Ranking of Listed Concerns for Pre-SL Listed Concern Privacy Concerns Networks and Security Issues Computer Crime Issues Social Issues Intellectual Property/ Copyright Issues

Average Ranking 2.06 2.54 3.27 3.45 3.63

4 Post-Second Life (SL) Survey After these Second Life class activities, a secondary post-SL survey was administered to the same class sections to see if there was an overall change in student perception. The questions were the same as in the pretest, except for a few additional questions that were asked related to avatar appearance and their overall thoughts on Second Life. For the post Second Life questionnaire, nine main questions were asked of the students. There were 70 student respondents yielding a 76% response rate from the total 92 possible students. The Average age from these students was 19.95 years old, with 35 students identified as “Male” (50%) and 35 students as “Female” (50%). Question #1 was in the form of a table, where students could select their choice for each question as presented in Table 4. The results of the student’s responses are summarized in Table 4. Table 4. General Reactions to Second Life for Class Activity Question Would you be interested in using Second Life for a class Lab / activity in the future? Do you think virtual environments could have an educational benefit? Rather use another tool / virtual world other than SL Enjoyed using SL in class Intend to use SL in the future (outside of class)

YES

NO

Maybe

41

15

14

37

0

29

8 59 10

31 5 34

31 6 26

Table 5. Avatar Appearance Question How important is your avatar's appearance? How important is the appearance of your friend's avatar? How important is the appearance of a stranger's avatar? How important is the appearance of a significant others/ partner' avatar?

Very Important 4

Important 15

Somewhat Important 32

Not Important 19

1

3

35

31

1

7

30

32

3

14

32

21

Gaining Insight into the Application of Second Life in a Computing Course

25

Question #2 was also presented in the form of a table where students could select the appropriate response. The authors were interested in the perceptions students had on the appearances of other avatars. Specifically, we wanted to know if students felt threatened by the appearances of anonymous users, or if they had an affinity towards higher quality avatars for making friends [10]. Additional research is needed in this area. Table 5 presents the results of the general questions of the importance of appearance. In Question #3, students were also asked “Do you think you could feel threatened by the appearance of another avatar?” 9 students responded “Yes” (13%), 53 responded “No” (76%) and the remaining reported “Maybe” (11%). Some students commented on appearance stating that “Some appearance[s] could be offensive” and that appearance could relate to “stereotypes of how people appear”. One student commented that avatar appearance was something they did not feel was important because interaction in SL was “No different than an IM or chat session”. In Question #4, Students were asked about their general feeling of safety while using Second Life. “How would you rate your level of safety while in Second Life? (1-Unsafe/threatened: 10-Very Safe)”. It was mentioned in class about safety concerns of virtual worlds (as with other online environments) not to reveal personal information and also to choose avatar names that were dissimilar to their real-life name and not to choose words that could identify them or their location. When asked to rate their level of safety on a scale of 1 to 10 (1-Unsafe/threatened: 10-Very Safe), from the 70 respondents the average rating score was a 7.74. In Question #5, students were asked to list any difficulties they had with the SL platform. Many students left this question blank. For those that did have difficultly, Table 6 describes the problems noted. A frequency is noted for difficulties mentioned in some cases. These described difficulties are similar to previous studies [6]. Question #6 asked students “What did you like the most about Second Life?”. These items and frequency of responses are described in Table 7 below. Students were then asked in Question #7 what they generally disliked about Second Life. These items and frequency of responses are described in Table 8 below. Students were then asked in Question #7 what they generally disliked about Second Life. These items and frequency of responses are described in Table 8 below. Table 6. Difficulties in Second Life Difficulty Learning the controls Confusing for new users at first Appearance difficult to change/edit Didn’t know where to explore Flying Building things Needed more time to learn Teleporting Difficulty getting the concept Lag Awkward to control

Frequency 9 6 6 3 2 2 1 1 1 1 1

26

J. Braman et al. Table 7. Liked about Second Life General Comments Flying / Ability to Fly Being able to visit/ explore new places It was a different/ unique experience Freedom to be anyone / do anything Social aspects Graphics Fun Editing appearance Other avatar’s appearances Building Making a avatar Easy to use It’s more fun than being in a real class Whole other world online Like a game but with more interaction

Frequency 11 10 8 8 6 4 3 1 1 1 1 1 1 1 1

Table 8. Disliked about Second Life General Comments Controls are difficult Not into virtual worlds Lag Trying to figure out where to go / explore Building Too much to understand Hard to focus on the task I was bored Graphics Rather create things in real life Security could be an issue Different feel safe or protected Creating an avatar/account Movement Complicated Hard to understand Wanted more time in class Everyone in class can see what you are doing Seems overwhelming Couldn’t change my clothes Rude people Characters look unreal “It scared me a little bit”

Frequency 5 4 3 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

The authors again wanted to see what types of concerns students had about Second Life or using virtual worlds and posed it as another question in the post-SL survey. Question #8 asked “List any ethical concerns you have about Second Life or other virtual worlds:” The results from the pre-SL questionnaire and post-SL questionnaire were different. To compare both results refer to Table 1.

Gaining Insight into the Application of Second Life in a Computing Course

27

In Question #9, the authors wanted to see what types of concerns from the same list as mentioned previously in question #5 in the pre-SL survey, was most important to the students after using Second Life. Again the same five general issues were listed randomly. Students were asked to order these concerns from 1 (highest) to 5 (lowest). Table 9 lists the concerns and what the average ranking score was for all students. All respondents answered this question. The only main change in the average ranking of these general concerns regarding differences were between “Social Issues” and “Intellectual Property/ Copyright Issues”. Previously Social Issues were initially ranked higher, but in the second assessment, it was listed as the least issue of concern. The other issues remained in the same order. The importance of students’ perceived concerns generally remained the same before and after interacting within Second Life. Table 3 compares the results from pre and post-SL questionnaire. Table 9. Average Ranking of Listed Concerns for Post-SL Listed Concern Privacy Concerns Network and Security Issues Computer Crime Issues Intellectual Property/ Copyright Issues Social Issues

Frequency 2.11 2.97 3.09 3.16 3.66

5 Summary of Survey Results As indicated in the above reported results, the feedback from the pre and post survey questionnaires provided basic insight on the impact of using Second Life as an instructional tool. In addition, the responses of the students indicated their perceptions on the application and use of this software. The purpose of this study was for instructors to gain insight on how effective the inclusion of Second Life technology impacted their computing course. However, the instructors were aware of some limitations surrounding this study. For example, the questions on the pre and post surveys can be more in-depth to cover more pedagogical issues; inclusion of analyses and testing of the data would have yielded stronger results; and a reporting of larger population (different types of computing courses) may have also yielded interesting results. In future investigations the authors propose by means of a pre and post SL survey to compare two separate groups of students with differences in class activities to see if there is a significant difference in responses. A major strength of this study is the feedback from students’ perception of using this technology. It is the opinions of the authors that while this study has limitations, it adds value for further examination in this area. The feedback did yield interesting results which demonstrated that there were concerns related to the perception of virtual words as reported in Table 1. These main concerns were still present after interacting with Second Life, but reporting of these concerns was much less in the post-SL survey. In addition, as outlined by Table 3 and Table 9, the ranking of importance of the randomly selected issues, remained similar with both the pre and post surveys. Using this preliminary data as a starting point, we plan to conduct more in-depth studies on concerns related to virtual world usage.

28

J. Braman et al.

6 Conclusions As technology-enhanced learning increases in the classroom coupled with virtual worlds and other computer mediated environments, it is important to gain insight on the best practices using these types of mediums for the benefit of both students and educators. Cultivating on our experiences with implementing Second Life activities in the classroom from several types of activities has encouraged experimentation and potential improvement with these types of technologies. Not only has feedback from students been generally positive, usage in the classroom has generally been encouraging as students are interested in being engaged in learning from within a 3D space. Using the results of this study as a preliminary baseline on the inclusion of Second Life in classroom instruction, we can begin to enhance our in-world activities by trying to overcome potential barriers; such as, misconceptions and concerns or technical / usability problems. The feedback from this study suggests that there are still many concerns that are held by students about interacting in a 3D world and it may take some time to dispel some of the concerns. While there is still much needed research in this area, from both a student and instructor point of view regarding overall successfulness in long term studies. Many of the questions posed in our surveys simply serve as a starting point for other research initiatives by the authors, as it is our intention to explore other areas of education and social interaction through virtual worlds. It is out hope that other educators will explore virtual worlds.

References 1. Laughlin, D.: Overcoming Objections to MUVEs in Education. In: Vincenti, G., Braman, J. (eds.) Teaching through Multi-User Virtual Environments: Applying Dynamic Elements to the Modern Classroom. Information Science Reference, Hershey (2010) 2. McArthur, A.: Real Ethics in a Virtual World. In: Proceedings of CHI 2008 (2008) 3. Antonacci, D.M., Modress, N.: Envisioning the Educational Possibilities of User-Created Virtual Worlds. AACE Journal 16(2), 115–126 (2008) 4. CMP Media LLC.: Second Life Tries for a Second Act The online virtual world has weathered a boom-and-bust cycle. Now it has a New CEO. Can this Social Networking Phenom Attract a New Wave of Consumers and Enterprise Users? (2008), http://www.commonwebnews.com (retrieved) 5. Vincenti, G., Braman, J.: Teaching through Multi-User Virtual Environments: Applying Dynamic Elements to the Modern Classroom. Information Science Reference, Hershey (2010) 6. Wang, Y., Braman, J.: Extending the Classroom through Second Life. Journal of Information Systems Education. Special Issue: Impacts of Web 2.0 and Virtual World Technologies on IS Education 20(2) (2009) 7. Wang, Y., Dudley, A., Braman, J., Vincenti, G.: Simulating Ethical Dilemmas: Teaching Ethics through Immersive Virtual Environments. In: Proceedings of the 12th IASTED International Conference on Computers and Advanced Technology in Education (2009)

Gaining Insight into the Application of Second Life in a Computing Course

29

8. Dudley, A., Braman, J., Wang, Y., Vincenti, G., Tupper, D.: Security, Legal, and Ethical Implications of using Virtual Worlds. In: Proceedings of the 14th World Multi-Conference on Systemics, Cybernetics and Informatics: WMSCI 2010 (2010) 9. Filimon, S.: Machinima. In: Braman, J., Vincenti, G., Trajkovski, G. (eds.) The Handbook of Research on Computational Arts and Creative Informatics, Information Science Reference, Hershey (2009) 10. Yee, N., Bailenson, J.: The Proteus Effect: The Effect of Transformed Self-Representation on Behavior. Human Communication Research 33(3), 271–290 (2007)

Third Agers and Social Networking in Higher Education Christopher M. Connolly and Gabriele Meiselwitz Applied Information Technology Computer and Information Science Towson University 8000 York Road, Towson, MD 410-704-2835 {cconno4,gmeiselwitz}@towson.edu

Abstract. With an aging population and the social networking boom significant research has been performed in three areas. The first is the challenge higher education institutions face integrating social networking sites in class offerings. Students are embracing this medium at an accelerated rate; however, the benefits of social networking for students are not always clear. Secondly, this paper will look into the trend of seniors re-entering institutions of higher education; especially the challenges relating to Web 2.0 technologies such as social networking sites. Although a large number of seniors are online today; many are not comfortable using social networking [20, 14]. Finally, studies relating to the use of social networking sites by third agers outside the classroom will lead to a better understanding of potential challenges seniors and instructors may experience when seniors decide to return to the higher education roles. Keywords: Social networks, computer science education, higher education, third age.

1 Introduction The population of the United States is aging. Recent demographic research has shown that within the next ten years the populace of third agers will become a full 20% of the American population [1]. Due to seniors being in better health today and staying active long after retirement; many decide to engage in occupational pursuits or return to higher education [2]. But as third agers re-enter the workforce as retirees, start their own business, or enroll in a university; they face new challenges that are going to transform how they perceive themselves and how they are perceived by others [2]. These challenges can include barriers linked to age such as limited mobility or cognitive problems; they can be attitudinal like age based discrimination and their own perceptions of themselves, and they may be structural like lack of transportation or financial considerations [2]. The PEW Internet and the American Life Project indicates that while the technology skills of many seniors may initially be lacking, once they are introduced, many become avid e-mailers, internet searchers, and gamers [2]. Social media and A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 30–39, 2011. © Springer-Verlag Berlin Heidelberg 2011

Third Agers and Social Networking in Higher Education

31

networking websites have seen the number of users aged fifty and over nearly double in the last year [2]. Although there is the potential for great benefit; the number of seniors who actually utilize technology solutions remains low when compared to other groups [2]. As of January 2010 38% of people age 65 and over use the internet; a relatively low percentage considering the national average of 74% for other age groups [2]. Another study performed by PEW investigated this increase in internet use by seniors. This particular survey showed that while all age groups experienced an increase; the percentage experienced by users 55 and over was more significant. The study measured the percentage of seniors that used the internet in 2000 and then again in 2010. The result was a dramatic increase, people aged 50 to 64 increased from about 48% in 2000 to 78% in 2010. Further, the study showed that the number of those over 64 accessing the internet nearly tripled from about 15% in 2000 to 42% in 2010 [2]. With numbers increasing at these rates they become a demographic segment that is available, and often has disposable income to make purchases, pursue educational goals, and subscribe to services. As third agers are making efforts to improve their lives; they are turning to their local community colleges and universities for support. They are doing this in record numbers causing the educational community to take notice and pursue new and better ways to support this age group [3]. In fact, the number of students who fall into the 50 and over group has increased by 144% [3]. This trend shows that these learners are going to be a wonderful resource, as both a student pool and even a group from which to recruit professors, for educational institutions in the future.

2 Social Networking in Higher Education A Social networking site is defined as a site that allows individuals to construct public or semi public profiles, articulate a list of other users to share with, and view and traverse their list and those made by others [4, 5, 6, 8]. What separates social networking sites is that they are built on these connections [8]. According to Social Networking and Education: Emerging Research within CSCL; the study of how students are using this software is still in its early stages and the results of these surveys show the practice of using social networking sites in higher education is developing and relatively informal [6]. In looking at how social networking sites should be used, a few suggestions were put forth by Ryan Alexander in Social Networking in Higher Education [5]. The first is that institutions should identify which social networking architectures are most successful and look into whether or not a particular architecture is suitable for their environment. One feature that has made these sites so successful is the ease with which users can post their content. There is no need for web design expertise anymore; users simply fill out a form. This is true for posting pictures and graphics, submitting a blog update, and sending an online message to a fellow student. A second suggestion is to examine other social networking sites that have experienced a high degree of success. Other suggestions are to encourage professors to work on their own Wikis, and have students contribute and create entries [5]. While many campuses have utilized

32

C.M. Connolly and G. Meiselwitz

these technologies others have not; but that has not stopped creative students from building their own micro-versions of these sites specifically for their class or group. There are also some downsides for universities not participating in this trend; information and research that the University has created may end up on one of these informal sites without that University’s review and/or approval. Universities also run the risk of having proprietary information appear to an unrestricted audience. Universities may look into the uses and benefits of participating in Web 2.0 technologies and find the rewards will far outweigh the costs and difficulties [5, 8]. In looking at how social networking sites can be used in higher education there are two schools of thought that are being investigated. One is that students have already garnered how to get the most out of social networking sites and that the study of how they are currently being used is most important [6]. This is informal research where a portal is created for students; then researchers study how students utilize it. The main point of these studies is to explore the opportunities that are being created by the explosion of use among students; with a focus on using social networking as a collaborative tool for improving student achievement [6]. Students received many benefits in the process of making these connections. Students are shown as using the sites to create and test versions of their own personality, create relationships, and as creative outlets; aside from the simple sharing of information and knowledge. Other benefits of social networking sites to learners include creating a social learning environment, creating a practicing community, and creating team building environments. These activities contribute to the learning in the traditional classroom by bringing it outside the classroom and class-time. One can not underestimate the importance of the social aspects of learning [7]. There are other factors that will influence the creation of these social networks. One is the institution itself and its policy toward social networking. Another would be the technology policy of the institution. Are there Terms of Use policies that create a sense of Big Brother over the shoulder? This will limit social interaction and the depth of the relationships and collaborative efforts being forged as students are reluctant to break rules or to be exposed to online monitors that, while university sanctioned, are not the intended recipient. The type and quality of the network at the institution can be a factor as network speeds may influence the performance of a social networking endeavor. If the speeds are adequate; there are other factors like network congestion and down time for maintenance or repairs users can expect to experience that may also affect use. Aside from the physical network; things like the current social climate, political atmosphere, and student emotions can create barriers to social networking. The other position that researchers take when looking at social networking sites in higher education is that the best method for using social networking sites has yet to be discovered [7,8]. The role that social network sites play in education is the focus of the paper by Liccardi [8]. One point that was noted was that social networking group size should be a focus of study as the optimal number has not been calculated [8]. Another point that should receive attention was the importance of the group skill-set when creating groups, since much of the work is done online and groups have their own online dynamics. [8]. Hamasaki and Takeda studied the phenomenon of the “matchmaker” or the “friend of a friend” [9]. In this method, a friend of a friend introduces two people that are unfamiliar with each other. The article states that the manner in which social

Third Agers and Social Networking in Higher Education

33

networking sites facilitate the creation of relationships can be helpful in understanding their application to the student learning process. Another challenge this paper discusses is the community of practice [10]. This research looks into the dynamics that occur when a group forms for the completion of a task. This paper poses the question: “What will be the effect of social software on the formation of these groups?” Cho investigated the way advanced computer users were using social networking sites to develop a collaborative learning environment [11]. The project was with a homogeneous group of students working on the design of an aerospace system. The researchers took two groups of geographically dispersed students and emphasized the importance of collaboration in the assignment; the researchers then created and delivered a social collaboration web site for them to work with. A survey was distributed at the beginning, and at the end of the study to measure any pre-existing networks. There were two hypotheses being tested, the first was those with a better social status, or closer to the center of the friend circle would perform better. The second was that those in the center would not be as inclined to make new connections. The result showed that both hypotheses of the research were supported. But for this paper the results are not as important as the amount of study that went into the research and the scientific method that was used in the definition of a social network friend group, the positions of the individuals in the group, and the effect that can be observed from this. Indeed, the true result of this study was the observance that those who take advantage of their social contacts tend to achieve better results. How they use the sites was not covered.

3 Seniors in Higher Education As seniors become more attuned to the possibilities of internet activities many experience a desire for enhancing their knowledge. The desire for education is a natural response once third agers are exposed to the opportunities for learning that permeate the internet. Education is not strictly about the traditional subjects such as reading and mathematics, it is also learning about entertainment, local and national news, causes and charitable works, sporting events, and informal discussion of people’s lives. Especially in the recent economic downturn; where the stock market lost 27% of its value and job losses are at record levels, many older Americans are experiencing difficult situations. Indeed, as of November 2010 more than one million people aged 55 and over are jobless. Although this seems like a difficult situation; this same population of people are experiencing better health and have had more, and better educational opportunities [12]. This study has also shown an increase in the participation of older adults in educational activities. In 1998, only 3.6% of the population of Slovenia was active in educational pursuits, whereas in 2004 that had risen to 12.1%. This trend is true elsewhere and America is no exception; third agers are returning to the roles. In Education: a Possibility for Empowering Older Adults, the authors examine the environment older adults experience as they re-enter higher education [13]. The actual

34

C.M. Connolly and G. Meiselwitz

study was related to the motivations and successes that older adults were experiencing in higher educational institutions. There were three of these studies performed to measure trends. The first was in 1987, and second in 1998, and the third and last in 2004. When seniors in 1987 and 1998 were surveyed about their motivations for more education they listed the “joy of learning” as their primary motivator. Other choices that had a significant response were “enable self-development” and “improve my situation”. The newer evidence does point to the fact that overwhelmingly older learners are returning to the educational roles from an intrinsic motivation. In 2004, the primary reason (64%) listed for pursuing education was a need for companionship. They were satisfying this need by becoming a classmate and teammember. The second most popular reason (61%) was the “joy of learning”, again indicating the value this demographic places on education. This paper goes on to state that it can be assumed from these results that these learners have had a lifelong love of learning and that they are simply continuing to indulge it. This is attributed to another result of the study that indicated those that generally listed “love of learning” as a motivator had higher educational levels. Another area that was questioned were the barriers that seniors experienced when trying to re-enter school. The important data here was older learners did not list a lack of programs as a barrier. The choice that was most common not having time, 36% in 1987, 35% in 1998 and 45% in 2004. Interestingly, the researchers did not trust this result and stated it was more likely a lack of income is the true barrier as the older population is at a higher risk of living below the poverty line and/or alone [13]. Other studies that have been performed look not to older people’s perceptions about activities or institutions but at their perception of themselves. In Forging New Identities: Older Adults in Higher Education the researchers looked first at the terms used to address third agers and found that most viewed being called “older” or “seniors” negatively [14]. These people have their wealth of experience, vast knowledge, and understanding of the human condition. The challenge for younger generations is to view older learners through a different lens, it means to do away with many of the old perceptions and see the aging population as living intentionally, learning, adapting and growing [13, 14]. Unfortunately, these adults are also influenced by societal perceptions of what they can and cannot do. This is closely tied to learning. If these older adults perceive themselves as “done learning” it may be because they feel they are too old to learn. Overcoming this invisible barrier has more to do with changing the way older adults think of themselves than dealing with their actual abilities. For these adults a solution is remaining engaged, whether through work, learning, or community. For many of them it means forging a new identity, or a new perception of themselves. This is an important concept because many people’s personal identity is established by the kind of work they do [14]. As a society we look at each other in terms of what we each contribute. We are not simply asking adults to take a few classes but really we are asking them to be open to the possibility of changing the way they see themselves at a fundamental level [14].

4 Seniors and Social Networking Many social networking sites are experiencing the explosion of users that are age 50 and over [2]. In fact, during the period from 2009 to 2010 social network use by

Third Agers and Social Networking in Higher Education

35

Americans over 50 increased from 22% to 42%. Their motivation can be traced to any number of personal causes, such as combating loneliness, trying to keep in touch with loved ones in other parts of the country, or just staying active and using these as tools to stay involved in their culture [15]. Still there is a great opportunity; for although the numbers are growing rapidly, there are a significant number of older people not getting involved. Social networking sites are fighting a battle of perception among many older people. Many perceive them as a little silly and empty headed [16]. A study by Lehtinen looked at the perceptions older adults have of social networking sites and after conducting surveys found that: Older adults find the internet unwelcoming, and that social networking is not important. Many seniors lack a reference community, that is, they would be more interested in joining a social group that communicated online if they were familiar with the people in it. This study suggests third agers are not inclined to simply go online and find a group of people to talk to. This study used Netlog, a social networking site that provided space for a profile, file sharing like pictures and video, friend lists, and blogs. The subjects were instructed how to use the site; but not specifically why to use the site. The findings indicated that the users largely did not use the services; there were some that experimented with the site and made some postings. Some of the reasons cited for not using the site were: • A simple lack of interest – users may have used computers for many years but stated they view them as tools, not sources of entertainment. • Some users only felt comfortable using computer systems when someone else was maintaining them – they used machines at work but felt they were more trustworthy because they were maintained by professionals. • Many other people their age do not use it – older users have trouble finding each other; and when they do they find they would rather chat on the phone than online. • It is “cold” – many older users responded that the interactions were impersonal and less engaging than other mediums of communications. When understanding older adults entering higher education it cannot be assumed that they will immediately adopt all the technological communications mediums that are routinely used by students in their late teens and twenties. While older users may be more inclined to use a social networking site to meet with a study group or share notes, once the class task or assignment is complete it cannot be assumed they will maintain their account or remain active on that site [16]. It is important to realize that social networking in education is not just occurring through the classic internet site. An important social networking tool group is highlighted by Twitter and other mobile based applications. When looking at the senior population and taking into account the physical limitations they experience, texting (a life-skill for younger users) can be very difficult. The application of SMS by older people lags behind other generational groups [17]. Research is being performed into how to increase the adoption of texting by older people. It should result in some changes to the existing technology or new technologies that will make SMS (Simple Message System) easier to use. When the current twenty-something population ages, they may find texting has become an antiquated method of communication; but that does not necessarily mean they will wish to abandon it. They may simply need a modification to the way it is performed.

36

C.M. Connolly and G. Meiselwitz

Relating to this topic, social sites like Twitter require knowledge of the mobile device’s SMS system and how to send and receive messages. These are two options that have yet to be truly adapted for the physical limitations of older users.

5 Conclusion and Further Research Educators are working to find ways to engage third agers, and are dealing with a series of steps that can be taken to improve accessibility and acceptance [13, 14]. The first task is to identify the factors that older adults list as their objections to using social networking sites. A few were described as: They are simply not interested. The answer is motivation; however, it is important to reach out to your audience correctly with a sense of enthusiasm and excitement. At the same time it is equally important not to become pushy. As DeMicco discusses in Motivation for Social Networking at Work; “There has been an emergence of social networking sites that are targeted towards specific groups” [18]. The proper focus can be used as a motivating point for older users. Many listed a reason for their lack of interest was there was no one on the site they knew. Focusing on this audience will allow more users to feel the site was created for them, incline them to give it a try, and perhaps create regular users. Other lacks of motivation are simply seeing the sites a frivolous or not meaningful. However, another article in Seniors Online states that seniors most often go online to engage in social activities [19]. It then states that those seniors that have active social calendars and lives say that they have a higher quality of life. Education is part of this social calendar and therefore part of that quality of life. Studies performed by the PEW research institute state that when seniors go online they are also looking for information about health care, news, and events [3]. Online communities that target that population present an excellent opportunity to accomplish these activities all in one place. Many seniors may find great benefit from the social interaction they would experience if they could be convinced to access, and had the ability to use, social networking sites. They simply require an introduction to the information available. Many social networking sites do not have enough accessibility options for those with physical impairments [20]. In early November 2010 the social networking giant Facebook reduced the size of its font for status updates and names. This is easily corrected if the user knows to hold the Control key while hitting the plus sign (+); but that is not indicated anywhere on the site. There are many online resources detailing the process but how are older users to know to go there? Research into making sites easier to access for older adults is not new, but newer technologies like touch-screens and tabletop computers can make things easier to use by expanding the screen and removing the need for the double-click [21]. What can be encouraged, that may increase use of social networking sites among third agers, is the publicizing of these aids so seniors may find more and better solutions to accessibility problems. Privacy is a concern for older adults. Many consider the internet a frightening and unfriendly place. They are concerned about their information falling into the wrong hands and doubly concerned about how the information they put on the internet may be used. Consequently they are reluctant to join a site and provide information over a connection regardless of the privacy controls and security in place [22]. While the information and the privacy issues are well addressed by the creators of the big social

Third Agers and Social Networking in Higher Education

37

networking sites, it is important to educate the users, in layman terms, how to secure their profiles. Then inform them of policies that are in place and software engaged to make sure that the information they post to their friends is not going to be taken and used against them. It is also important to make sure that the company providing the social networking site lives up to these promises. Establishing and cultivating trust is the most important ingredient when asking people who are naturally mistrustful of the medium to embrace it. Once the barriers to using the medium have been addressed the second step is to introduce the benefits of the medium to these users. One of the benefits is how the collaborative nature of the medium can create new communities for seniors to join and enjoy. As third agers have expressed they are “not done yet”. They may be up for new challenges and embrace education as a terrific and rewarding undertaking [13, 14]. When they engage they will meet new people and expand their knowledge. They will develop new social structures and relationships as well as engage and form new identities in a new environment [14]. Many enjoy a love of learning and can turn what some see as a great challenge into an enjoyable pastime that is its own reward [13]. The possible benefit here is the enhancement of their lives which, although it is a more intrinsic and less tangible benefit, can be a powerful motivator. Many third agers enjoy feeling useful and providing important contributions to society after they have retired. Education, and the social networking involved, gives them a chance to work toward their goals. An important benefit experienced by their younger student peers will be the sharing of their considerable life experience in the classroom [14]. Further research can be done into the terminology that these sites use and the conflicts with pre-existing notions those seniors have for what each word means and what the site is asking them to do. These double meanings create needless barrier to access. Current research also suggests that interface methods for adjusting sites for the physical limitations that seniors experience should be given more attention, for example magnifiers for increasing text size, options for single click over double click on links, and quick and easy navigation back to their home or profile that include the action words like “Click here to return to your profile”. Investigation and progress into these particular areas may make these social environments more inviting to seniors. This research will be used in a larger study to create an interface that third agers can easily use to access Internet content. The goal is to encourage Internet use by creating an interface that makes it easier to navigate. The information in this paper about how and why third agers are currently using the Internet will be used to create a survey measuring seniors current internet use. This survey will be used to identify a group of seniors, age 70 to 75 with little or no internet experience. These seniors will be asked to volunteer their assistance in creating this interface. The interface design and function will be created using some of the research from this paper regarding design and accessibility features; and with information garnered from the volunteers for ease of use and adoptability. This application will run as a proxy server opening the Internet inside a window. The interface will surround the internet window but will still run in a traditional Java-enabled browser; making it accessible on most internet capable devices.

38

C.M. Connolly and G. Meiselwitz

The interface will eliminate double-clicking as a navigation tool by allowing users to associate single click buttons with links existing in the website. These large buttons will have clearly defined functions to simplify the interface and contain action words telling the user to “click here” and what will happen when they do. The interface will also provide “always present” tools for increasing text size and screen contrast that are clearly labeled. Finally, the input from the volunteer group will help in the organization of these features for screen location, size, and labeling. Our volunteers will then be asked to perform a series of internet tasks using Internet Explorer. The users will then perform a closely related but different set of tasks using the new interface. For both task sets the users will be timed and measured for accuracy, with accuracy being a measure of errant clicks and times the user asked for help. The research question being endeavored is whether users with limited internet experience can find the internet easier to use with an interface that provides action based buttons for navigation rather than hyperlinked text.

References 1. The Demographics of an Aging Population, http://www.ageworks.com ROI, http://www.ageworks.com/course_demo/200/module2/module2.htm (accessed on May 12, 2011) 2. Lakin, M., Mullane, L., et al.: Framing New Terrain: Older Adults & Higher Education. Reinvesting in the Third Age: Older Adults and Higher Education. American Council on Education, Washington, DC (2007) 3. Pew Internet and the American Life Project, http://www.pewinternet.org/ (accessed on January 12, 2011) 4. Boyd, D., Ellison, N.: Social Network Sites: Definition, History, and Scholarship. Journal of Computer Mediated Communication 13(1) Article 11 5. Alexander, B.: Social Networking in Higher Education. Selection from: The Tower and the Cloud. Copyright, Educause (2008) 6. Greenhow, C.: Social Networking and Education: Emerging Research within CSCL. CSCL 2009 ROI (2009) 7. Brown, J., Duguid, P.: Organizational Learning and Communities-of-Practice: Toward a Unified View of Working, Learning, and Innovation. Organization Science 2(1), 40–57 (1991) 8. Liccardi, I., Ounnas, A., Pau, R., Massey, E., Kinnunen, P., Lewthwaite, S., Midy, M., Sarkar, C.: The Role of Social Networks in Students’ Learning Experiences. In: ITiCSE – WGR 2009 (2009) 9. Hamasaki, M., Takeda, H.: Find Better Friends? Re-Configuration of Personal Networks by Neighborhood Matchmaker Method. In: SWAFT, pp. 73–76 (2003) 10. Wenger, E.: Communities of Practice: The Organizational Frontier. Harvard Business Review (2000) 11. Cho, H., Gay, G., Davidson, B., Ingraffea, A.: Social Networks, Communication Styles, and Learning Performance in a CSCL Community. Journal of Computers in Education 49(2) (September 2007) 12. Rix, S.: The Employment Picture, October 2008 – Mostly Grim News for Older and Younger Workers. Research Report (November). AARP Public Policy Institute, Washington D.C (2008)

Third Agers and Social Networking in Higher Education

39

13. Kump, S., Krasovec, S.: Education: a Possibility for Empowering Older Adults. International Journal of Life Long Education 26(6), 635–649 (2007) 14. Lankin, M.: Forging New Identities: Older Adults in Higher Education. Internationsl Journal of Continuing Education and Lifelong Learning 2(1) (2009) 15. Cross, D.: Older users Flocking to Facebook, Twitter. CNN ROI, http://www.cnn.com/2010/TECH/social.media/08/27/older.users. social.networks/ (accessed October, 2010) 16. Lehtinen, V., Nasanen, J., Sarvas, R.: A Little Silly and Empty Headed – Older Adults Understandings of Social Networking Sites. British Computer Society (2009) 17. Weilenmann, A.: Learning to Text: an Analytic Study of How Seniors Learn to Enter Text on Mobile Phones. In: CHI 2010. ACM, New York (2010) 18. DiMicco, J., Millen, D., Geyer, W., Dugan, C., Brownholtz, B., Muller, M.: Motivations for Social Networking at Work. In: CSCW 2008. ACM, New York (2008) 19. Seniors Online, Seniors online increase article (2004), http://www.seniorjournal.com/NEWS/SeniorStats/03-02-04 SnrOnline.htm (accessed on January 11, 2010) 20. Holmes, J., Powell-Griner, E.: Lethbridge –Cejku, M. Heyman, K.: Aging Differently: Physical Limitations Among Adults Aged 50 Years and Over. Center For Disease Control, http://www.cdc.gov/nchs/data/databriefs/db20.htm 21. Piper, A., Campbell, R., Hollan, J.: Exploring the Accessibility and Appeal of Surface Computing for Older Adult Health Care Support CHI. ACM, New York (2010) 22. Pfeil, U., Arjan, R., Zaphiris, P.: Age Differences in Online Social Networking – A Study of User Profiles and the Social Capital Divide among Teenagers and Older Users in MySpace. Computers in Human Behavior 25(3), 643–654 (2009) 23. Pfeil, U., Arjan, R., Zaphiris, P.: Age Differences in Social Networking – A study of user Profiles and the Social Capital Divide among Teenagers and Older MySpace Users. Center for HCI Design London, UK (2008)

A Real-time Disaster Situation Mapping System for University Campuses Kayo Iizuka1, Yasuki Iizuka2, and Kyoko Yoshida1 1

School of Network and Information, Senshu University, Kanagawa, Japan 2 School of Science, Tokai University, Kanagawa, Japan {iizuka,k-yoshida}@isc.senshu-u.ac.jp, [email protected]

Abstract. This paper proposes a real-time disaster situation mapping system for campuses. University campuses have various features and issues (e.g., difficult to determine the number of people on campus at any one time as there are many students studying at the library or in their rooms, visitors can also use the open spaces, building layouts are often very complicated etc.), and therefore unique disaster prevention efforts are required. In order to address these prevention issues, a system that facilitates disaster situation information transmission by users will be effective. PlaceEngine is implemented for this system to allow users to estimate the current location easily by utilizing Wi-Fi devices. Keywords: disaster prevention, campus map, situation information gathering, PlaceEngine, social media.

1 Introduction There has been great improvement in various disaster (e.g., capital earthquakes directly above their epicenters) prevention countermeasures and their researches for more than a decade. However, most of the countermeasures that have been implemented are classified as “Kojo” which are implemented by public sectors (in Japanese), therefore, it is considered that the importance of both “Jijo” (countermeasures implemented by the individual) and “Kyojo” (countermeasures implemented by mutual assistance) is increasing [1]. University campuses are one of the prime examples of places that require Kyojo (mutual assistance). In order to assist better evacuation, current situation information of each site is useful, and in most cases, the latest information of each site is grasped by people who were just there. The current situation of each site is necessary not only for evacuation, but also for administering first aid or relief operations. However, not all buildings are closely packed together on campuses, and that may create difficulty in gathering situation information inside the campuses. Under these conditions, a system that facilitates disaster situation information transmission by users (students, staff, faculties, visitors) will be effective. It facilitates supporting people to evade danger, and facilitates support of disaster countermeasures offices to grasp the situation of the many places on campuses. The system proposed in this paper was designed and A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 40–49, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Real-time Disaster Situation Mapping System for University Campuses

41

prototypes were developed for such needs for real-time information sharing during disasters. In order to stimulate mutual assistance, information infrastructures are required. “Recent disasters highlight the importance of social media supporting critical information gathering and dissemination efforts by members of the public” [2]. Recently, social media such as twitter have been effective for disaster evacuation or rescue [3], but these need to be even more effective to gather disaster situation information, categorize them and visually show them on maps. The real-time disaster situation mapping (RDSM) system proposed in this paper can function as one of the social media.

2 Background As mentioned previously, the need for mutual assistance seems to be increasing, because in actual fact many people survive in times of disaster by mutual assistance. In this chapter, disaster prevention works are reviewed and classified from several perspectives (Table 1). Related papers were extracted that matched the specified criteria: those that examined disaster prevention, those that developed an information system, and those whose publication date is post the Great Hanshin Earthquake (Jan.17, 1995). Twenty six papers were extracted using this process [4]-[30]. Public sectors are the overwhelming majority for target users of the system, one paper targets citizens, and two papers target schools or companies (Table 1 (a)). For assistance type, Kojo (assistance by public sectors) is the overwhelming majority (72.1%), and only five papers dealt with Kyojo (mutual assistance) (Table 1 (b)). More than half of the papers dealt with disaster a priori (Table 1 (c)). For the area scope, “broad” (area includes public sector) has the highest percentage (84.6%), and “medium” (schools, shopping center etc.) occupies only 15.4% (Table 1 (d)). As seen above, disaster prevention systems are mainly targeted at public assistance from the public sectors across broad areas. The RDSM system proposed in this paper targets university campuses whose area scope is medium sized, and Kyojo (mutual assistance) because mutual assistance is effective for this scope. Considering the accessibility and role of social media, common devices (mobile phones, PCs) are adapted to the RDSM system.

3 Requirement for Campuses Disaster Prevention University campuses have various unique features and issues when considering disaster prevention. For example, it is difficult to determine how many people there are on campuses. There will be many students studying at libraries, spare rooms, or cafeterias. Visitors can also use the open spaces. Issues with places also exist. The layouts of buildings are often very complicated. They differ depending on when they were built, and construction of buildings differs by research field. Sometimes public roads cut through the campuses. Some campuses are located on inclined ground. Layout of the rooms (e.g., classrooms, experimental laboratories, practical rooms etc.) is not always familiar, therefore, arriving at one's destination is sometimes not easy. In addition, administrative structural issues exist, which are also different from those

42

K. Iizuka, Y. Iizuka, and K. Yoshida

in companies. On-campus disaster countermeasures offices will be set up in times of emergency. However, not all buildings are closely packed together in campuses, and that may cause difficulty in gathering situation information inside the campuses. Under these conditions, a system that facilitates disaster situation information transmission by users (students, staff, faculties, visitors) will be effective. It facilitates support of people to evade danger, and it facilitates the support of disaster countermeasures offices to grasp the situation of many places on the campuses. This system compensates for the area not covered by the public sector (national response system) (Figure 1). Table 1. Classification of Disaster Prevention Related Works (a) Target users Number of papers Public sectors 23 Citizens (individual) 1 Others (*1) 2 Total 26 (b) Assistance type Substances Number of papers Kojo 19 (Assistance by Public Sector) Kyojo (Mutual 5 assistance) Jijo (Self assistance) 2 Total 26 (c) System operating point Substances Number of papers A priori 15 Substances

(e) Point of use Number of papers Outdoor only 19 Outdoor / Indoor 5 Indoor only 2 Total 26 (f) Equipment used Ratio Substances Number of (%) papers 71.2 PC・mobile-phone 15 (via Web browser)

Ratio (%) 88.5 3.8 7.7 100.0

Substances

19.2 Special-purpose 11 terminal 7.7 100.0 Total 26 (g) Adapted technology Ratio Substances Number of (%) papers 57.7 GIS・web-GIS 17

Ratio (%) 73.1 19.2 7.7 100.0 Ratio (%) 57.7

42.3

100.0 Ratio (%) －

(first response)

A priori A priori and a posteriori A posteriori

7 4 0

26.9 Wireless sensor 15.4 GPS 0.0 Others (QR Code, etc.)

6 4 1

－

28

－

－－

(reconstruction support)

Total Substances Broad

26 (d) Area scope Number of papers 22

100.0

Ratio (*1) Companies, Schools, etc. (%) (*2) Ratio is not applicable because some papers deal with more than one technology. 84.6

(Area includes public sector)

Medium (Schools, Shopping center etc.) Narrow

4

15.4

0

0

26

100.0

(Home, Private concern)

Total

Total

A Real-time Disaster Situation Mapping System for University Campuses Formally included In National Response System

Formal response organizations Affiliated Volunteers Spontaneous Volunteers

43

Incident management system Victims

Community members

Private Sector No defined role In National Response System

Media

Fig. 1. Positioning of Disaster Prevention Systems [31]

4 Real-time Disaster Situation Mapping (RDSM) System 4.1 Overview The Real-time Disaster Situation Mapping (RDSM) system is a web based system that can handle disaster situation information on the campuses by the level of detail, such as classrooms or laboratories in the school buildings. The RDSM system can aggregate the information that is sent by users (informers), store it in a situation database on the server and display a disaster situation map to users on request. Assumed users are people staying on the campus and disaster staff of countermeasures offices (Figure 2, Figure 3). 4.2 Service Conditions of RDSM System The RDSM system is developed based on the following service conditions: • Disaster type: The RDSM system mainly targets the countermeasures for earthquakes. The quake damage assumed that buildings on campus were not fully but may partially destroyed, and mobile devices (PC or phone) are in use and connected to servers inside the campuses or other campuses located remotely. Disaster Information Map

Real-time Disaster Situation Mapping system PlaceEngine General Users -Students -Administration Stuffs -Faculties -Visitors

Updated on request

Disaster Countermeasures Offices

Stored in DB (server)

Situation information by location

Get entry screen

Store

・ situation text voice image

on request Need map

It seems to be safe here .

Situation mapping subsystem General Users （informers） -Students -Administration Staff -Faculties -Visitors

Fig. 2. Overview of System Use

Situation Database ・location

（SGS） Send situation data Use recent information for better decision making

That person who has fallen over needs to be rescued

(RDSM) Situation gathering subsystem

Get location data

mapping on request

show to users

(SMS) information request

Fig. 3. System Flow of RDSM

44

K. Iizuka, Y. Iizuka, and K. Yoshida

• Place of use: People can access the RDSM system from inside the building on campuses, and they can also access it from outside of campuses such as from evacuation centers. Although the RDSM system is developed for campuses, this system can be used for facilities that have similar features to campuses. • Targeted users: People such as students, administration staff, faculties, and visitors are assumed to be informers. People who are assumed to use maps are disaster countermeasures office staff, in addition to people listed as informers above. 4.3 Subsystems of RDSM and Functions The RDSM system consists of two subsystems: the “situation gathering subsystem (SGS)” and the “situation mapping subsystem (SMS)”. Situation gathering subsystem (SGS). The situation gathering subsystem (SGS) facilitates the support of people on the campus to send situation information (e.g., safety or danger, damaged situation) of the location where they are staying at that time, using mobile devices (PC or phone). Informers can access the disaster situation entry screen (Figure 4). Their location information (degree of latitude and longitude) can be specified by PlaceEngine [32]. PlaceEngine is a service that allows users to easily estimate the current location by utilizing Wi-Fi devices. Users can send situation information from the “situation data entry screen” by selecting an answer from check boxes or radio buttons. A free description column is prepared if required. Speech to text transformation is feasible using w3voice [33]. "w3voice skeleton" is a development kit for voice-enabled Web applications. Voice data and image data as well as movie data are also available for sending (Figure 5).

5. Identify location by intensity of Wi-Fi signals

PlaceEngine server

Real-time Disaster Situation Mapping system (RDSM)

Situation Database

Situation gathering subsystem (SGS) 4.Access to PlaceEngine Server

6.Get location data from PlaceEngine 1.Get access to SGS (entry screen)

Situation mapping subsystem (SMS)

3.Receive Wi-Fi signals 2.Activate client software of PlaceEngine

Fig. 4. Entry Screen (SGS)

7． Access to SMS (Situation Map)

Fig. 5. Location Data Acquisition Scheme by PlaceEngine

Location information (degrees of latitude and longitude, or name of the place at which the informer is staying) can be identified by manual entry; in addition, automatic acquisition of location information is feasible by using PlaceEngine. By this function, users (informers) can send location information, even if they do not recognize their location. Entered situation data are stored in databases.

A Real-time Disaster Situation Mapping System for University Campuses

45

Fig. 6. Screens of SMS

In order to set up PlaceEngine for automatic acquisition of location information (degrees of latitude and longitude) the following requirements must be met: • • • • •

Access point data of Wireless LAN are saved in the database of PlaceEngine. At least one access point is reachable from the place where the informer is staying. Access to the PlaceEngine server is feasible. Mobile devices (PC or phone) are equipped with Wireless LAN (Wi-Fi) equipment Client software is installed on the mobile devices (PC or phone) of the informer.

Location information (degrees of latitude and longitude or place name) can be registered by each access point, so identifying places such as classroom or hallways is feasible because there are many access points of wireless LAN on campuses. Situation mapping subsystem (SMS). The situation mapping subsystem (SMS) reflects situation data that are transmitted by users onto maps. It reads data of the situation database that is updated by SGS on request from users. Maps are composed of three levels: overall campus map level (it presents a birds-eye view), building and floor map level, and detailed information of disaster situation by room type (e.g., classrooms, offices, laboratories) level. The situation information amount is displayed at the overall campus level (Figure 6). By clicking a certain building on the overall campus map, a building and floor map appears. Disaster situation information such as fire, rescue required, building damaged, property destruction, and flood damage is shown by icons. The time when those situations were informed is also shown on this

46

K. Iizuka, Y. Iizuka, and K. Yoshida

level map. By clicking a certain room on the buildings and floors map, detailed information of the disaster situation is shown. Detailed information is displayed by time-series on this screen, and voice data or image data are available for access here. The map displayed in the situation mapping subsystem (SMS) can be created from existing map image data that are used for university brochures. The existing map data would be transformed to png format. Situation data that are described or depicted by text data or icons are rendered on this png map. In order to insert disaster-affected points on the map, obtained degrees of latitude and longitude of two different points (place) from Googlemaps are used. By calculating the distance between users’ location and the point obtained from Googlemaps, users’ location is identified from the reduced scale distance that is started from two different points. 4.4 Significance and Features of RDSM System The significance and features of RDSM system are as follows: • Assists users (students, staff, faculties, visitors) in evacuation or rescue: Since the RDSM system is a web based system and uses common mobile devices (PC, phone), users can find out the latest information of a disaster situation (e.g., safety or danger, damaged) according to location easily. PlaceEngine was adapted to assist users in locating and entering their position. Users can choose disaster situation information from options on the screen, instead of having to type everything about the situation they want to convey; however, users can send text data, voice data, and graphic data to the server, if necessary. • Assists the disaster countermeasures office inside/outside the campuses: Although countermeasures office staff inside the university need to gather situation information, it is difficult to keep tabs on the situation overall because of the limited human resources. The RDSM system can cope with this problem by gathering information on the places from the people who are actually there. • Available for use in everyday situations (not just in times of disaster): It is often said that disaster prevention systems do not work in times disaster because the users are not used to operating them. The framework of the RDSM system is effective for everyday situations in the universities, for posting event notifications, maintaining buildings, and so forth (this idea was obtained through feedback from administration staff in the university). In addition to the features mentioned above, disaster situations (fire, rescue required, building damaged, property destruction, and flood damage) are indicated by icons, in order to improve usability. The latest situation data is presented on the map face; however access to time series data is also available for cases where information of the degree of situation change is required. This feature is practical for assisting with rapid rescue, or preventing the damage from spreading. To ensure the communication environment in times of disaster, a backup system at remotely-situated campuses of universities is recommended.

A Real-time Disaster Situation Mapping System for University Campuses

47

5 Conclusion and Future Works As mentioned in previous chapters, the RDSM system facilitates assistance of general users (students, staff, faculties, visitors) and disaster countermeasures office on campuses for effective evacuation and rescue. Thus, the RDSM system facilitates Kyojo (mutual assistance) that is expected to reduce disaster damage, and it can play the role of a social media. The RDSM system is an updated version of a disaster mapping system that was developed by the Yoshida-Iizuka Project 2009 in Senshu University. Comments from attendees of the project exhibition were as follows: The majority of comments were “It makes us feel safe should a disaster occur”, and “It must be available for various sites on campuses” was also heard often. For agita factors, comments such as, “feel anxious about the communication environment in times of disaster”, “worry about mischief and malevolent literature”, “information security must be considered”, were heard. The RDSM system is also premised for use in everyday situations, so this answers the comment “It must be available for various situations on campuses”. RDSM is not supposed to save personally identifiable information. However, if the requirement for dealing with personally identifiable information appears in the future for reasons such as corresponding to the need for urgency, authority setting for accessing data must be carefully considered. Countermeasures for mischief and malevolent literature are not considered now, however, the analysis function of informed data by comparing time series data or data of the surrounding area, is recognized as one of the items to be considered. When considering the “feel anxious about the communication environment in times of disaster” comment, the communication environment is not always completely destroyed by disaster as shown when the communication environment was restored on the day following that of the Great Hanshin-Awaji Earthquake (Magnitude:7.3) in 1995 [6]. By setting up a backup system at remotely-situated campuses of universities, ensuring the communication environment becomes more feasible. Crowd computing would be effective for the universities that do not have remotely-situated campuses. Customizing for each university, upgrading the control panel of RDSM system is planned in order to allow flexibility of controlling system parameters such as disaster emergency level. The input screen of the building layout can be improved to facilitate use. Even though the operating systems that are supported by PlaceEngine are currently limited, they will be widespread in the future. By improving these factors, the role of the RDSM system as social media will be more firmly established. Acknowledgement. Part of this work is supported by a research grant from the Institute of Information Science, Senshu University. We thank Hayato Tsuchiya, Masahiro Urabe, Toshiaki Tsunashima, Tokinori Suzuki, and Masayuki Ishikawa for their contribution to the development of the RDSM system. We would also like to thank members of Yoshida-Iizuka Project 2009 in Senshu University.

References 1. White Paper on Disaster management, Cabinet Office, Government of Japan, http://www.bousai.go.jp/hakusho/h20/index.htm 2. Guido, L., Raquel, B.-F.: The Use of Social Media in Disaster Situations: Framework and Cases. International Journal of Information Systems for Crisis Response and Management (IJISCRAM) 2(1), 11–23 (2010)

48

K. Iizuka, Y. Iizuka, and K. Yoshida

3. Emergency Services embrace Social Networking in Queensland Floods, http://socialmediadaily.com.au/emergency-services-embracesocial-media/ 4. A large-scale disaster and the Internet - How did the Internet correspond to the Great Hanshin Earthquake?, All-in-One INTERNET magazine 2.0, http://i.impressrd.jp/e/2008/01/17/343 5. Nanbu, S., Takahashi, I.: Personal Safety Confirmation System Utilizing Emergency Earthquake Warning and Position Information System by Local Mobile Phone Network. In: Proceedings of the Annual Conference of the Institute of Social Safety Science, vol. (14), pp. 13–14 (2004) 6. Kuwata, Y., Shinjo, A., Ohtani, H., Inoue, U.: A GIS-based Real-time Disaster Information Sharing System. Journal of Information Processing Society of Japan 43(11), 3419–3428 (2002) 7. Terada, M., Sano, Y., Inoue, A., Kaneda, S.: A Web-GIS for Initial Stage of Disasters. Doshisha University Policy & Management Review 8(2), 257–266 (2006) 8. Asakawa, K., Hirano, K., Tsukada, T., Hayashi, Y., In, Y., Omiya, Y., Hamai, T., Murakami, H.: Collection and presentation system of safety information in disaster: An application of GPS mobile phone. ITE Technical Report 33(11), 123–126 (2009) 9. Ichihara, T., Ito, S., Mase, K., Kunifuji, S.: Information gathering system using IR tag in environment with difficult construction of network, IPSJ SIG Technical Reports. Groupware and Network services (30), 37–42 (2005) 10. Hada, Y., Takizawa, O., Shibayama, A., Gyoda, K., Suzuki, T., Kawabata, K., Kaetsu, H., Asama, H.: Research on Ubiquitous Network Technology for Information Acquiring in Disaster. Journal of The Japanese Society for Artificial Intelligence 23(4), 480–485 (2008) 11. Jeong, B., Zama, S., Endo, M., Takizawa, O.: Development of a Prototype Disaster Damage Information Collecting System Based on Cellular Phone. In: Proceedings of the Annual Conference of The Institute of Social Safety Science, vol. (21), pp. 15–16 (2007) 12. Aoki, M., Yonemura, S., Shimokura, K.: Information presentation based on cellular phone in a disaster. Correspondences on Human Interface: Human Interface 9(2), 169–174 (2007) 13. Hirabayashi, Y., Hasegawa, A., Hasegawa, S.: Campus Evacuation Routes Information to be delivered to Mobile Phones. Journal of Nagoya Bunri University 7, 57–64 (2007) 14. Shibayama, A., Takizawa, O., Hosokawa, M., Ichii, T., Hisada, Y., Murakami, M.: A Study on the Information System Using Radio Frequency Identification. Journal of Social Safety Science (8), 135–144 (2006) 15. Ichii, T., Shibayama, A., Murakami, M., Sato, T., Hisada, Y., Namai, C.: Development of the Disaster Information Mutual Support WebGIS as purposes of use in advance or disaster situation. AIJ Journal of Technology and Design (22), 553–558 (2005) 16. Ohno, K., Furuhashi, T., Ishimaru, S., Hata, I.: Basic Research on Earthquake Disaster Preventing Bulletin System using Web Sites: The Outline and Usage of The System, Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, B-2, pp. 799–800 (2008) 17. Kurata, K., Fukuwa, N., Tobita, J.: Development of Web-Based Geographic Information System for Supporting Disaster Mitigation Activities. Journal of Social Safety Science (10), 293–300 (2008) 18. Nishimura, T., Nakada, Y.: Design of an Open Disaster Information System Using GIS. The 57th National Convention of IPSJ (3), 607 (1998) 19. Ishida, E., Fukuwa, N.: Development of GIS for Seismic Disaster Mitigation on Internet Using JAVA. AIJ Journal of Technology and Design (5), 287–291 (1997)

A Real-time Disaster Situation Mapping System for University Campuses

49

20. Tani, S., Fukuhara, M.: Disaster Prevention System by Web GIS in Rural Area. Geoinformatics 17(2), 146–147 (2006) 21. Takabayashi, S.: Information acquisition in disaster crisis management: implementation of image and location data. Bulletin of Saitama Women’s Junior College (20), 1–10 (2009) 22. Narita, S., Shibata, Y.: Disaster Safety Support System Using both Indoor and Outdoor Location Information. The 70th National Convention of IPSJ (4), 875–876 (2008) 23. Nozaki, K., Fukui, Y., Shibata, M., Tanaka, H., Matsuda, Y., Yoshimoto, N., Tsukada, K.: A Proposal for Dynamic Emergency Navigation System in a Time of Disaster. IPSJ SIG Technical Reports DBS (141), 185–190 (2007) 24. Shiwaku, K., Sasaki, M., Kakumoto, S.: Introduction Process of Information System for Community Based Disaster Management. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, F-1, pp. 371–372 (2008) 25. Park, Y., Wang, J., Furuya, T., Sadohara, S.: Development of The Crisis Management System for Improvement on the Emergency Ability Part1: Construction of Emergency Information Database system for the Disaster Support Using GIS. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, D-1, pp. 605–606 (2005) 26. Aoyama, T., Ichii, T., Murakami, M., Hisada, Y.: WebGIS used in normal and disaster situations for sightseeing spot: Application for the ITO in IZU peninsula. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, A-2, pp. 491–492 (2006) 27. Tsuchiya, M.: “TSUJIDO Area” Disaster Prevention Map using GIS (Geographic Information System). Annual Design Review Japanese Society for the Science of Design 11(11), 12–15 (2006) 28. Kurata, K., Murakami, H.: Development of Simple GIS software for Community Disaster Preparedness: A Case Study for Ube City. In: Proceedings of the Annual Conference of The Institute of Social Safety Science, vol. (17), pp. 29–32 (2005) 29. Sato, T., Ichii, T., Murakami, M., Shibayama, A.: Development of the Disaster Information Mutual Support WebGIS: Part 1: Construction of the Date Base on Disaster Information. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, A-2, pp. 501–502 (2005) 30. Imahashi, Y., Kawaguchi, J., Morino, S.: Study on the Earthquake-proof Hazard Map of Mie University Campus by Using M-GIS. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, B-2 pp. 1027–1028 (2004) 31. http://www.gwu.edu/~icdrm/publications/PDF/Harrald%20HSGAC%2 0testimony_rev.pdf 32. PlaceEngine, http://www.placeengine.com/PlaceEngine 33. w3voice, http://w3voice.jp/skeleton/

Recent Trends in Software Support for Online Communities for Teaching and Research Projects in Higher Education Daniel Kadenbach and Carsten Kleiner University of Applied Sciences and Arts Hannover, Germany {daniel.kadenbach,carsten.kleiner}@fh-hannover.de Abstract. This paper investigates the evolution of software project support for teaching and research projects in higher education. It therefore analyses the results of two surveys conducted with students of the department of computer science of the University of Applied Sciences and Arts Hannover in February 2009 and November 2010. The surveys focused on examining the experiences and requirements of the students. Beneath mere technical requirements the surveys also investigate the acceptance of social software elements in project support. Since containing a core of equal questions the answers to the two surveys allow to deduce trends. These trends towards more sophisticated supportive software tools and environments are further investigated not only to allow a better support for projects, but also to make projects more visible and valuable, preserving their achievements and making them accessible for others. Keywords: software support, projects in higher education, online collaboration, social software.

1 Introduction and Motivation The University of Applied Sciences and Arts Hannover offers bachelor and master’s degree programs in applied computer science. Within these programs a multitude of different projects is conducted each semester and beyond. Similar to many institutions, in addition research projects of longer durations are also executed. These projects differ greatly in their type, contents, lifetime, the number of people involved and their goals. They vary from one person projects such as seminar works, comprehensive homework or final theses to those with many project members like capstone projects or scientific projects led by the faculty or staff. In particular in the scope of software development projects, but also beyond that, a trend becomes more and more apparent to make use of sophisticated supporting software tools, which can provide an enormous benefit if a suitable set of tools is chosen for a project, accepted by the project team and used correctly. This paper analyses how project support by software in the department has changed over time and which possibilities this change offers, but also which challenges are involved with it. By this other institutions shall be given the opportunity to have an insight into the gathered experiences and findings, especially by drafting trends in the usage of supportive software tools for projects, and to encourage an information exchange leading to an improved situation for projects in all higher education institutions. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 50–59, 2011. c Springer-Verlag Berlin Heidelberg 2011

Recent Trends in Software Support for Online Communities

51

1.1 Example: A “Traditional” Project A software-development project for students without supportive software-systems shall serve as an example for this evolution. Such a project could begin with a few meetings of all project members in which the requirement analysis is conducted, the softwaresystem is designed and further tasks and responsibilities are assigned to the participants. The students henceforth will work on their own on their particular tasks between further team meetings; or they could use frequent meetings to work together in a room on the development of the software. The latter approach, although offering far better communication, will be more expensive in terms of time effort, because the students need time to reach the team’s working-room and because the meeting place usually lacks some comfort, so probably students will try to avoid the additional effort if possible. In addition it is sometimes difficult for the institutes to provide each project team with an individual lab room due to space constraints. Finally the students will put together the modular developed parts of their application and in the best case will test the complete system extensively, and maybe even document it properly if enough time is left. The results of their work will be delivered and if applicable also presented in front of other students, but in most cases they will not be trackable even after a short time. Thus the value of the software developed itself as well as lessons learned during project execution will soon be lost. No sustainable project improvement process is achievable for student’s projects on one hand. On the other hand projects cannot build on previously implemented software in order to achieve higher goals. 1.2 Example: A Project with Software-Support This process can be improved in many ways by supportive software systems for software development projects, which appeared mainly within the development of Open Source Software, such as version control systems, wikis, trackers and others. Because they are used in nearly every real software development project nowadays, the knowledge of these different tools has become an absolute necessity in an up-todate study program in applied computer science to improve student employability. In addition these tools also support the development project itself and thus may improve the results. That implies the knowledge of their operating mode, their advantages and disadvantages in different projects and their handling. In order to make the right choice in the upcoming projects, a student has to know about the qualities and features of a variety of supportive software systems. In using the adequate supporting functions, totally new possibilities could arise in this exemplary software development project. To just name a few: • The use of a distributed version control system like Subversion offers the possibility to work together on the whole source-code of the application from any place. Additionally, one can be informed about the results of each member of the team, avoid conflicts of different versions and make the results of the projects accessible and useful even after the run-time of the project. • A wiki provides a central, up to date, documentation which matches the specific requirements of the project. On one hand the created content can also be accessed and

52

D. Kadenbach and C. Kleiner

preserved beyond the project life-time. On the other hand it can always be changed and updated easily at all times. It is therefore remarkable how simple it is for all team members to manage and find project related information which increases the likelihood that a useful documentation is created. The wiki can provide an external online presentation of the project for other students or even on the Internet as required. It increases the project’s visibility and, if desired, can be a way for the project to receive feedback. • Issue-tracking systems can be used to keep track of bugs, feature requests, or even the planned tasks, forming milestones of packages of this tasks. Team members can be attached to every issue as responsible persons for it. Also the progress of the issues and milestones and therefore of the project itself can easily be tracked by other developers, instructors or even customers. The motivation of the students to reach a high quality in every aspect of the project can be increased through a greater visibility of the project and therefore a much higher possible value and benefit of the project outcomes. This effect increases the more the project generates a value for any type of customer, be it a research project at the school, a company or some non-profit organization. Especially the last aspect allows to use the mechanisms and the potential of social software. In social software system the users are often inherently rewarded by gaining reputation when they create content, which is useful for others (see [12]). Thus the motivation of project participants for providing and preserving quality information and improving existing content can be increased. In section 2 we will be reviewing related publications on software support for student projects. In section 3 the analysis of the conducted surveys follows, starting with a general description of the surveys, describing their methodology and summing up results from which trends are deduced. Also further observations of the daily project support practise will be enclosed at the end of this section. Section 4 sums up the results and deduces conclusions from them. Finally section 5 will draft future work and the focus of future investigations.

2 Related Work Stanfill et al. document their experiences with version control systems and project management tools like Trac [3] for the use in capstone projects for product design teams in [8]. Also the use of wikis such as the MediaWiki is investigated. Problems and solutions with these tools are described. The acceptance of these tools and their benefit is analysed. The importance of learning effective collaboration techniques with these tools is stated, which will also be the case in this paper. Also Radermacher et al. investigate the benefits of using Subversion and Trac to improve capstone projects [9]. Munson investigates the use of wikis to create an organizational memory of projects which is also a subgoal of our efforts [7]. Grudin et al. explore the challenges in using wikis in scientific and engineering organizations, and at least a part of the results also apply in our environment [4]. Denzer describes the importance of teaching students communication skills and tools, which also applies for the supportive software systems described within this paper [1].

Recent Trends in Software Support for Online Communities

53

Another interesting aspect in which the use of collaborative, supportive online tools can greatly improve the results is their use in distributed student projects across mulitple universities, which is investigated in [2] [10]. Also this paper partially continues to investigate and substantiate some directions of our previous one in [6], which focused more on the aspect of social software.

3 Analysis The following considerations are quantitatively based on two conducted surveys, but also qualitative experiences which where gained in two years of daily project support practise are finally shared. In these two surveys (the first one carried out in February 2009, the second one in November 2010) all students of the department were invited to give their answers to an online questionnaire1 . The aim was to gather the experiences of the students with supportive systems for their projects and their requirements for the support they desire for their future projects to meet their needs as good as possible. Of course it had to be considered, that a probably significant part of the students will not be used to many supportive systems and therefore will not be able to estimate their possible value. Nevertheless both surveys allow to take a look at the level of awareness of the students in this matter, and at their demanded requirements for project support. The first insight allows to further optimize the project offerings of the department, the big advantage of the second one is that knowing which systems the students want to use themselves means to know which system will reach the highest acceptance among them. Therefore they will be used with the greatest motivation. But also this is highly dependent on the knowledge of the students about the advantages and disadvantages of different system components, so they are able to make a good choice at all. The surveys however exceeded mere technical requirements. So they also investigated the form of administration the students demanded for their project support. They were asked if some administrators should set up functions for a project on inquiry or if there should be a system where the students could do that autonomously. The acceptance of social software was also analyzed. The students were asked if they are willing to use social functions in their project support – thus generating content for other students through their project work and offering more ways of interaction between projects – and also how deeply they are interested in accessing the contents of other projects. Furthermore because the core of the questions in both surveys was identical, it is possible to conclude developments and trends in the answers by comparing the results. So the summed up results will be discussed, right after a short look at the methodology of the surveys has been taken. 3.1 Methodology Both surveys were carried out online with the open-source tool limesurvey [11]. Therefore the students where invited through mailing-lists to take part in the particular survey. 1

The results can be found at https://proanvil.inform.fh-hannover.de/surveys

54

D. Kadenbach and C. Kleiner

The first survey consisted of 30, the second of 32 questions, each within the categories: Experiences of the students, desired supportive systems, acceptance of social functions, administration of the software support and miscellaneous questions (like for example the usage of virtualisation to provide server instances for projects). Most questions of the survey could be answered with yes, no or abstention. Sometimes the students could choose between a set of predefined answers (for example when asked which version control system they prefer), but were also able to add further answers themselves. Of the about 250 active students in the two programs 60 participated in the first survey and 94 in the second one. This return rate is pretty good as the majority of students is enrolled in the first semesters of the program where no projects are executed. Thus among the students active in projects the return rate is significantly above 50%. 3.2 Results of the Surveys Detailed results for all of the questions cannot be presented due to space constraints. Thus we focus on a summary of the most interesting results here: Basic Acceptance. The surveys showed that 100% of the participants desire supportive software systems for their projects, so that such systems can expect a high acceptance. However it can be assumed that some students who would refuse supportive systems for their projects have not taken part in the surveys in the first place but their number is probably rather small compared to the ones who just did not care about it enough. But this is not really a drawback of the surveys, because supportive systems should always only be an offering for the students, they can choose to use them if they want. They should never be forced to use them in their projects, because this would very likely far decrease the acceptance of the students and render the supportive systems useless. Students should be convinced with arguments to use the provided systems autonomously. Version Control Systems. In both surveys all students voted for the use of version control systems (VCS). In the second survey the questions regarding the usage of VCS were more precise which showed that 96% of the participants even suggested the use of VCS for small projects with only one person like final theses. Furthermore it was investigated which VCS the students preferred, which is shown in table 1. The acceptance of Subversion increased while there is a slight trend towards the more sophisticated Git. Table 1. Survey: Preferred VCS Type First Survey Second Survey Subversion 56% 74% CVS 23% – Git 19% 22%

Recent Trends in Software Support for Online Communities

55

Wikis. The students where also questioned, if they would recommend the use of wikis for the documentation of projects. The acceptance rate in the first survey was at 48%, in the second at 50%. In both cases the other half of the participants abstained from voting. Again the second survey questioned deeper and the students were asked about their experiences with wikis in their previous project work. It could be seen that just the 50% who suggested the use of wikis had already worked with them in their projects, and that the other half which abstained had not gained any experiences with their use. This also illustrates how important it is that students get to know the different supportive systems, so that they can choose the right ones for their future projects. Tracker. The use of issue-tracking systems like the open-source Mantis or Bugzilla, or integrated as parts of other systems like in Trac or Redmine [5], was suggested by 66% of the students in the first survey. The acceptance increased to 85% in the second survey. Mailing-Lists. Automatically generated mailing-lists for projects were favoured by 72% in the first and 80% of the participants in the second survey. Personal Homepages. Personal homepages are the only point in which the demand was decreasing between the two surveys, it sank from 53% to 30%, which shows the lower importance of this more traditional function. Administration. The results of both surveys indicated (71% and 90%), that the students would prefer a system where they can independently choose and establish supportive systems for their projects – like creating Subversion repositories or wiki instances on demand. Usage for Personal Projects. A growing part of the students (from 69% to 93%) would appreciate it if they could use supportive tools for their own small projects, for example using a VCS and a wiki to support them in their final thesis. Social Functions. An Internet portal page, where all projects are listed, and thus which offers the possibility to find, look at and learn from projects of other students and which therefore increases the visibility and sustainability of their own projects alike, was favoured by 71% of the participants of the first and 80% of the second survey. The surveys additionally not only investigated if the students would like to see the projects of others but also if they are willing to share the contents of their own projects with other students. The positive answers increased from 61% up to 82% in the second survey. Similar high acceptance rates where reached by functions such as the following: • The use of instant messengers to support team communication. • The commenting of content of other projects and the receiving of comments for own projects in return. • The creation of user-profiles with interests and abilities to find project-partners and help. • The creation of a central knowledge base with solutions to frequently arising problems. All in all the surveys show a trend towards more experiences of the students and a greater acceptance of and demand for supportive systems as can be seen in figure 1.

56

D. Kadenbach and C. Kleiner

Fig. 1. Trends of the Surveys

Also an increasing willingness and wish to create a sustainable overvalue with their project work for other students and to profit from other students projects likewise was identified. 3.3 Further Observations Other observations were made next to the surveys in the daily project support work. So figure 2 shows the number of projects which where supported with software systems (most of them with VCS, wikis and issue tracker, or integrated systems like Trac or Redmine). A slightly more than proportional increase can be investigated over the years. Figure 3 shows the accumulated frequency of commits in the version control systems of all projects, which fluctuates depending on the current projects and project phases.

Fig. 2. Number of Projects

Recent Trends in Software Support for Online Communities

57

Fig. 3. Overview of Monthly Subversion Commits

It can be seen that project work takes place continuously, even in semester breaks, and that the systems are used extensively. A similar trend can be seen in the wiki usage. In the last year there was a stronger demand for supportive systems which integrate multiple functions like VCS, wiki, tracker, shared file storage and more in one userinterface like Trac and Redmine. Also an increased interest in Git from students and staff members has been observed, which also showed slightly in the surveys.

4 Summary and Conclusions All in all, there is a trend towards more sophisticated software support for projects. This can not only be proved by the increased number of participants in the second survey and their increased level of awareness of available tools and their handling. Also the higher acceptance of these tools by students and staff members in their daily project work demonstrates this trend. The results of the survey point out how important it is to make the students familiar with the possibilities of recent supportive tools. They do not only have to be able to handle these tools efficiently but also, which may be even more important, have to have the knowledge and experience to decide what kind of tool is the best to use in a certain situation. The survey also showed a trend towards tools such as Trac and Redmine for example. Both integrate multiple supportive functions in a single user-interface. The acceptance of Subversion is enhanced within the version control systems. Surprisingly, there was a slight increase in the interest in Git, which offers a far more sophisticated use as Subversion for example with different workflow models. Therefore, its utilization should be absolutely encouraged. Also important and carrying a great potential is the higher acceptance of social functions in project support and the increased demand for a central portal page for projects.

58

D. Kadenbach and C. Kleiner

If the projects can be presented there in an easy way to browse and search, their value will be increased beyond their development time because other students might be inspired. They could learn from conducted projects or even interact with currently carried out projects, giving comments, advice, or criticism. Therefore, through the higher visibility and possible higher benefit of a project for others, the students would be more encouraged to aim for high quality and reusability of their project artifacts, especially the documentation. In using these tools students are supported furthermore in this effort. The overall sustainability of both the project results as well as the process to achieve them is significantly increased. Finally the results of the survey show, that supporting software systems for projects can expect the highest acceptance when the students have the freedom to choose them themselves and are not forced to use a predefined set of tools or to use them at all. This increased responsibility certainly requires that the students are aware of a profound knowledge of the different types of tools available. They need to know about their use, possibilities, strengths and weaknesses. Then they can choose the right tools for their work. This implies that these tools, their attributes and handling early have to be a part of the curriculum in courses and exercises, in particular because the students can profit from their usage already very soon in their studies.

5 Future Work With these results the project support at our department shall be improved to better fit the requirements of the students and staff members. This includes giving them the facility to easily create supportive software systems like wikis, VCS, trackers and others by themselves using a web interface based on a central user-management. This also means building a portal for the projects and establishing functions of social software. All this will cause great efforts to be accomplished. Therefore it is planed to further investigate how other institutes are satisfying this needs and to work together with them striving for solutions which will be generally applicable and an enrichment for all. Additionally future investigation will be made to keep track of the always changing requirements and to gain a finer understanding which types of projects can benefit the most from which type of supportive systems. Finally an eye has to be kept open to look at other collaborative, supportive tools, new versions of used tools and newly developed ones.

Acknowledgments This paper owes various persons a debt of gratitude, particularly the students and staff of the department of computer science at the University of Applied Sciences and Arts Hannover, who participated in the surveys, stated their requirements and wishes, evaluated and used supportive functions and gave valuable feedback. Especially Oliver Bruns and Noomi M¨uller helped to improve this paper. Also the department itself which invests into the challenging research and development of a preferably optimal support for its projects earns gratitude.

Recent Trends in Software Support for Online Communities

59

References 1. Denzer, A.: Long-Distance Multidisciplinary Collaboration: Some Lessons Learned. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf. org/resources/2010%20Proceedings/index.html 2. Brereton, O.P., Lees, S., Bedson, R., Boldgreff, C., Drummond, S., Layzell, P., Macaulay, L., Young, R.: Student group working across universities: a case study in software engineering. In: IEEE Transactions on Education, p. 394. IEEE Education Society, Los Alamitos (2000) 3. Edgewall Software: Trac, http://trac.edgewall.org/ (visited, December 2010) 4. Grudin, J., Poole, E.S.: Wikis at work: success factors and challenges for sustainability of enterprise Wikis. In: Proceedings of the 6th International Symposium on Wikis and Open Collaboration, WikiSym 2010, pp. 5:1–5:8. ACM, New York (2010), http://doi.acm. org/10.1145/1832772.1832780 5. Lang, J.-P.: Redmine, http://www.redmine.org/projects/redmine (visited, December 2010) 6. Kadenbach, D., Kleiner, C.: Benefits and Challenges of Using Collaborative Development Environments with Social Software in Higher Computer Science Education. In: Ozok, A.A., Zaphiris, P. (eds.) OCSC 2009. LNCS, vol. 5621, pp. 479–487. Springer, Heidelberg (2009) 7. Munson, S.A.: Motivating and enabling organizational memory with a workgroup wiki. In: Proceedings of the 4th International Symposium on Wikis, WikiSym 2008, pp. 18:1–18:5. ACM, New York (2008), http://doi.acm.org/10.1145/1822258.1822283 8. Keith Stanfill, R., Blackwelder, E.I.: Adapting Lightweight Source Control and Project Management Software for Use by Multidisciplinary Product Design Teams. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf.org/ resources/2010%20Proceedings/index.html 9. Radermacher, A., Helsene, A., Knudson, D.: Improving Capstone Courses with Content Management Systems and Virtualization. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf.org/resources/ 2010%20Proceedings/index.html 10. Tabrizi, M.H., Collins, C.B., Kalamkar, V.: An international collaboration in software engineering. In: Proceedings of the 40th ACM Technical Symposium on Computer Science Education, SIGCSE 2009, pp. 306–310. ACM Press, New York (2009), http://doi. acm.org/10.1145/1508865.1508976 11. The Lime Survey Project Team: Limesurvey, http://www.limesurvey.org (visited December, 2010) 12. Tørning, K.: Position Paper: CSCW and the Web 2.0. In: Workshop of the Tenth European Conference on Computer-Supported Cooperative Work (2007)

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments: Recommendations for Educators and Wiki Designers Portia Pusey1 and Gabriele Meiselwitz2 1

Department of Educational Technology and Literacy Department of Computer and Information Sciences, Towson University 8000 York Road Towson, MD, 21252 USA {ppusey,gmeiselwitz}@towson.edu

2

Abstract. This paper discusses assessment practice when wikis are used as learning environments in higher education. Wikis are simple online information systems which often serve user communities. In higher education, wikis have been used in a supporting function to traditional courses; however, there is little research on wikis taking on a larger role as learning environments and even less research on assessment practice for these learning environments. This paper reports on the assessment techniques for large- and small scale- learning environments. It explores the barriers to assessment described in the studies. The paper concludes with a proposal of five improvements to the wiki engine which could facilitate assessment when wikis are used as learning environments in higher education. Keywords: Wiki, Wiki Learning, Wiki Learning Environment, Assessment.

1 Introduction Theoretically, wikis have the potential to harness the benefits of Web 2.0 technology to support collaborative learning. While not specifically designed as a learning environment, a previous review of the literature revealed that wikis have many features which enable collaborative learning to occur [1]. However, since wikis are not designed as an educational tool, a repeated weakness reported in the literature is the inadequacy of the wiki to assist assessment. Assessment is a critical part of the wiki learning process. Angelo and Cross [2] write that “the central purpose of Classroom Assessment is to empower both teachers and their students to improve the quality of learning in the classroom” (p. 4). This is because Classroom Assessment “provides faculty with feedback about their effectiveness as teachers, and it gives students a measure of their progress as learners” (p. xiv). The assessment process plays the same significant roles in wiki learning environments as in traditional learning environments – informing and improving A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 60–68, 2011. © Springer-Verlag Berlin Heidelberg 2011

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

61

instruction and evaluating student growth. However, in wiki learning environments, assessment also serves to ensure students participate. This is because there is evidence that indicates that students will not contribute fully to the wiki learning experience if they are not being graded for their participation [3, 4, 5, 6, 7, 8, 9]. The authors of this work have taught five semesters in a mid-Atlantic college of education using wikis – in classes as small as 12 and combined groups as large as 180 students. They have used best practice for implementing wikis for instruction but continue to find assessment a time consuming process – this limitation of using wikis for instruction is an often reported barrier to implementation [5, 10, 11, 12, 13]. This work represents a synthesis of examples and recommendations from the literature of assessments used for wiki. Assessment has not been the focus of many research studies; however, many reports and studies mention the methods they used to evaluate student progress. After identifying these assessment methods, the barriers to effective assessments in wiki learning environments will be listed. Many of these barriers can be overcome with improvements to the wiki engine. Therefore, this work will conclude with recommendations to improve wiki engines to better support assessment practice.

2 Wiki Assessment in the Literature Wikis are well suited to fulfill the needs of instructors looking for a collaborative, student-centered learning environment. They provide dynamic repositories that allow students to learn by constructing knowledge based on their experiences in a selfdirecting manner; becoming active contributors to each other’s learning; and developing a sense of community ownership of this content [14, 15, 16]. However, wikis were not designed as learning environments for formal education. Therefore, instructors and wiki researchers have used traditional method of assessment to evaluate student work. Assessment methods have rarely been the focus of a research study so this list represents the types of assessment referred to in the studies on wikis published in peer review journals over the past 5 years. Many of the studies found in the body of literature on instructional wikis are qualitative; many of the journal articles included a reference to the type of assessment used. In the descriptive studies, traditional assessments using quizzes and rubrics were used; peer assessment also was frequently used by instructors. Another method that was evaluated quantitatively as well as qualitatively was the use of the wiki usage data. The qualitative studies explained methods instructors used to attribute an individual’s contribution to group work. One of the qualitative studies attempted to automate the assessment process using the portfolio record kept in the wiki [11]. While there was no consensus in the literature about the best method to use, there was wide agreement that assessment in wiki learning environment remains a time consuming process. 2.1 Quizzes Quizzes have also been used to assess knowledge gained from the creation and use of the wiki. Formats suggested by Hazipanagos & Warburton [17] include multiple choice, short answer, true/false, and matching. In many cases students created the quiz questions and answers [17, 18, 19]. In these studies, and in research by the

62

P. Pusey and G. Meiselwitz

authors of this work, students write the questions and answers. They submit the questions and answers to the instructors who create an automated quiz using a quiz tool or course management system such as Blackboard. Quizzes can be taken without the aid of the wiki; or, when students can use the information in the wiki, we have called the “quiz” a scavenger hunt. Quizzes and scavenger hunts are tools that can be used formatively to improve or drive instruction. They are used summatively when the goal is to improve knowledge or awareness of the wiki content. The benefit to using a quiz, especially an online quiz, is that the feedback to the student can be immediate. However, wikis can be used for more than generating knowledge and awareness and at least one wiki researcher called for a change in practice away from summative, norm referenced tests to meet the demands of the authentic project-based learning that occurs online [20]. 2.2 Rubrics Several of the studies used rubrics to assess their students [4, 18, 21]. The use of rubrics to assess authentic activities reflects the evidence in the body of literature that rubrics are effective tools to assess complex authentic student work [22, 23,24 25, 26, 27].The use of rubrics to assess student work was recommended by one research because rubrics allow the instructor to set clear expectations up front [28]. Best practice on the use of rubrics underscore the rubrics’ ability to inform student work as it is being created when the rubric is given to students prior to instruction [29]. When rubrics are given to students when learning activities are assigned, students are able to review their own work for errors prior to submission for feedback or grading [30]. The authors have used rubrics to create an all-in-one assessment that not only assessed the content and quality of the student’s contribution but also the quantity and quality of their collaborative posts. Rubrics are used both summatively and formatively to assess multiple facets of a student project. 2.3 Peer Assessment Since wiki is a social learning experience a social evaluation tool has been often used. Several studies reported using peer assessment [11, 18]. One study, which used several semesters and repeated measures, reported peer assessments to be equivalent to instructor assessments [19]. Irons [31] recommend using peer feedback as a way to improve student learning and reduce the time commitment on the part of the instructor. In one account, students completed the wiki learning activity over several class periods [18]. This study reported that students anonymously evaluated each team member’s, as well as their own, contribution to the work at the end of each day. The author reported that this self and peer rating system valued each students’ contribution. Peer assessment also can contribute to interaction between students. In a study that reported on 5 semesters of teaching with wikis in a college of education [11] this study suggested that peer assessment encouraged interaction between the author and the other students. In this study, feedback was provided to the authors by the other students in the class using the threaded discussion area of the page. By participating in the conversations occurring in the threaded feedback the instructor has an opportunity to monitor and assess the learning as well as the

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

63

collaborative experience [12]. The authors also use the threaded discussion area as a formative feedback forum for students. In addition to the improved interaction/ collaboration and improved student learning (evidenced in other studies) the authors also found the threaded feedback discussion a wealth of information about student learning and understandings about the content. Studies suggest that students may not be able to provide quality feedback without being trained to do so [22, 23, 24, 25, 26, 27]. So, a tutorial or practice is beneficial to be sure the peer assessment will reflect the instructor’s expectations. Formative peer assessment when conducted in the threaded discussion, is more time consuming than if the instructor were to complete this evaluation herself. This is because the instructor not only needs to read and provide feedback to the author; the instructor also is assessing the student feedback to the author. However, the threaded discussions also allow the instructor the ability to participate in the discussions occurring among students and remediate any misconceptions and inaccuracies as they are occurring. Summative peer feedback, by well trained students, has been reported as an effective means to evaluate students while reducing instructor workload [28]. 2.4 Wiki Usage Data A final method of assessment, which is most frequently reported in the literature, is to use a student’s quantity of work as part of his final grade. In most cases, studies report that students are not only graded on the quality of the content they are producing but on the quantity of their contribution to the end product. Wiki usage data has been used in several studies to provide the instructors with a portfolio of information that tracks posts to individual users [5, 11, 12, 28, 32, 33]. The literature on assessment for wikis used as learning environments suggests that while students should work collaboratively, they should be assessed individually [7]. Several studies reported higher student satisfaction and a perception of fairness in the assessment of group work because of the capability of individualized assessment in wiki learning [34, 35, 36, 37, 38, 39, 40, 41]. However, there are concerns that have surfaced in the literature that the usage data may not accurately reflect the group process [4, 42]. First, quantities of posts are not an indication of learning; and second, the quality of the post is a more important consideration as evidence of learning. Infrequent posts do not necessarily indicate that a student is not learning [3, 42]. Student teams may choose to collaborate face-to-face and designate one student to post their work. And qualitative studies have indicated that students read and re-read wiki content without contributing to the conversation [43]. Paradoxically, one study hypothesized that if more in class face-to-face time had been allocated to the wiki activity, perhaps greater evidence of collaboration would have been seen in the wiki learning environment [44]. Therefore, the recommendations found in the literature are that the usage data should only be used as a small part of the assessment process.

3 Recommendations for Practice The authors of this work contend that the barriers to effective assessment stem from the fact that wikis were never designed to be used as a learning environment. Therefore, several changes to the wiki engine need to be considered before effective and efficient assessment can be designed and tested.

64

P. Pusey and G. Meiselwitz

3.1 Facilitate Communication within the Wiki Wikis offer several modes of communication – the threaded discussion associated with each content page, email, and direct edits to the content page. However, the only way to provide private feedback to students is to use email. The authors of this article have found that the email tool within some wikis often require several clicks to communicate directly with students. Furthermore, email communications within some wiki engines leaves no record of the private conversations. Therefore, creating a communication system within the wiki that allows for private conversations to occur would be helpful in building a virtual community [17]. An effective private way of communicating with students would also resolve some of the frustrations suggested by another study which reported student frustration and confusion because multiple learning platforms are used [5]. The students in this study used course management software for most tasks and used the wiki to communicate with their tutors. In order to encourage students to use the wiki platform, the authors of this study suggested that the instructor should refuse to answer questions by email [46]. However, improvements to the wiki email to link directly into the email systems that the instructors and students currently use would be a better solution. 3.2 Comment Tool Learners benefit when they are provided frequent and detailed feedback [10, 40]. In one study English as a Second Language students required a large amount of immediate feedback [5]. Zorko [5] reported that the ability to react quickly when there was a problem and to provide guidance was motivating for his students. Furthermore, Zorko reports that students can respond to the comments more quickly and produce a better end product. Currently, all comments must be made to students using the threaded discussion, direct edits to the page, or in email. A comment tool, similar to the one used in popular word processing revision tools, would facilitate the feedback process. First, it would enable the instructor to provide targeted feedback directly linked to sentences, paragraphs, or words. Therefore, when the instructor is providing feedback they no longer need to identify where, in the text, the feedback is targeted. Second, it would enable peers to provide feedback without directly editing the page. Students don’t like other students directly editing their pages [3, 4]. Formatting issues have occurred, and changes have been undone [43]. Furthermore, students have reported that they are reluctant to edit each other’s work [9]. Therefore, a targeted comment --using a callout box or stickynote type tool – would allow students to suggest changes to the page without actually making them. 3.3 Colored Instructor Comments The threaded discussion area is still a valuable source of evaluation data. By participating in the conversations occurring in the threaded feedback the instructor has an opportunity to monitor and assess the learning as well as the collaborative experience [12]. Several studies reported on the use of peer assessment in threaded discussions to help learners improve the content of their wiki page [17, 19]. Using the threaded discussion section is very helpful for students to provide feedback and to enable them to defend their suggestions [12]. However, one study reported the

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

65

concern that in the large body of information and threaded discussions it was feared that instructor comments would get lost and students would not see the feedback [32]. If the instructor comments appeared in a unique color they would be easily identifiable by the students and would not get lost. 3.4 Connections to Plagiarism Checking Software It is easy for students to cut and paste content directly from their web browser into their wiki content pages. Several studies reminded students not to directly cut and paste into the wiki [3, 3]. In fact, in one study, students expressed concern about material copied directly from books and Internet sites [43]. With so much of wiki content being text-based, a tool that checked pages for plagiarism within the wiki would be ideal. Sites such as Turnitin are used by educational institutions to check their students work for originality (http://turnitin.com/static/index.php). As a first step, giving instructors the ability to test content pages with a single click of the button would help make the assessment process more efficient. However, providing students access to this tool would also serve as a metacognitive tool to improve their awareness of plagiarism and the need for proper citation. 3.5 Integrated Rating Tool It has been suggested that when students have been properly trained to use the tools, peer assessment can be equivalent to instructor assessments and can improve end products [19, 22, 45, 46, 47]. Furthermore, peer feedback, especially formative feedback, can reduce the time commitment on the part of the instructor [31]. However, wikis only currently facilitate qualitative feedback. The qualitative feedback is an integral part in helping students improve their contributions to the wiki, however most grading systems are quantitative (A-F, 0-4, 25 points, etc.). Therefore, the addition of a flexible rating tool would aid in the assessment process. The proposed tool would work similarly to the feedback systems in many online shopping and auction sites which use a star system to accumulate buyers’ feedback into information about the reliability of the seller or product. The system would link directly to a rubric and accommodate multiple ratings based on the number of subtasks being assessed [29]. The rating scale should be flexible to accommodate local assessment scales – i.e. ranges of letters and numbers. Finally, to facilitate best practice for peer assessment the system should allow for anonymous ratings [27, 47, 48]. Of course, a reporting tool for the instructor, which would export the rating into a spreadsheet, would also be necessary.

4 Conclusion This work is part of a larger study that evaluates the use of wikis to change attitudes in preservice teachers. It began in 2009 with a review of best practice for implementation of wiki as a learning environment [1]. After five semester of implementing wiki learning environments it was clear that the best practice found in the literature could help an instructor successfully facilitate collaborative work. However, the struggle to effectively and efficiently assess collaborative groups up to 180 students led these authors to the literature for answers.

66

P. Pusey and G. Meiselwitz

Attempts have been made to automate the assessment process in a wiki but these attempts were not successful, therefore, the assessment process is still a laborintensive process on the part of the instructor [4, 41]. The examples from the literature quizzes, rubrics and peer assessments are useful for effective formative and summative assessment but the processes are not efficient ones. Therefore, we suggest improvements to the wiki engine to better facilitate instructor and peer assessment. We suggest improved communication tools -- private conversations between instructors and students, colored instructor comments, and a commenting tool -- a flexible rating tool that directly connects to the assessment rubric and exports reports to spreadsheets; and plagiarism checking within the wiki. Whether these improvements to the wiki engine are ever implemented, future work should focus on empirical work that evaluates assessment practice so that instructors have complete body of work on best practice for implementing wikis as learning environments.

References 1. Pusey, P., Meiselwitz, G.: Development of Heuristics for Implementation of Wiki technology as Learning Environments in Higher Education. In: Stephanidis, C., et al. (eds.) 2009 Human Computer Interaction Conference Proceedings, pp. 507–517. ACM, New York (2009) 2. Angelo, T.A., Cross, K.P.: Classroom Assessment Techniques: A Handbook for College Teachers, 2nd edn. Jossey-Bass, San Fransisco (1993) 3. Wheeler, S., Yoemans, P., Wheeler, D.: The Good, the Bad, and the Wiki: Evaluating Student-Generated Content for Collaborative Learning. Brit. J. of Ed. Tech. 39, 987–9956 (2008) 4. Lin, H., Kelsey, K.D.: Building a Networked Environment in Wikis: The Evolving Phases of Collaborative Learning in a Wikibook Project. J. of Ed. Computing Research 40, 145– 169 (2009) 5. Zorko, V.: Factors Affecting the Way Students Collaborate in a wiki or English Language Learning. Australasian J. of Ed. Tech. 25, 645–665 (2009) 6. Choy, S.O., Ng, K.C.: Implementing Wiki Software for Supplementing Online Learning. Australasian J. of Ed. Tech. 23, 209–226 (2007) 7. Cubric, M.: Wiki-Based Framework for Blended Learning. In: Proceedings International Symposium on Wikis, pp. 11–24. ACM, Montréal (2007) 8. Ebner, M., Kickmeier-Rust, M., Holzinger, A.: Utilizing Wiki-Systems in Higher Education Classes: A Chance for Universal Access? Universal Access in the Information Society, vol. 7, pp. 199–207 (2008) 9. Tal-Elhasid, E., Meishar-Tal, H.: Naples Models for activities, collaboration and assessment in wiki in academic courses. In: Electronic Proceedings of the Eden 2007 Annual Conference (2007), http://www.biu.ac.il/bar-e-learn/eden2007/tal_tal.doc 10. Beebe, R., Vonderwell, S., Boboc, M.: Emerging Patterns in Transferring Assessment Practices from f2f to Online Environments. Electronic J. of e-Learning 8, 1–12 (2009) 11. Trentin, G.: Using a Wiki to Evaluate Individual Contribution to a Collaborative Learning Project. J. of Comp. Asst. Learning 25, 43–55 (2009) 12. Campbell, K., Ellingson, D.E.: Cooperative Learning at a Distance: An Experiment with Wikis. American J. of Bus. Ed. 3, 83–89 (2010)

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

67

13. AL-Smaki, M., Guetl, C., Kappe, F.: Peer Assessment System for Modern Learning Settings: Towards a Flexible E-Assessment System. Int. J. of Emerg. Tech. in Learning 5, 5–11 (2010) 14. Reinhold, S.: Wiki trails: Augmenting Wiki Structure for collaborative interdisciplinary learning. Paper presented at WikiSym 2006 (2006) 15. Schaffert, S.: IkeWiki: A Semantic Wiki for Collaborative Knowledge Management. In: Enabling Technologies: Infrastructure for Collaborative Enterprises. IEEE, Los Alamitos (2006) 16. Wang, C., Turner, D.: Extending the Wiki Paradigm for Use in the Classroom, pp. 255– 259. IEEE Press, New York (2004) 17. Hatzipanagos, S., Warburton, S.: Feedback as Dialog: Exploring the Links between Formative Assessment and Social Software in Distance Learning. Learning, Media & Tech. 34, 45–59 (2009) 18. Tharp, T.L.: Wiki, Wiki, Wiki – What? Assessing Online Collaborative Writing. English J. 99, 40–46 (2010) 19. Ren, Z., Baker, P., Zhang, S.: Effects of Student-Written Wiki-Based Textbooks on PreService Teachers’ Epistemological Beliefs. Ed. Comp. Res. 40, 429–449 (2009) 20. Williams, P.: Assessing context-based learning: Not Only Rigorous but also Relevant. Assmt. & Eval. In Higher Ed. 33, 395–408 (2008) 21. Lending, D.: Teaching Tip Using a Wiki to Collaborate on a Study Guide. J. of Information Systems Ed. 21, 5–13 (2010) 22. Bloxham, S., West, A.: Understanding the Rules of the Game: Marking Peer Assessment as a Medium for Developing Students’ Conceptions of Assessment. Assmt. & Eval. in Higher Ed. 29, 7333–7721 (2004) 23. Prins, F.J., et al.: Formative Peer Assessment in a CSCL Environment: A Case Study. Assmt. & Eval. in Higher Ed. 30, 417–444 (2008) 24. Sluijsman, D., Prins, P.: A Conceptual Framework for Integrating Peer Assessment in Teacher Education. Studies in Ed. Eval. 32, 6–22 (2006) 25. Sluijsmans, D.M.A., et al.: Training Teachers in Peer-Assessment Skills: Effects on Performance and Perceptions. Studies in Ed. Eval 41, 59–78 (2004) 26. Sluijsmans, D.M.A., et al.: The Training of Peer Assessment Skills to Promote the Development of Reflection Skills in Teacher Education. Studies in Ed. Eval 29, 23–42 (2003) 27. Wen, M.L., Tsai, C.C., Chang, C.Y.: Attitudes Towards Peer Assessment: A Comparison of the Perspectives of Pre-Service and In-Service Teachers. Innov. in Ed. & Teaching International 43, 83–92 (2006) 28. Minocha, S., Thomas, P.G.: Collaborative Learning in a Wiki Environment: Experiences from a Software Engineering Course. New Review of Hypermedia & Multimedi 13, 187– 209 (2007) 29. Castle, S., Arends, R.: Developing credible performance assessments. In: Castle, S., Shaklee, B.D. (eds.) Assessing Teacher Performance. Rowman & Littlefield Education, Lanham (2006) 30. Andrade, H., Du, Y.: Student perspectives on rubric-referenced assessment.. 3. Practical Assessment Research & Evaluation 10, 159–181 (2005) 31. Irons, A.: Enhancing learning through formative assessment and feedback. Routledge, New York (2008) 32. Elgort, I., Smith, A.G., Toland, J.: Is Wiki an Effective Platform for Group Course Work? Australasian J. of Ed. Tech. 24, 195–210 (2008)

68

P. Pusey and G. Meiselwitz

33. Slovie, P.A.: Use of the Wiki: Encouraging Preservice Teachers’ Construction of Knowledge in Reading Methods Courses. J. of Literacy & Tech. 9, 0975–1535 (2008) 34. Bruns, A., Humphreys, S.: Building Collaborative Capacities in Learners: The M/Cyclopedia Project, revisited. In: Proceedings International Symposium on Wikis, pp. 1–10 (2007), http://www.wikisym.org/ 35. DePedro, P.: New Method Using Wikis and Forums to Evaluate Individual Contributions in Cooperative Work While Promoting Experiential Learning: Results from Preliminary Experience. In: Proceedings International Symposium on Wikis, pp. 87–92. ACM, Montréal (2007) 36. Clark, D.J., Redmond, M.V.: Small group instruction diagnosis: Final report. Seattle, WA: University of Washington Seattle. ERIC Document Reproduction Service No. ED217954 (1982) 37. Hassaien, A.: A Qualitative Student Evaluation of Group Learning in Higher Education. Higher Ed. in Europe 31, 138–150 (2007) 38. McGraw, P., Tidwell, A.: Teaching Group Process Skills to MBA Students. In: A Short Workshop. Ed. & Training, vol. 43, pp. 162–170 (2001) 39. Elliott, N., Higgins, A.: Self and Peer Assessment – Does it Make a Difference to Student Group Work? Nurse Ed. in Practice 5, 45–48 (2004) 40. Gupta, L.M.: Enhancing Student Performance Through Cooperative Learning in Physical Sciences. Assmnt. in Higher Ed. 29, 63–73 (2004) 41. Armellini, A., Aiyebayo, O.: Learning Design and Assessment with E-tivities. British J. of Ed. Tech. 41, 922–935 (2010) 42. Wang, H.C., et al.: An empirical exploration of using Wiki in an English as a second language Course. In: Proceedings of the 5th IEEE International Conference on Advanced Learning Technologies, pp. 155–157. IEEE, Los Alamitos (2005) 43. Matthew, K.I., Felvegi, E., Callaway, R.A.: Wiki as a Collaborative Learning Tool in a Language Arts Method Class. J. of Res. On Tech. in Ed. 42, 51–72 (2009) 44. Judd, T., Kennedy, G., Cropper, S.: Using Wikis for Collaborative Learning: Assessing Collaboration Though Contribution. Australasian J. of Ed. Tech. 26, 341–354 (2010) 45. Hubert-Williams, N.: Facilitating Collaborative Learning using Online Wikis: Evaluation of their Application within Postgraduate Psychology Teaching. Psych. Learn. & Teach. 9, 45–51 (2010) 46. Schneider, D.K., et al.: Conception and Implementation of Rich Pedagogical Scenarios Through Collaborative Portal Sites: Clear Focus and Fuzzy Edges. In: International Conference on Open and Online Learning (2003), http://tecfa.unige.ch/proj/seed/catalog/docs/icool03schneider.pdf 47. Tsai, C.C., Liu, E.Z.F., Lin, S.S.J., Yuan, S.M.: A Networked Peer Assessment System Based on a Vee Heuristic. Innov. in Ed. & Teaching Int. 38, 220–230 (2001) 48. Freeman, M.: Peer Assessment by Groups of Group Work. British J. of Ed. Tech. 20, 289–299 (1995)

Teacher Agents: The Current State, Future Trends, and Many Roles of Intelligent Agents in Education Kevin Reed1 and Gabriele Meiselwitz2 1 Department of Information Technology, Northern Virginia Community College, Alexandria Campus, Bisdorf 352, 3001 N. Beauregard St., Alexandria, VA 22311 USA 2 Computer & Information Sciences Department, Towson University, 8000 York Rd. Towson, MD 21252 USA [email protected], [email protected]

Abstract. Since their development in the 1980's, Intelligent Tutoring Systems (ITS) have experienced a widespread success in such varying areas of education as military training, personal tutoring, and vocational instruction[9], [12], [19], [36]. ITSs are not without limitations, however, and have often proven to be costly and inflexible. Combining these systems with Intelligent Agents (IA), first proposed in the 1990s, is intended to address some of the shortcomings of ITSs; notably the cost of building new learning objects. While IA provide a mechanism for generating dynamic content tailored to a specific learner, a lack of standardization in IA ontologies and a narrow focus on pedagogy provides rich veins for reseach. In this paper we broadly survey the development of IAs in education with an eye towards further exploration of their possibilities. Keywords: Intelligent agents, education, chatbot, tutoring systems.

1 Introduction Speaking at the 2010 commencement of Hampton University, U.S. President Barack Obama remarked First and foremost, your education can fortify you against the uncertainties of a 21st century economy. In the 19th century, folks could get by with a few basic skills, whether they learned them in a school like Hampton or picked them up along the way. For much of the 20th century, a high school diploma was a ticket to a solid middle class life. That is no longer the case. Jobs today often require at least a bachelor's degree...[2] Mr. Obama's observations iluminate a movement that has been emerging since the century mark. U.S. Census data shows that college enrollments nationwide are up 17 percent since 2000, while college costs continue to climb rapidly [42] and government support for higher education continues to decline. Of particular concern in these diverging trends is the fact that many more of the students enrolling in college, and particularly community colleges, since 2000 require intervention services and remediation in the core subjects of mathematics, reading, and writing [31]. Rising A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 69–78, 2011. © Springer-Verlag Berlin Heidelberg 2011

70

K. Reed and G. Meiselwitz

enrollments, too few instructors, stagnant budgets, and students poorly prepared for the rigors of a college education produce a toxicity that must be diluted if the U.S. is to retain its competitiveness in the global economy of this century. The traditional method of addressing student deficiences and individual needs has been the personal tutor, generally an individual who is both a subject domain expert, and takes the time to get to know his or her students personally, so that instruction can be matched to the student's particular learning style and goals. With increasing success, this role has been shifting (or at least supplementing) from human to human interaction to human to machine sessions; starting in the early 1960's [9]. Initially, these systems, typically known as Computer Aided Instruction (CAI), automated routine testing and drilling, with little differentiation between their operation and earlier wholly mechanical systems -- except they ran on computers. But advances in the fields of artificial intelligence and computer hardware, along with evolving perspectives in education and cognitive psychology, enabled ambitious researchers in the 1970s to model and develop 'intelligent' computer based instruction, commonly referred to as 'Intelligent Tutoring Systems' (ITS) [9].

2 Intelligent Tutoring Systems: Design, Characteristics, and Types The term 'Intelligent Tutoring System' is expansive in scope, entailing any computer application that possesses intelligence and is used in instruction. Generally, an ITS should be comprised of one or more autonomous software agents (discussed in greater detail below) that interact with students, present content, and provide assessment, without the immediate oversight of human instructors [9] [11]. ITSs have been constructed from a variety of languages and tools, and run on different platforms; and vary in the degree of intelligence they exhibit [11]. All of them, however, share specific architectural characteristics: the domain or expert model, the student model, the tutor model, and the interface or environment model. [9], [17]. Each of these is a component of a larger system. 2.1 Design Domain Model. The domain or expert model contains the knowledge, behaviors, and formalisms that make up the content being taught, both declarative (facts) and in some measure, procedural (processes for manipulating facts). It is a repository of the rules and inferences a subject-matter expert would follow would to solve a particular problem, such as doing addition in mathematics or diagnosing a breakdown of an automobile air conditioning system. Called an 'expert system' in artificial intelligence (AI), many researchers also refer to this component as the 'cognitive model', as there is an implicit relationship between the content and the 'knowledge state' of the expert working through the problem. The knowledge state, in turn, are those points transitioned to along the path in which the expert possesses increasing understanding of the task at hand. Student Model. The student model tracks and saves the active knowledge states of the student user in constructs such as Bayesian networks, so that progress can be

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

71

mapped. This component provides the student with a map of his or her learning, and dovetails into the tutor model. Student models have three tasks ([45] as cited in [41]): 1.

2.

3.

By explicit (asking questions) or implicit (tracking the student's navigation) means, they must collelate data about the student and compare it to other student responses. This data must be used to map the student's knowledge and learning. Taking the form of 'buggy' models (representation's of the student's knowledge as deriviations from established 'expertise'), the model will attempt to predict future student responses. These are compared to the actual responses, and a more precise map of the student's knowledge states can then be made. The model must further diagnose the current state of the student's knowledge, so that optimal learning strategies can be chosen to present subsequent domain information for mastery.

Tutor Model. Wenger stated that when 'learning is viewed as successive transitions between knowledge states, the purpose of teaching is accordingly to facilitate the student's traversal of the space of knowledge states' ([45] as cited in [41]). The tutor model, which Wenger called 'pedagogical expertise', monitors the progress captured in the student model, and intervenes in situations where there are discrepancies between the rules established in the domain model. Aptly named, it is the role of the tutor model to manage adaptive teaching strategies, such as learning path maps and recommendations; sending the student remedial instruction and positive reinforcement through the interface model as warranted. These goals are addressed through two modes of support states ([45] as cited in [41]): 1. 2.

Diagnosis, in which the ITS extrapolates state information from the student's observable behavior, knowledge level, and learning style Didactics, in which an appropriate curriculum is dynamically created and adjusted to match the student's current level.

Interface Model. Not surprisingly, the interface model is the means through which an ITS user interacts with the system – both human student and human instructor. The complexity or 'richness' of the interface is dictated by the detail needed for the student to view and complete a problem, and for human monitor to administer the system, making refinements to the domain model as needed. Wenger suggested that the Interface must consist of two component ([45] as cited in [41]): 1. 2.

The discourse model, which detects abiguities in student answers and provides a corrective responses. Knowledge presentation, which provides guidance to keep the student from missing key elements of the knowledge being conveyed.

2.2 Characteristics Woolf et al., 2001 [46] describe six characteristics or abilities of ITSs. They must: • generate 'appropriate instructional material' − This quality distinguishes intelligent tutors from computer aided instruction, which uses a finite set of problems and responses.

72

K. Reed and G. Meiselwitz

• model student performance − Assessments need to be formative and immediate, so that a catalog of student strengths and weaknesses can be created and used to update instructional material and provide feedback. Formative assessments are used to determine if students are learning the material currently being presented. They are intended as a 'feedback loop' for the instructor. This type of assessment can be contrasted with summative assessements, which are generally used to measure a student's overall performance with the objective of assigning a grade or mark. [16] • model expert performance − This is the location of the tutor's knowledge base, which gives the tutor domain intelligence. • model pedagogical strategies − Provides the tutor with the adaptibility to accommodate students at various levels of learning. To accomplish, the tutor must be 'aware' of a student's current knowledge state and learning style; and possess the ability to define learning strategies dynamically. • model natural dialog − A tutor must at least give the appearance of natural communication. (This is being realized now through developments in natural language processing). • self-monitor performance − A tutor should be capable of reflective activity that will improve its 'teaching' over time. Interestingly enough, Woolf et.al., stated that every ITS should possess at least one of these qualities; however it seems more likely that every truly intelligent tutor should possess all of them. Domain knowledge and student tracking are features of CAIs, necessary but insufficient for genuine teaching.1 Reflecting on the ongoing processes, developing strategies for aiding students; in effect 'getting to know' the learner are essential virtue for any teacher. There is a clear and self-validating relationship between Woolf et.al.'s list of ITS features and the architecture of these tutors that has evolved over the last twenty years, as shown in figure 1. 1

Meeriam and Caffarella identify four distinctive theories of education ([22] as cited in [35]): 1. Behaviorist: Focused on skill development and training, views the learning process as tantamount to behaviorial change 2. Cognitive: emphasizes 'learning how to learn', sees learning as an internal mental structuring 3. Humanistic: defines the learning process as an act of self-fulfillment and stresses selfdirected learning 4. Social/situational: proposes that learning is the relationship that evolves between student and environment, and that the role of teaching is to create or identify communities of practice. In addition, an often referenced theory of learning in ITS literature is constructivism. Based on the work of Vygotsky and others, constructivist-based learning articulates a model similar to that of the social/situational theory: learning is a process of 'enculturation' into a knowledge community [35].

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

73

Fig. 1. Relationship between Woolf et.al.'s list of ITS features

Intelligent Agents. In a typical ITS, these abilities are manifested by a collection of autonomous but tightly coupled intelligent agents. A working definition of 'intelligent agent' can be elusive, because they are the subject of extensive research and reporting. Generally, though, they can be described as autonomous enities that observe and impinge upon their environment, which is a combination of application, data storage, knowledge source, and network. They are rational, goal directed agents [32], and usually software constructs [3]. 2.3 ITS Types ITSs can further be classified by the cognitive models they employ: Cognitive or Constraint Based [37]. Constraint Based, often referred to as the Behavioral approach, has its roots in the work of B.F. Skinner and other behavioral psychologists and identifies learning in terms of 'cause and effect' chains [8] or constraints [25]. An example of such a constraint would be: if is true, then <satisfaction condition> must be true, otherwise something is wrong. This learning model is focused on an immediate, declarative problem and has no contextual awareness beyond the chain. Consequently, it is easier to implement as a rules based tutor, but is limited in its ability to provide feedback and reinforcement to the learner [37]. The Cognitive model is based on the theories of John Dewey, a philosopher of education, and the research of educational pyschologists Lev Vygotskey, Jean Piaget, and others [8]. It is holistic in approach and views learning as both declarative (facts) and procedural knowledge construction in a social context. Within the framework of an ITS, rules built on this model could take the following path [37]: if the goal is to classify a shape and the shape has three sides then classify the shape as a triangle because triangles have three sides Facts are easy to code, procedure is more challenging. However, the cognitive model has the clear advantage in its inherent feedback loop.

3 An Additional Classification We observe that there is a further classification of ITSs, and specifically those that constructed on the cognitive model. This additional subdivide is predicated on the

74

K. Reed and G. Meiselwitz

source and substrate of their knowledge domain. While hardly trivial, there are domains in which declarative knowledge is well established and the procedures for deriving and manipulating facts are well known. Disciplines such as mathematics and the sciences are in this area. Their rules and knowledge construction can be described by human experts and crafted into ontologies2 that can be used to built a knowledge base for an ITS. It's no coincidence that the most successful and widely known tutors in existence today provide training in mathematics and the sciences. Projects discussed in the literature include the PUMP Algbra Tutor Project, a partnership between Carnegie Mellon University and the Pittsburgh School System; SHERLOCK, a U.S. Air Force initiative to train jet plane mechanics; AutoTutor, a Physics instruction system at the University of Memphis; and CIRCSIM-Tutor, a natural language ITS project at the Illinois Institute of Technology, which teaches 1st year Medical students about circulatory system pressure. These systems are 'successful' because they encompass clearly defined rules and facts [9] [11]. In other domain areas in which content and knowledge is more unstructured, constructing an ontology is more difficult [28]. This is both the opportunity and challenge of the World Wide Web. An almost unlimited amount of data can be found there, but its minimal organization makes it hard to locate and use information efficiently. This is the issue that the 'Semantic Web'3 is intended to redress [13], though the degree of uncertainty associated with web content makes this an arduous undertaking [18]. Still, the web is an unequalled source of material, and can be mined as a knowledge base for an ITS. We propose, then, to prototype an ITS that helps teach adult learners to read by using content from the web as source material. Reading is a complex activity that typically takes years to master, but is a foundational skill in education and in the workplace. Without this skill, young adults can never hope to obtain the educational levels that President Obama stated are 2

Ontologies are formal representations of knowledge within a domain. They define a vocabulary and semantics to describe the concepts within a domain and the interrelationships between those concepts [13]. More concretely for this discussion, an ontology is a text-based construct of reference/knowledge that is built with an ontology representation language by a domain expert; and can be consulted by intelligent agents as part of their knowledge base [10]. They are conceptually similar in purpose to XML. A simple example, decribed in the OIL (Ontology Inference Layer) ontology representation language, would be [10]: class-def defined herbivore subclass-of animal, NOT carnivore slot-contraint eats value-type plant OR (slot-constraint is-part-of has-value plant) A number of ontology representation languages have been created, with perhaps the most 'successful' being OWL (Web Ontology Language), which became a W3C recommendation in February 2004 [20]. Typically, domain experts will use ontology authoring tools such as Protege-2000 (http://protege.stanford.edu/) to create and modify ontologies, as their syntatic complexity makes hand coding error prone and tedious. Developing ontologies is time consuming and very expensive [23], and so a significant amount of research is ongoing to find ways to automate the process [37]. 3 The term 'Semantic Web' was coined by Tim Berners-Lee and describes the technologies and mechanics that will allow machines, and specifically intelligent agents, to understand the meaning, the 'semantics', of information on the web [4]. This is an area of much active research.

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

75

essential in his commencement speech. Nevertheless, this remains an ongoing problem in post secondary education. A 2004 U.S. Department of Education study reported that 42% of 1st year students in public two-year institutions needed remediation [38]. At Northern Virginia Community College, the second largest school of its kind in the nation, and situation in one of the country's best educationed regions; over 60% of traditionally aged college students require remediation in reading [24]. There are numerous reasons why so many students come out of high school reading poorly. Social promotions and lack of adequate diagnosis in K12 impact student performance [27]; as do school district retention efforts (school systems 'dumb down' their programs to reduce student dropout). Additionally, a reluctance to inflict a sense of 'low self-esteem' on students [5]; and the simple fact that more students are enrolling in colleges today than in previous generations [43] all contribute to a situation that now needs to be addressed on an ongoing basis. The use of a machine based tutor can help in aleviating this educational and social problem, by providing reading level appropriate material and assessments for students to practice. There are currently a number of machine based tutoring systems with ITS characteristics, but none appear to use the web as source of reading material. The use of 'canned' material can be problematic, as the content may be of no interest to the student and subsequently discourage additional practice. [29], [30].

4 Towards an Intelligent Reading Tutor We envision a Cognitive model ITS in which remedial students select topics of personal interest, such as Sports news, or that have some utility, like want ads. A repository in the ITS would then be checked for matching material and, if found, presented to the student. Otherwise, the system would dispatch an agent to retrieve candidate documents from the web and place them into a database. This process would be ongoing. In the database the documents would be checked for reading level difficulty and semantically annotated with the ITS's 'priming' ontology – a basic set of patterns and responses that will provide a substrate for subsequent assessments. Because we are not attempting to map complex domain knowledge about the document contents, we believe that machine annotation will be possible. The material will then be annotated with Artificial Intelligence Markup Language (AIML), an XML compliant language that is used to create life-like dialog for an A.L.I.C.E chatbot [1]. We expect that this will enable the ITS to use a chatbot4 for assessment, which will be informal and conversational. A possible use scenario would be: 1. Student selects reading material, which is fetched from database 2. The reading material is preannotated with agent generated AIML statements that ask the most fundamental of questions, such as 'Is this article about X' and 'When do the events talked about happen?' 4

Chatbots, chatterbots, and conversational agents are 'talking' computers with antecendents in the research of Alan Turing. They are software constructs that, through various mechanisms used to acquire the appearance of natural language understanding, provide an interactive conversation with human users. The first actual implementation was 'ELIZA', developed at MIT in 1966 by Joseph Weizenbaum; with subsequent agents created to interact in commerce, education, and entertainment, including Second Life. [44]

76

K. Reed and G. Meiselwitz

3. Using a simple web interface, the student will read the selected material and answer the chatbot's questions. We will explore the feasibility of using a Baysian network to analyze the probability that the student's responses are correct. 4. The student's 'correct' responses will be used to formulate additional AIML statements. Answers marked as 'incorrect' will be flagged for review by a human expert. The student will be aware that he or she is 'training' the chatbot to understand the article being read, and is thereby contributing to a community of readers, in support of a constructivist or social/situational learning process. Obviously, the chatbot cannot truly ascertain whether or not the student actually comprehends the material read, but we will look for emerging patterns in the student responses indicative of student learning – one of which will be the 'success' of subsequent readers in responding to the chatbot's questions. Scoring will then be dispatched to the agent that will maintain student profiling (and a human instructor for any arbitration), so that future reading selections are more finely tuned to the student's reading level and interest. We will also study the chatbot dialog to determine whether this type of unstructured communication might be useful in creating more complex ontologies. The ITS we are proposing is not intended to replace human to human interaction and instruction. Learning is a highly social activity, especially in one as fundamental as reading comprehension. We wish to examine the interesting ramifications such an ITS might have on human-computer teaching and learning support, but ultimately, we hope to provide a tool that adults can use to improve their reading.

References [1] Foundation, A.I. (n.d.). AIML – The Artificial Intelligence Markup Language, http://Alicebot.org , http://www.alicebot.org/aiml.html (retrieved December 1, 2010) [2] B., O.: Educated Citizens in a Changing World. Vital Speeches of the Day 78(8), 364– 366 (2010) [3] Bellifemine, F., Caire, G., Greenwood, D.: Developing multi-agent systems with JADE. John Wiley & Sons Ltd., West Sussex (2007) [4] Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American Magazine (2001) [5] Brandon, C.: Nearly 30% of college freshmen can’t read, write, or calculate (2008), http://www.thefiveyearparty.com:http://www.thefiveyearparty. com (retrieved December 23, 2010) [6] Capuano, N., Carrolaggi, P., Crestani, J., Gaeta, M., Herber, E., Sangineto, E., et al.: A Virtual Organization for e-Learning. In: 1st International Kaleidoscope Learning GRID on Distributed e-Learning Workshop [7] Cartelli, A.: Semantics, Ontologies and Information Systems in Education: Concerns and Proposals. Issues in Informing Science and Information Technology 3 (2006) [8] Conway, J.: Educational Technology’s Effect on Models of Instruction. Educational Technology (1997), http://udel.edu/~jconway/EDST666.htm (retrieved November 26, 2010) [9] Corbett, A.T., Koedinger, K.R., Anderson, J.R.: Intelligent Tutoring Systems. In: Helander, M., Landauer, K., Prabhu, P. (eds.) Handbook of Human-Computer Interaction, 2nd edn., pp. 849–874. Elsevier Science B.V, Amsterdam (1997) [10] Devedizic, V.: Education and the Semantic Web. International Journal of Artificial Intelligence in Education 14, 39–65 (2004)

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

77

[11] Freedman, R.: What is an Intelligent Tutoring System? In: Ali, S.S., McRoy, S. (eds.) Intelligence, vol. 11 (3), pp. 15–16 (2000) [12] Gasevic, D., Hatala, M.: Ontology mappings to improve learning resource search. British Journal of Educational Technology 37(3), 375–389 (2006) [13] Hendler, J.: Agents and the Semantic Web. IEEE Intelligent Systems, 30-37 (2001) [14] Hermann, J.F.: The Silicon Tutor: A critical examination of the uses of intelligent tutoring system in writing assessment and pedagogy. Doctoral Dissertation, Indiana University of Pennsylvania (2005) [15] Huang, W., Webster, D., Wood, D., Ishaya, T.: An intelligent semantic e-learning framework using context-aware Semantic Web technologies. British Journal of Educational Technology 37(3), 351–373 (2006) [16] Interactive Educational Systems Design, Inc. (n.d.). How Jamestown Reading Navigator Supports Research-Based Instruction for Struggling Adolescent Readers Formative and Summative Assessment, http://www.readingnavigator.com: http://www. readingnavigator.com/mkt/assets/formative_and_summative_asse ssment.pdf (retrieved December 21, 2010) [17] Jia, J.: An AI Framework to Teach English as a Foreign Language: CSIEC. Association for the Advancement of Artificial Intelligence, pp. 59–71 (2009) [18] Laskey, K. J., Laskey, K. B., Costa, P. C., Kokar, M. M., Martin, T., Lukasiewicz, T. (eds.). Uncertainty Reasoning for the World Wide Web (2008), http://www.w3.org:http://www.w3.org/2005/Incubator/urw3/ XGR-urw3-20080331/ (retrieved November 19, 2010) [19] Maddux, C.D., Johnson, D.L.: Information Technology in Education: Some Reasons for Optimism. Computers in the Schools 26, 83–88 (2009) [20] McGuinness, D. L., & van Harmelen, F. (n.d.). OWL Web Ontology Language Overview. OWL Web Ontology Language Overview, http://www.w3.org/TR/owl-features/ (retrieved November 19, 2010) [21] Memon, Q., Khoja, S.: Academic Program Administration via Semantic Web. In: Proceedings of World Academy of Science, Engineering, and Technology, vol. 37, pp. 695–698 (2009) [22] Merriam, S., Caffarella, R.: Learning in Adulthood: A Comprehensive Guide. JosseyBass Incl., San Francisco (1998) [23] Murray, T.: Authoring intelligent tutoring systems: An analysis of the state of the art. International Journal of Artificial Intelligence in Education, 98–129 (1999) [24] Office of Institutional Research, Student placement rates across different developmental areas: 2004-2007. Northern Virginia Community College, Office of Institutional Research (2010) [25] Ohlsson, S.: Constraint-Based Student Modeling. In: Greer, J., McCalla, G. (eds.) Student Modeling: The Key to Individualized Knowledge Based Instruction, pp. 167–189 (1994) [26] Pahl, C., Holohan, E.: Applications of Semantic Web Technology to Support Learning Content Development. In: Koohang, A. (ed.) Interdisciplinary Journal of E-Learning and Learning Objects, vol. 5 (2009) [27] Pennington, M.: Why Johnny still can’t read (2008), http://www.edarticle.com:http://www.edarticle.com/k-12subject-areas/reading/why-johnny-still-cant-read.html (retrieved December 23, 2010) [28] Rajput, Q.N., Haider, S., Touheed, N.: Information Extraction from Unstructured and Ungrammatical Data Sources for Semantic Annotation. In: Proceedings of World Academy of Science, Engineering and Technology, vol. 40, pp. 295–303 (2009) [29] Ramachandran, S., Atkinson, R.: An Evaluation of Intelligent Reading Tutors. In: Woolf, B.P., Aïmeur, E., Nkambou, R., Lajoie, S. (eds.) ITS 2008. LNCS, vol. 5091, pp. 731– 733. Springer, Heidelberg (2008)

78

K. Reed and G. Meiselwitz

[30] Ramachandran, S., Stottler, R.H.: An Intelligent Tutoring System for Adult Literacy Enhancement. In: Proceedings of the fifth Internation Conference on Intelligent Tutoring Systems. Montreal (2000) [31] Reed, C.B., Conklin, K.D.: Enrolling in College, Ready or Not. The Chronicle of Higher Education 52(8) (October 2005) [32] Russell, S., Norvig, P.: Artificial Intelligence: A Modern Approach, 2nd edn. Prentice Hall, Upper Saddle River (2003) [33] Shabajee, P., McBride, B., Steer, D., Reynolds, D.: A prototype Semantic Seb-based digital content exchange for schools in Singapore. British Journal of Educational Technology 37(3), 461–477 (2006) [34] Shafrir, U., Etkind, M.: e-Learning for depth in the Semantic Web. British Journal of Educational Technology 37(3), 425–444 (2006) [35] Smith, M.K.: Learning theory, the encyclopedia of informal education (2009), http://www.infed.org:http://www.infed.org/biblio/b-learn.htm (retrieved December 22, 2010) [36] Snae, C., Brueckner, M.: Ontology-Driven E-Learning System Based on Roles and Activities for Thai Learning Environment. Interdisciplinary Journal of Knowledge and Learning Objects 3 (2007) [37] Stamper, J.C.: Automatic Generation of Intelligent Tutoring Capabilities via Educational Data Mining. Charlotte, NC (2010) [38] Stotsky, S., Wurman, Z.: College Students Who Can’t Do Math Or Read Well. Minding The Campus Reforming Our Universities (2009), http://www.mindingthecampus.com/originals/2009/12/by_sandra_ stotsky_and_zeev.html (retrieved December 22, 2010) [39] Tankeleviciene, L., Damasevicius, R.: Characteristics of Domain Ontologies for Web Based Learning and their Application for Quality Evaluation. Informatics in Education 8(1), 131–152 (2009) [40] Tomsett, C.: Reconfigurability: creating new courses from existing learning objects will always be difficult. Journal of Computer Assisted Learning 21, 440–448 (2005) [41] Urban-Lurain, M.: Intelligent Tutoring Systems: An Historic Review in the Context of the Development of Artificial Intelligence and Educational Psychology (1996), http://www.cse.msu.edu/rgroups/cse101/ITS/its.htm (retrieved December 21, 2010) [42] Census Bureau, U.S. (n.d.). College Enrollment Up 17 Percent Since (September 17, 2000). M2 Presswire [43] USA Today. One-third of students need remedial college math, reading (2010), http://USAToday.com:http://www.usatoday.com/news/education/ 2010-05-11-remedial-college_N.htm (retrieved December 23, 2010) [44] Wallace, R.: Alicebot, Alice Blog (2009), http://alicebot.blogspot.com/ (retrieved December 23, 2010) [45] Wenger, E.: Artificial Intelligence and Tutoring Systems: Computational and Cognitive Aproaches to the Communication of Knowledge. Morgan Kaufmann, Los Altos (1987) [46] Woolf, B.P., Beck, J., Eliot, C., Stern, M.: Growth and maturity of intelligent tutoring systems: A status report. In: Forbus, K., Feltovich, P.J. (eds.) Smart machines in education: The coming revolution in educational technology, pp. 99–143. AAAI Press, Menlo Park (2001) [47] Woolf, B., Eliot, C.: Customizing the Instructional Grid. Applied Artificial Intelligence 19, 825–844 (2005)

Interpreting User-Generated Content: What Makes a Blog Believeable? Rahayu Ahmad and Wayne G. Lutters Department of Information Systems, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA {arahayu1,lutters}@umbc.edu

Abstract. Blogs have democratized participation for everyone who wants to publish on the Internet. This trend, however, is problematic for readers trying to determine trustworthy and believeable sources. This blog diary study, allowed us to understand what aspects influenced believability and ultimately utililty through capturing routine blog-surfing behaviors. We found that literary appeal constituted by authentic and personalized stories positively influence the believability of blogs. Blogs that stimulated readers’ desire for novel and rare information also enhance their believability. Other factors that significantly correlated with believability included aesthetics, matching information needs and user expectations.

1 Introduction Web logs (blogs) have become an essential cord in the social fabric of the Net. A blog is usually defined as a frequently updated webpage and generally have customized navigation forms with dated entries organized as an archive in reverse chronological order [18]. Blogs can come in many different forms: filters, personal journals, and notebooks [13]. Filter blogs are mainly comprised of selective links by the authors, which often become pointers to other information resources. Personal journal blogs are more diary-like, containing personal thoughts and records of daily events. Lastly, notebook style blogs are usually in the form of long and focused essays. While blogs’ form and function have shifted over time, these solo-authored, widely read, personal information sources remain the bedrock of our user-generated infoculture [16]. Blogs have democratized participation for everyone who wants to publish on the Internet. Their popularity is largely dpendent on commercial blog software, such as Blogger [4]. Through such user-friendly software, any individual can feasibly create and publish contents on a potentially international scale with little technical knowledge. Given the personal voice, opinionated writing style, and opportunity to author anonymously, how do readers evaluate the believability of individual blogs?

2 Background Most research on the concepts of credibility or believability in social media has been conducted on web pages, a more general genre of online authorship than blogs. It has A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 81–89, 2011. © Springer-Verlag Berlin Heidelberg 2011

82

R. Ahmad and W.G. Lutters

focused primarily on trustworthiness and expertise. Trust is described as the willingness of the author to make valid assertion while expertise is the ability (competency) of the author in making those assertions [12]. In perceptions of website credibility, researchers have emphasized three factors: navigation, usability, and author background/affiliation. Their studies have demonstrated that websites with a professional look and feel and ease of use positively influence the perception of credibility [12]. The site author’s background and identifiable personal information also has positive impact [10]. Scholars have argued that mechanisms for establihsing a perception of credibility among readers may be meaningfully different from general websites and traditional media [15][19]. It has been speculated that trust with readers are facilitated by developing a personal relationship through disclosing the intimate personal and professional lives of bloggers, allowing readers to see the world through their eyes [22]. These authentic experiences, shared in the form of narratives, are a quintessential trait of blogs [2][20]. Compared with credibility in web sites, author identifiable information has no significant influence towards blog believability or credibility perception. One study demonstrated that there were no differences in perceived credibility when the blogger was identifiable versus when they were anonymous [7]. This implied that the identity of the writer was not being used to evaluate the credibility of the blog content. Another study suggested that bloggers build their credibility by showcasing their intellectual abilities through their analytical arguments and insightful commentaries of issues or events [2]. In addition, among experienced Internet users, blogs are perceived as more credible compared to traditional media such as newspapers or TV. [14]. Futhermore, unlike other media, bias in blogs was considered acceptable to readers [14]. These current studies suggest that blog readers may use distinctive factors in evaluating the believability of blogs, which have not yet been explored in greater detail.

3 Study Design Informed by the limitations of prior studies, we designed a contextual diary study that best preserved natural user behavior. An email invitation was sent out to undergraduate and graduate students and their friends who regularly maintain or read blogs. Twenty people participated in this study. Of the participants, 20% were male and 80% were female. The range of participant ages was 24 to 45 years, with a mean age of 28.5 years and a mean of 16 years of education. Our participants’ demographics were representative of the population of bloggers as indicated by Pew Internet and American Life Project [16]; in which half of all bloggers are under the age of 30. Based on answers to the self report question about their blog reading habits 20% reported doing so on daily basis, 50% were weekly, and 30% monthly. Participants were randomly presented with a search scenario to identify blogs that they would want to recommend to a friend within three broad interest areas: health, parenting, and travel. The rationale of having diverse types of blogs was to see whether there are any significant differences of factors influencing believability. The three areas were selected to represent various degrees of consequences and opinionated contents.

Interpreting User-Generated Content: What Makes a Blog Believeable?

83

For example health blogs would likely have higher consequences of believability and a lower degree of opinionated contents compared to parental and travelling. Based on the scenarios, participants were asked to search for blog at home on their own time, bookmarking each one that they found to be useful, with a minimum request of five. For each discovered blog, they were asked to complete an online diary consisting of a brief online survey with both open and 5-point Likert-scale questions. The open ended questions asked participant to briefly summarize the blog to ensure that they were actually reading it. The particants were also asked to describe their reasons for selecting the blog. As for the 5-point Likert scale questions (1=strongly disagree and 5=strongly agree) participants rated the blog on dimensions developed from Flanagin and Metzger [11] like believability, bias, persuasiveness and interest. The participants also rated each blog using 10-point scale (1=very low <property> and 10= very high <property>) in property terms proposed by Rubin and Liddy [19]: design, accuracy, timeliness of information, transparency of author information, literary appeal, profit purpose, appeal, and believability. Participants were given a maximum duration of two weeks for completing these online diaries. After analyzing the participants’ compilation of selected blogs and their evaluation for each one, we invited 5 people for a follow-up interview session to better understand their rationale for selecting those blogs and to provide further explanatory detail for our data analysis. These participants were selected based on the extremes of their property ratings, focused on those with large variance in their factors that might influence believability. The interview session was contextualized in the sense that both participants and the interviewer revisited the blogs of interest during the interview session. This interview was conducted through the Skype application which permits screen sharing between the two parties.

4 Analysis In total 77 blogs were analyzed. We performed correlational analyses to identify factors that correlated with believability. We also coded the participants’ responses to the open ended question "why do you think the blog is useful?” to help interpret the statistical results. The relationship between all variables and the blog’s believability was investigated using the Spearman correlation coefficient since the measures were ordinal in nature. There was a moderate correlation between the two variables literary appeal and believability (rs = .37, p < .05; n=77). The variable curiosity trigger also had moderate correlation with believability (rs = .32, p < .05; n=77). The other three variables that were weakly correlated with believability were aesthetic, (rs = .26, p < .05), matching needs (rs = .22, p < .05) and matching expectation (rs = .25, p < .05). The following section provides a discussion of our results.

5 Result and Discussion 5.1 Literary Appeal and Believability Similar to the web credibility literature [10], we observed that participants valued reliable citations and references in judging the believability of the blogs:

84

R. Ahmad and W.G. Lutters

I think this blog is very useful for parents because it provides a scientifically proven result. (Tracy)1 There are a lot of links and information about diabetes. The information is a concrete, supported by links, videos, academic research and other reliable resources. (Melissa) However, the importance of these factors was most dominant in the medical scenario given the serious consequenses of applying fraudulent knowledge. For other scenarios where citations are absence, participants relied on self disclosure of the bloggers and authenticity of their personalized stories in making a believability judgement. Self-disclosure has been shown to increase trust in workgroups [5][24]. For bloggers, the self disclosure tended to be holistic covering not only aspects of her/his work, but her/his self and personal life [22]. Through stories, photos, and personal links a blog’s personality can reveal much about the author’s values, orientation, and biases. In our study, participants valued the personalized diaries which enforced the believability of the blog. One participant explained: It gives a daily account of a person living with diabetes, which gives a balanced and realistic story to believe in. (Kate) The participants also emphasized the authenticity of the blogs’ content as one of the criteria they used in recommending it to others. Participants perceived real people’s experiences as authentic and trustworthy, as one noted: It [the blog] provides a perspective on Bali done by a foreigner who is actually living there instead of simply visiting as a tourist. (Sam) The authentic and genuine experience allowed readers to relate to the author and immerse themselves in the situation, thus enabling greater believability of the content: “When I want to go blogs, I like them to be like talking to themselves, don’t act like they are at work, or a teacher or university professor. I just want to have their pure feeling because it helps me to relate and be more attune to the situation.” (Cheryl) This finding potentially explains why the author background has not been found to be critical for determining credibility. The personalized voice and first-hand experience of the bloggers are cues for determining their competency and hence the believability of their blogs. This observation is also supported by a recent finding that authors’ authenticity and passions are a reliable measure for determining blog’s content credibility [15].

1

Our stylistic convention is to identify quotations from the interviews in italics and direct excerpts from the diaries in Courier font.

Interpreting User-Generated Content: What Makes a Blog Believeable?

85

5.2 Curiosity Trigger and Believability In the studies of online trust, curiosity has been perceived as critical component in a two staged trust model [17]. Based on this model, trust is formed gradually, initially started with exploratory stage and continue to commitment stage [17]. In exploratory stage, the user did not engaged yet in a specific web site and trying to decide whether or not to explore the web site. Egger [9] argued that users initial reaction to a web site is influence by the its appeal and its ability to arouse the users’ curiosity for further exploration. In this study, blogs that have meaningful headings/title and offers variety of information aroused their desire to explore the blog in details as described by our participants: “I could see that thediabetesblog.com would be more informational. Thus I would explore this blog more.” (Melissa) “It combines food and travel, so i'm curious about the topic, on how food varies around the world.” (Adam) Most of our participants were motivated by the desire to seek out novel information, also known as diversive curiosity [3]. Thus, authors who satisfied the readers’ desire for novel and intriguing information, enhanced the believability of their own blogs.One participant noted: “I was curious to know the different between the types of diabetes stages,and from reading her blogs and links provided from this blog, I get to know them.” (Tania) As described by other researchers [1][14], blogs are perceived to be a credible medium due to the perception of their ability to provide insightful, independent information. Our study found that participants were seeking blogs that provided exquisite and interesting information that was hard to find elsewhere, which eventually lead to higher believability ratings. The following diary notes explain: This blog shows the more "hip" side of Bali life. As opposed to other blogs about Bali/Indonesia which tend to be on the denser, more intellectual side of the blogosphere spectrum, this blog provides a lighthearted view of the nightlife, style and beach culture of Bali. (Sam) If someone wants to travel around the world and eager to know the hidden treasures of the destinations, this [the blog] is the right place to find. (Adam) In a prior study, the novelty of a blog’s content significantly influenced the playfulness perception of the travel’s blog which eventually led to a positive intention to visit the places[6]. This implied that the novelty of the blog content is used as a reliable indicator to actually make decisons. Our study provides further explanation of this novelty impact for the believability perception of blogs.

86

R. Ahmad and W.G. Lutters

5.3 Aesthetics and Believability Aesthetics, described as the overall look and design of a site, has been established as one of the factors that influence the credibility of websites. In Fogg’s [12] credibility study involving 2500 participants, the general look and feel was the top most factor. Professional looking web sites were often rated as more credible. For this study, generally particpants described blogs with high aesthetic qualities as: “A good blog should have a neat design, not too colorful, and has a appealing color scheme. It also has to have contemporary design” (Adam) Look and design impacts the judgement of the competency of the blogger, and, hence, influence the beleivability of the contents, as one participant highlights: “I think that an author who would spare some time on finding a good layout is actually dedicative about his/her blog.” (Melissa) There was also a strong reliance on images for verifying the blogs’ contents and the blogger’s experience particularly in travel blogs as illustrated by the following excerpt: This blog is based on his real experience which inclusive tips and photo as a proof. (Lisa) Our findings suggest that similar to web credibility perception, the look and design of blogs is being actively considered in evaluating believability. However, it has a somewhat weaker correlation with believability when compared to literary appeal or writing style. This finding might be attributed to the nature of blogs which are diary like, with a higher degree of personal disclosure, resulting in greater attention to its writing style and contents. The look and design, however, supplemented other cues in determining the competencies of the bloggers. 5.4 Matching Information Needs, Prior Expectation and Believability Increasing numbers of people are relying on blogs as primary sources of information. It is reasonable to argue that the ability of blogs to provide sufficient and appropriate information to the readers enhances their believability perceptions. Our diary responses demonstrated that users acquired the information they were looking for which eventually influence their believability perceptions: The links provided on the site are suitable for our needs of information. (Lyn) It [the blog] helps me to plan out when should I go and expect upon vacation to Langkawi by next month. (Lisa) It has been demostrated elsewhere that readers are often attracted to blogs for their depth of information [23]. The comprehensiveness of information may signal the expertise of the blogger to either compose or compile useful information, hence

Interpreting User-Generated Content: What Makes a Blog Believeable?

87

positively influencing the believability of the blog. The excerpts below demonstrated our participants’ views: It consists of everything that a diabetic needs to know. (Abby) The current posts and seem to mostly be relevant to child development and the responses seem to be appropriate for the topic of the original post. (Gilmore) In the public relations literature, an important property of trust is the trustee’s expectation about the motives and behaviors of a trustee [8]. If those expectations are met, trust is seen as having been warranted. If not, either the level of trust will be lowered or distrust may guide future actions. Similarly, in the blog literature, the audience of a blog has expectations for the authenticity of the blog [20]. In Baumer et al.’s [2] study, readers recognized the personal opinion and narratives in a blog as part of the blog’s authenticity. We shared a similar observation in our study where participants expected the blogs to be based on firsthand experience as precisely described by Tania: “My expectations are personal experience from the author and also other useful links related to the topic” (Tania) On top of that, readers also had some expectation for the frequency of updates, visual style, navigability, responsiveness, appropriateness, and other aspects [2]. From the previous discussions we have observed that aesthethics (visual style) and matching information needs (appropriateness) positively correlate with believability. Having most of the readers’ expectation met, led to perception of believability for a blog.

6 Conclusion This study has a few key limitations. Firstly, it employs a correlational analysis method which prevents us from making cause and effect conclusions. Future work can establish a reliable model of believability perceptions by incorporating the aspects identified in this study. Secondly, although we tried to cover a wide range of blogs topics, our three categories are just the beginning of a comprehensive study. Blogs that are dedicated to crisis management (in event of natural disaster), for example, may give greater importance to the “timeliness” factor for asessing the believability. The value of this particular study is twofold. Firstly, previous studies have mainly focused on credibility perceptions of blogs with other traditional media. These studies have highlighted that blogs are perceived as more credible particularly among experience Internet users. Previous studies also demonstrated that traditional measures like author identifiability did not influence the credibility perception in blogs. These suggested that other factors may be more important in determining the believability and usefulness of the blogs. Our multi-method study is a step toward addressing these gaps. In our field analysis, some key factors like literary appeal and aesthetics emerged as central traits, confirming past literature. However, literary appeal has taken a different meaning where the emphasis is on the personalized and

88

R. Ahmad and W.G. Lutters

authentic experience of the authors which positively influence the believability and utility of their blogs. We also found an emerging factor, curiosity trigger, described as the ability to stimulate readers desire for novel information, to influence the believability of blogs. Readers were attracted to blogs that provided exquisite and unique information which could not be found anywhere else. This finding supports other studies who found that people considered blogs as source of rich and deep alternative information source, which eventually led to positive credible perceptions [1]. This study also provided explanatory details that readers have certain expectations of look and design, writing style and appropriateness of information presented. Secondly, as our information sources are increasingly dominated by user generated content, the question of credibility or believability has become more relevant. Although blogs offer myriad opportunities for learning and gaining new information, improper consideration of the contents’ believability and reliability may lead to harmful consequences especially in riskier scenarios such as with medical information. This research is useful in highlighting the measures used by readers in believing the blogs and hence can be used for assessment of such potential harmful consequences and the formation of appropriate Web policies and education both formal and informal.

References 1. Andrews, P.: Is Blogging Journalism? Nieman Reports 57(3), 63–64 (2003) 2. Baumer, E., Sueyoshi, M., Tomlinson, B.: Exploring The Role Of The Reader In The Activity Of Blogging. In: Proceedings of ACM Conference on Human Factors in Computing Systems (CHI), pp. 1111–1120 (2008) 3. Berlyne, D.E.: Conflict, Arousal, And Curiosity. McGraw-Hill, New York (1960) 4. Blood, R.: How Blogging Software Reshapes The Online Community. Communications of the ACM 47(12), 53–55 (2004) 5. Bos, N., Olson, J., Gergle, D., Olson, G., Wright, Z.: Effects Of Four Computer-Mediated Communications Channels On Trust Development. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 135–149. ACM Press, New York (2002) 6. Chen, Y.C., Shang, R., Li, M.J.: The Effects of Traveler’s Blogs on the Behavioral Intention to Visit a Tourism Destination. In: International Conference on Business and Information, Malaysia (2009) 7. Chesney, T., Seong, D.S.K.: The Impact Of Anonymity On Weblog Credibility. Int. Journal of Human-Computer Studies 68, 710–718 (2010) 8. Doney, P.M., Cannon, J.P.: An Examination Of The Nature Of Trust In Buyer-Seller Relationships. Journal of Marketing 61, 35–51 (1997) 9. Egger, F.N.: Affective Design Of E-Commerce User Interfaces: How To Maximize Perceived Trustworthiness. In: Proceedings of the International Conference on Affective Human Factors Design. Academic Press, London (2001) 10. Eysenbach, G., Kohler, C.: How Do Consumers Search For And Appraise Health Information On The World Wide Web? Qualitative Study Using Focus Groups, Usability Tests, And In-Depth Interviews. BMJ 324, 573–577 (2002)

Interpreting User-Generated Content: What Makes a Blog Believeable?

89

11. Flanagin, A.J., Metzger, M.J.: Perceptions Of Internet Information Credibility. Journalism and Mass Communication Quarterly 77, 515–540 (2000) 12. Fogg, B.J., Marshall, J., Laraki, O., Osipovich, A., Varma, C., Fang, N., Paul, J., Rangnekar, A., Shon, J., Swani, P., Teinen, M.: What Makes Web Sites Credible? A Report On A Large Quantitative Study. In: CHI Conference Proceedings (2001) 13. Herring, S., Scheidt, L., Bonus, S., Wright, E.: Bridging The Gap: A Genre Analysis Of Weblogs. In: Proceedings 37th Annual HICSS Conference (2004) 14. Johnson, T.J., Barbara, K.K.: In Blog We Trust?: Deciphering Credibility of Components of the Internet among Politically Interested Internet Users. Computers in Human Behavior 25(1), 175–182 (2009) 15. Kang, M.: Measuring Social Media Credibility: A Study on a Measure of Blog Credibility. Institute for Public Relations (2010) 16. Lenhart, A., Fox, S.: Bloggers: A Portrait Of The Internet’s New Storytellers. Pew Internet & American Life (2006) 17. McKnight, D.H., Choudhury, V., et al.: Trust In E-Commerce Vendors: A Two-Stage Model. In: International Conference on Information Systems (ICIS). Association for Information Systems, Brisbane (2000) 18. Nardi, B.A., Schiano, D.J., Gumbrecht, M., Swartz., L.: Why We Blog. Communication ACM 47(12), 41–46 (2004) 19. Rubin, V., Liddy, E.: Assessing Credibility Of Weblogs. In: Proceedings of the AAAI Spring Symposium: Computational Approaches to Analyzing Weblogs, CAAW (2006) 20. Schiano, D.J., Nardi, B.A., Gumbrecht, M., Swartz, L.: Blogging By The Rest Of Us. In: Proceedings of CHI, Vienna, Austria (2004) 21. Trammell, K.D., Keshelashvili, A.: Examining The New Influencers: A Self-Presentation Study Of A-List Blogs. Journalism & Mass Communication Quarterly 82(4), 968–982 (2005) 22. Van House, N.: Weblogs: Credibility and Collaboration in an Online World (prepared for CSCW Workshop on Trust, unpublished) (October 2004) 23. Wall, M.: Blogging Gulf War II. Journalism Studies 7(1), 111–126 (2006) 24. Zheng, J., Veinott, E., Bos, N., Olson, J., Olson, G.: Trust Without Touch: Jumpstarting Long-Distance Trust With Initial Social Activities. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 141–146. ACM Press, New York (2002)

Extracting Events from Wikipedia as RDF Triples Linked to Widespread Semantic Web Datasets Carlo Aliprandi1, Francesco Ronzano2, Andrea Marchetti2, Maurizio Tesconi2, and Salvatore Minutoli2 1 Synthema Srl Via Malasoma 24 56121 Ospedaletto (Pisa) - Italy [email protected] 2 Institute of Informatics and Telematics (IIT) CNR Via G. Moruzzi, 1 56123 Pisa - Italy {francesco.ronzano,andrea.marchetti,maurizio.tesconi, salvatore.minutoli}@iit.cnr.it

Abstract. Many attempts have been made to extract structured data from Web resources, exposing them as RDF triples and interlinking them with other RDF datasets: in this way it is possible to create clouds of highly integrated Semantic Web data collections. In this paper we describe an approach to enhance the extraction of semantic contents from unstructured textual documents, in particular considering Wikipedia articles and focusing on event mining. Starting from the deep parsing of a set of English Wikipedia articles, we produce a semantic annotation compliant with the Knowledge Annotation Format (KAF). We extract events from the KAF semantic annotation and then we structure each event as a set of RDF triples linked to both DBpedia and WordNet. We point out examples of automatically mined events, providing some general evaluation of how our approach may discover new events and link them to existing contents. Keywords: Knowledge Representation, Knowledge Extraction, Semantic Web, Natural Language Processing, Semantics.

1 Introduction The core aim of the Semantic Web is to provide a set of methodologies, standards, technologies and best practices to make explicit the semantics that lies behind the data exposed over the Web. As a consequence, it is possible to support an easy and serendipitous automatic integration of the great variety of Web contents, thanks also to the exploitation of shared knowledge references like ontologies, lexicons and semantic resources. In this scenario, the Resource Description Framework (RDF) [1] and the Ontology Web Language (OWL) [2] currently constitute the two core W3C standards useful to respectively represent knowledge over the Web and to specify a formalized semantic reference frame. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 90–99, 2011. © Springer-Verlag Berlin Heidelberg 2011

Extracting Events from Wikipedia as RDF Triples Linked

91

Many methodologies to extract knowledge from Web contents have been proposed. They can be divided into two wide groups: - relation or fact extraction systems: they usually apply specific Web mining procedures to extract facts from online contents. Facts are usually represented as attribute-value pairs describing some feature of a given entity of interest, i.e. the population of a specific country. Extracted facts are usually shown to users as search results referring to the part of Web documents. Examples are WebKnox [3], the Grazer System [4] and TextRunner [5]. - interlinking systems: they mine information from poorly structured Web contents and, unlike the previous group of systems, they expose the extracted knowledge over the Web as RDF data. In this context, the Linked Data initiative [6] has defined a set of best practices to represent such knowledge by exploiting RDF and to unambiguously identify entities on a Web scale by means of URIs. A central role is played by DBpedia [7], an extract from Wikipedia contents, representing a hub for many other Linked Data datasets. Many other proposals to produce Semantic Web interlinked datasets have been made in parallel to Linked Data. They usually deal with some particular kind of Web contents and propose specific methodologies, like: - systems to enrich Web content: they apply procedures for keyword extraction or Named Entity recognition over Web pages, automatically producing a set of relevant terms to be annotated through the URI of the referred Wikipedia/DBpedia entity. Open Calais [8], for example, parses documents and points out entities, facts and events. When possible, entities are linked to DBpedia, Freebase or GeoNames URIs. Wikify [9] performs keyword extraction from Web pages, and disambiguates mined terms linking them to the referred Wikipedia entity. - systems to enrich social tagging service: systems like Faviki [10] and LODr [11] allow users to associate a tag to the Wikipedia page describing the referred concept. A different approach to automatically connect user tags to a specific concept of Wikipedia has been adopted by [12], exploiting also Tagpedia [13], a semantic resource for Tag Sense Disambiguation. Considering the attempts to build semantic resources by mining Wikipedia, [14] describes a set of methodologies adopted to extract from Wikipedia an association thesaurus, by exploiting the set of internal links, the taxonomy of Wikipedia categories and by mining the contents of each article. [15] describes how to build Tagpedia, a semantic reference useful to support the disambiguation of tags, by mining the structure of Wikipedia articles. Proposals to extract text snippets from Wikipedia representing facts have been also defined. Specific extraction techniques have been tuned to gather relevant text snippets from other articles [16] or to mine Wikipedia Named Entities over time [17]. Mining of Wikipedia has also been carried out by applying Natural Language Processing: a dump of the English Wikipedia has been shallow parsed and semantically annotated [18]. Applying both shallow and deep parsing to Wikipedia, methodologies to build a common sense knowledge base have been proposed [19].

92

C. Aliprandi et al.

In this paper we propose an approach to mine the unstructured textual data of Wikipedia so as to extract events, representing them through sets of RDF triples and integrating them in the Linked Data dataset. In section 2, after having clarified the notion of event, we present the KYOTO Annotation Format (KAF) and we describe Synthema English Slot Grammar, a deep parser used to automatically produce KAF annotations. In section 3 we describe the procedure to identify events and to represent them through RDF triples exploiting also WordNet and DBpedia. In section 4 we provide some meaningful examples of this procedure as well as some initial evaluation. We conclude in section 5, discussing our future plans to improve the event extraction process from KAF annotated texts.

2 Annotating Textual Documents by Exploiting KAF The goal of this paper is to explore the possibility to mine events from linguistically and semantically annotated textual documents and to try to propose possible ways to represent them as sets of RDF triples tightly connected with core Semantic Web datasets like DBpedia. For this reason we need first of all to define what we mean by an event, but also to specify a possible RDF representation of events as useful Semantic Web knowledge. We assert that an event is something that happens having some relevance in providing information in a particular context: it could be characterized by specific spatial and temporal coordinates. As a consequence, an event is usually built around a specific action or happening. In order to mine events from documents and, in particular, from the contents of Wikipedia articles, we consider the results of their linguistic and semantic analysis encoded in KAF [20], the deep semantic annotation format that we developed in the context of the KYOTO Project. KAF is a language neutral annotation format representing both morpho-syntactic and semantic annotation of documents through a layered structure. Starting from the lower of all its annotation layers, where tokens, sentences and paragraph are identified, in KAF each additional layer is built on top of the lower one, referring to its constituent elements. In this way, several levels of text annotation can be added by different linguistic processors. Moreover, specialized linguistic processors can be developed to generate incremental annotations for each specific layer. In KAF there are three macro-layers of document annotation (see also Figure 1): • morpho-syntactic layer: it groups all the language-specific text annotations. Tokens, sentences and paragraphs are identified in a specific document. Terms made of words or multi-words are pointed out, along with their Part Of Speech. In this layer also chunks and functional dependencies are represented. • level-1 semantic layer: it includes linear annotation of expressions of time, events, quantities and locations. • level-2 semantic layer: it is mainly devoted to represent facts, in a non linear annotation context, thus possibly aggregating evidences from the lower layers of multiple textual sources.

Extracting Events from Wikipedia as RDF Triples Linked

93

Fig. 1. The three macro-layers of KAF document annotation

Our RDF event extraction process mainly exploits the results of the morphosyntactic annotation layer of KAF. In this layer, the following elements are annotated: • word forms: each word form is unambiguously determined and linked to the sentence and paragraph of the text it belongs to. • terms: terms, also composed of two or more word forms, are identified and characterized by lemma, Part Of Speech and, when possible, by the type of referred Named Entity. The link of a term to an external reference can be represented. This is the KAF feature that we have mostly exploited to represent term meaning, either as a WordNet concept, disambiguated by proper WSD, or as a DBpedia entity. • dependency relations: they represent functional relations among terms, such as Agent, Action, Object, Qualifier, When, Where, How. • chunks: they are used to identify structured phrases, spanning one or more terms, like noun phrases, verbal phrases and prepositional phrases. Synthema English Slot Grammar (Syn ESG), an inverse parallel deep parser, is used to automatically produce KAF annotations. It carries out complex Natural Language Processing (NLP) tasks like word tokenization, segmentation, Part Of Speech tagging, dependency parsing, anaphora resolution and functional analysis. Syn ESG is intended to identify relevant knowledge from the raw text, by detecting concepts and semantic relations in texts. Concept extraction and text mining are applied through a pipeline of linguistic and semantic processors that share as a common ground McCord's theory of Slot Grammar [21]. Syn ESG parser - a bottomup chart parser - employs a parse evaluation scheme used for pruning away unlikely analyses during parsing as well as for ranking final analyses. By including semantic information directly in the dependency grammar structures, the system relies on the lexical semantic information combined with functional relations. Besides Named Entities, locations, time-points and dates, Syn ESG detects relevant information like

94

C. Aliprandi et al.

chunks, noun phrases and verbal phrases. The detected terms are then extracted, reduced to their Part Of Speech and functional tagged base form. Finally, syntactically and semantically tagged words are properly encoded in the corresponding KAF annotation, and the specific KAF layer is produced.

Fig. 2. Example output produced by the Syn ESG text parser

In Figure 2 we show the output sample for the sentence: “The Battle of Cannae took place near the town of Cannae in Apulia, during the Second Punic War”, taken from the Wikipedia article describing the Battle of Cannae. The figure shows functional dependencies among terms. Note the reference resolution for the word “Cannae”, that is correctly co-referred to a unique URI (represented by the id #2583). Starting from the described KAF features, in the following sections we detail the extraction of events from KAF annotated documents as well as their representation as RDF triples.

3 Events Extraction and RDF Representation As mentioned before, in a KAF annotated text, mined terms are linked, thanks to a specific Word Sense Disambiguation algorithm, to either WordNet or DBpedia. Summarizing, we assert that the meaning of a term in KAF documents can be denoted by a synsetID of WordNet and/or by a URI of an entity of DBpedia. Unlike DBpedia URIs, WordNet synsetIDs may represent ambiguous Web identifiers. In this paper we are proposing a tentative URI schema for WordNet synsets: the URI that identifies a synset of English WordNet version 3.0 is ‘http://www.kyoto-project.eu/wordnet/ English/30/synsetID’. In the next section we describe how to extract events from KAF documents and how to represent them as sets of RDF triples.

Extracting Events from Wikipedia as RDF Triples Linked

95

3.1 Identifying Events Inside KAF Annotated Documents To extract RDF representations of events from a KAF annotated document, we mainly consider the terms linked to WordNet synsets and to DBpedia entities, and the dependency relations linking terms. We base our event extraction process on the assumption that the nucleus of an event is an action expressed by a verb, i.e. by a term that has been classified as a verb and associated to a WordNet synset. We also assume that the features that characterize an event are included in and represented by the set of dependency relations of the sentence the verb belongs to and connected in some way to the same verb. In particular we consider the dependency relations of subject (the entity that performs the action), object (the entity affected by the action), when (time or interval of happening of the action) and where (place of happening of the action).

Fig. 3. Set of dependency relations characterizing an event

To identify an event we need a set of dependency relations, in particular we need at least a subject or an object dependency relation. The connection of terms with WordNet or DBpedia URIs is essential in order to represent events as Semantic Web RDF triples linked with other datasets. In Figure 3 the set of dependency relations that can be exploited to characterize an event is schematized. Each term is connected to WordNet or DBpedia except the time-interval characterizing the when dependency relation, which is a literal. Given the previous set of dependency relations, in the next section we describe how to represent events as a set of RDF triples. 3.2 Representing Events as RDF Triples Exploiting DBpedia URIs Each event that we identify in a document is characterized by a subset of the dependency relations shown in Figure 3. In addition, the verbal WordNet synset defining the action of the event identified by its synset URI and the terms of the subject, object and where dependency relations are identified by a URI of a WordNet synset or of a DBpedia entity. The when dependency relation is described by a string specifying a time or interval. We can express an event as a set of RDF triples by exploiting ontological properties taken from the DBpedia Ontology or from a properly structured ontology of dependency relations. In what follows we assume use of an ontology, referred

96

C. Aliprandi et al.

to as Dependency Relation Ontology, properly published at the stub ‘http:// dependencyrelat.org/ontology/’ namespace. The when and where dependency relations can be respectively expressed by the ‘http://dbpedia.org/ontology/Date’ and ‘http://dbpedia.org/ontology/Place’ OWL properties of the DBpedia Ontology. The subject and object dependency relations can be respectively represented in the Dependency Relation Ontology by the ‘http://dependencyrelat.org/ontology/Subject’ and ‘http://dependencyrelat.org/ontology/Object’ properties.

Fig. 4. RDF representation of an event

As a consequence, we obtain the RDF representation of an event shown in Figure 4. The two triples describing the subject and the object of an event are reified and thus referenceable by a URI: we assume to use for this purpose URIs published at the stub ‘http://linkedevents.org/events/’ namespace. The two URIs describing these RDF triples are in turn grouped under the same RDF Bag Container and thus both referenceable through a third URI that is also assumed to be published at the same stub namespace. This URI, referred to as the event URI, obtained by aggregating in the same RDF Bag Container the URIs defined reifying the subject and object triples, is represented in Figure 4 by the dashed line. It points out the core features of an event and represents the subject of the RDF triples describing the place and time of happening of that event (properties ‘http://dbpedia.org/ontology/Place’ and ‘http://dbpedia.org/ontology/Date’). To preserve in the RDF representation, also the URL of the Web page the event has been mined from, we also represent the event URI by means of the stub ‘http://dependencyrelat.org/eventProp/fromURL’ ontological property. To conclude, we can notice that the two RDF triples pointing out the subject and the object of an event could be collapsed into a single one if we manage to represent WordNet verbs, constituting the core actions characterizing an event, as ontological properties: we can exploit the hypernym/hyponym taxonomy of WordNet verbs to define a hierarchy of subsumed ontological properties related to them (i.e. in English WordNet 3.0 the verb consume subsumes the verb eat). Each of these properties

Extracting Events from Wikipedia as RDF Triples Linked

97

contributes to the definition of the main RDF triple of an event by linking the WordNet/DBpedia URI of the subject of the action to the WordNet/DBpedia URI that points out the object, in this case through the URI of the ontological properties representing the verb describing the actions.

4 Example of Mined Events To gather initial examples of facts, we annotated a set of Wikipedia articles, mainly related to wars and battles. Here we show a significant example of mined events from a sentence extracted from the Wikipedia page describing the ‘Battle of Gettysburg’:

Fig. 5. Concepts mined from the sample sentence: “The two armies began to collide at Gettysburg on July 1, 1863, as Lee urgently concentrated his forces there”

Figure 5 shows terms and dependency relations mined from the sentence. Two sentences are identified: “The two armies began to collide at Gettysburg on July 1, 1863” and “Lee urgently concentrated his forces there.”. Moreover, five terms are connected to WordNet synsets and two terms are identified through DBpedia entities. The date is a literal and thus managed as a string (July 1, 1863). From this sentence we can mine two events; their RDF representation is shown in Figure 6. Notice that two DBpedia URIs are exploited to refer to the two terms of the sentence: 1. Gettysburg: http:// dbpedia.org/resource /Gettysburg,_Pennsylvania 2. Robert E. Lee: http:// dbpedia.org/resource /Robert_E._Lee Moreover, the URI of four WordNet Synsets are exploited to represent the RDF triples of these events. In particular, the event related to the first sentence is described by a subject- The corresponding RDF triple is reified and further characterized by the place and date of happening of that event (Gettysburg, July 1st 1863). The event related to the second sentence is characterized by a subject and an object but no place and date are specified. Both events are linked to the URL of the Wikipedia page they have been mined from, through the RDF property: ‘http://dependencyrelat.org/eventProp/fromURL’.

98

C. Aliprandi et al.

Fig. 6. RDF representation of mined events from the sample sentence

5 Conclusions and Future Work In this paper we have presented an approach to extract events from documents, annotated using a deep text parser according to the KYOTO Annotation Format. We have defined an event extraction methodology that takes as input KAF annotated documents: it is based on functional dependencies and on the results of disambiguation of the terms through WordNet synsets or DBpedia entities. We have defined a representation of the mined events, as a set of RDF triples exploiting the URI of WordNet and DBpedia to point out the different entities taking part into events. We have provided a significant example of event mining procedure starting from a Wikipedia article. We are carrying out a global evaluation of the effectiveness of our event mining procedure by processing a larger set of Wikipedia articles. In this way we plan to provide also global quantitative evaluations of the quality of events. Even though the event extraction activities are at an early stage, we believe that the methodologies described can be useful to automatically produce Semantic Web data linked with other datasets. In particular, by applying this process to DBpedia, a considerable amount of new RDF triples can be generated. In conclusion, we can state

Extracting Events from Wikipedia as RDF Triples Linked

99

that our approach represents an example of synergy between NLP techniques and the Semantic Web to produce and link over the Web semantically-grounded contents. Acknowledgments. This work is partially funded by the European Commission (KYOTO project, ICT-2007-211423).

References 1. RDF W3C Web Page, http://www.w3.org/RDF/ 2. OWL W3C Recomm., http://www.w3.org/TR/owl-features/ 3. Urbansky, D., Thom, J.A.: WebKnox: Web Knowledge Extraction. In: 13th Australasian Document Computing Symposium, Hobart (2008) 4. Zhao, S., Betx, J.: Corroborate and Learn Facts from the Web. In: 13th International Conference on Knowledge Discovery and Data Mining, San Josè (2007) 5. Banko, M., Etzioni, O.: The Tradeoffs Between Open and Traditional Relation Extraction. In: 46th ACL: Human Language Technologies, Columbus (2008) 6. Linked Data Web Site, http://linkeddata.org/ 7. DBpedia Web Site, http://dbpedia.org/About 8. Open Calais Web Site, http://www.opencalais.com/ 9. Wikify! Web Site, http://www.wikifyer.com/ 10. Faviki Web Site, http://www.faviki.com/ 11. Passant, A.: LODr - A Linking Open Data Tagging System. In: Social Data on the Web Workshop at the 7th Int. Semantic Web Conference, Karlsrhue (2008) 12. Tesconi, M., Ronzano, F., Marchetti, A., Minutoli, S.: Semantify del.icio.us: automatically turn your tags into senses. In: Social Data on the Web Workshop at the 7th International Semantic Web Conference, Karlsrhue (2008) 13. Tagpedia Web Site, http://www.tagpedia.org/ 14. Nakayama, K.: Extracting Structured Knowledge for Semantic Web by Mining Wikipedia. In: Social Data on the Web Workshop at the 7th International Semantic Web Conference, Karlsrhue (2008) 15. Ronzano, F., Marchetti, A., Tesconi, M., Minutoli, S.: Tagpedia: a Semantic Reference to Describe and Search for Web Resources. In: Social Web and Knowledge Management Workshop at the 17th World Wide Web Conference, WWW 2008, Beijing (2008) 16. Adafre, S.F., Jijkoun, V., de Rijke, M.: Fact Discovery in Wikipedia. In: IEEE/WIC/ACM International Conference on Web Intelligence, Silicon Valley (2007) 17. Bhole, A., Fortuna, B., Grobelnik, M., Mladenic, D.: Mining Wikipedia and Relating Named Entities over Time. In: 13th International Conference on Knowledge Discovery and Data Mining, San Josè (2007) 18. Asterias, J., Zaragoza, H., Ciaramita, M., Attardi, G.: Semantically Annotated Snapshot of the English Wikipedia. In: 6th International Language Resources and Evaluation Conference LREC 2008, Marrakech (2008) 19. Suh, S., Halpin, H., Klein, E.: Extracting Common Sense Knowledge from Wikipedia. In: 6th International Semantic Web Conference, Athens, GA, USA (2006) 20. Bosma, W., Vossen, P., Soroa, A., Rigau, G., Tesconi, M., Marchetti, A., Aliprandi, C., Monachini, M.: KAF: a generic semantic annotation format. In: 5th International Conference on Generative Approaches to the Lexicon, Pisa (2009) 21. McCord, M.C.: Slot Grammar: A System for Simpler Construction of Practical Natural Language Grammars. Natural Language and Logic, 118–145 (1989)

Collaborative Sensemaking during Admin Permission Granting in Wikipedia Katie Derthick1, Patrick Tsao2, Travis Kriplean2, Alan Borning2, Mark Zachry1, and David W. McDonald3 1 Human Centered Design & Engineering, University of Washington Department of Computer Science & Engineering, University of Washington 3 The Information School, University of Washington Seattle, Washington, USA {derthick,pattsao,kriplean,borning,zachry,dwmc}@uw.edu 2

Abstract. A self-governed, open contributor system such as Wikipedia depends upon those who are invested in the system to participate as administrators. Processes for selecting which system contributors will be allowed to assume administrative roles in such communities have developed in the last few years as these systems mature. However, little is yet known about such processes, which are becoming increasingly important for the health and maintenance of contributor systems that are becoming increasingly important in the knowledge economy. This paper reports the results of an exploratory study of how members of the Wikipedia community engage in collaborative sensemaking when deciding which members to advance to admin status. Keywords: contributor systems, system administration, collaboration tools, sensemaking.

1 Introduction Decision making in an online community is often a difficult process. In large online communities like Wikipedia, administrators are key to keeping the community functioning, and those administrators come from the ranks of regular users. The decision to grant administrative permissions has important consequences for the entire community. In the process of deciding which candidates should be granted administrative permissions, reviewers must consider a user’s behavior relative to a set of criteria and come to some reasonably shared understanding of the merits of a new potential administrator. This study examines the collaborative activity of an online community deciding who is given administrative privileges. Understanding the processes of admin promotion in Wikipedia by its own members is challenging. Prior work has considered the edit histories of contributors to identify the characteristics of strong admin candidates (Burke & Kraut, 2008). A complementary study (Forte et al., 2009) provides an insider’s perspective on the process, using interviews with administrators and editors to reveal that the deliberation process has become increasingly difficult and unappealing to candidates. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 100–109, 2011. © Springer-Verlag Berlin Heidelberg 2011

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

101

These and other studies (e.g., Leskovec et al., 2010) illustrate that the decision making is complex, needing to account for diverse data and subject to intense scrutiny. However, researchers have yet to consider how the processes and tools available in deliberations about candidates for administrator status are employed. Our work is intended to help close this gap by considering the role of shared analytical tools in collective sensemaking (Heer & Agrawala, 2008).

2 Method We conducted an exploratory study to understand the processes, tools, and information that Wikipedians use to decide the outcome of Requests for Adminship (RfA) cases. In our study, we collected and analyzed three sources of data: (1) semistructured interviews with active Wikipedia editors who participate in the RfA decision process, (2) the archived record of successful and unsuccessful RfAs as maintained by Wikipedia, and (3) a review of the tools that Wikipedians use for viewing and distilling the contributions and activities of an RfA candidate. We designed this multi-perspective view of the RfA process to gain a rich understanding of the interplay among people, processes, and tools in the work of collaborative sensemaking about the promotion of select community members to admin status. 2.1 Interviews We conducted a series of semi-structured interviews with 10 experienced editors to discover how they think about their online interactions with other Wikipedians, and particularly how they develop their understandings of specific individuals. A primary focus area in these interviews was participation in the RfA process, including typical ways of participating in RfA cases and the use of different information tools (e.g., edit history counters). The interviewees worked from locations across the U.S., so all were interviewed remotely, using either telephone, instant message, or a combination of these methods. Interviews were transcribed and each transcript was coded by multiple team members using an open theme coding process. 2.2 Content Analysis We reviewed the online discussion transcripts of 6 RfA cases, examining instances in which candidates were approved, in which candidates were denied administrator status, and in which the candidate withdrew. We discussed all the transcripts individually and then identified prominent themes in the deliberative exchanges. 2.3 Tool Review When an RfA case is presented to the community for consideration, a set of tools are provided on the nomination page, allowing participants to explore different user characteristics and counts of system activities for the admin candidate. See Figure 1 for an example of the basic count- and history-based tools that point to information about all candidates. This group of tools offers a set of links to information on regular and special pages in Wikipedia. The one exception in this group is “count [quick],” which resides on Wikimedia Toolserver.

102

K. Derthick et al.

Fig. 1. A set of link-based tools to regular and special pages in Wikipedia that provide counts in different categories related to a given admin candidate’s actions and related status indicators in the system.

This predefined set of tools appears as a block of links on the nomination/ deliberation page for each admin candidate. Figure 2 below shows an additional “toolbox” of links available at the end of each RfA nomination. These links all point to tools residing outside Wikipedia itself on the Wikimedia Toolserver. Like the information tools in Figure 1, these tools allow reviewers to consider the candidates’ actions based on count-based criteria, some of which extend beyond Wikipedia itself. We examined this tool set, observing the affordances of each available tool for supporting an editor attempting to arrive at an evidence-based view of a candidate administrator.

Fig. 2. A toolbox provided with each RfA nomination, allowing reviewers to explore different dimensions of the candidate’s actions and presence in Wikipedia and other Mediawiki projects.

3 Results To understand how Wikipedia participants decide who should be promoted to admin status, we considered the formal process from the perspective of the participants, from evidence of their deliberative work, and from the system tools used to support the process. Our interviews and reviews of processes and tools yielded detailed insight into the RfA process. These results are presented below with selected quotes to illustrate each of three themes that emerged from our analysis: styles of interaction,

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

103

social networks, and counts of online activity versus the contexts in which those activities occur. These results are then explored in the discussion section. 3.1 Styles of Acting and Interacting A primary means by which an online community develops an understanding of its members is by interpreting the actions and interactions of individual participants. A sense of who a person is develops as that individual’s behavior is assessed—whether that behavior is related to individual acts performed in the shared space of the system, or whether it is his or her behavior when engaged in dialogic exchanges with others. Within Wikipedia, an editor’s actions and interactions are considered important sources of evidence when the community deliberates about the potential promotion of a candidate editor to admin status. Actions and interactions on user talk pages, on article talk pages, in edit summaries, and in private backchannels such as email or IRC chats, become evidence about which the community can deliberate. When deliberating about RfA candidates, editors seek evidence that an editor has been civil when engaging with others during their work. As one interview explained: I tend to value civil interactions. . . . Wikipedia has a few people who are good content contributors, but are also possessed of tempers and tend to blow up at people. In [such] case[s], it’s weighing value: are their contributions worth the people who might leave because they belittle or harass them, or the environment it creates for others? – Dylan Editors who engage in controversial issues with a level-headed approach are also viewed favorably, particularly since working productively in a heated exchange is a necessary part of doing administrative work. As one editor explained, Participation in highly controversial articles will gain people ideological enemies, but at the same time present a very stringent test of the person's ability to deal with complex and emotive issues, interact[ing] with disruptive editors. [These] qualities are often in demand of administrators. – Taylor Yet another kind of interaction style that is considered when admin candidates are being reviewed is how they have engaged “newbies,” whose initial contributions are sometimes more damaging than valuable. A productive, encouraging interaction style when dealing with new editors is important because the community as whole depends on the infusion of new contributors to conduct the overall work of the system. Admin candidates, thus, are expected to exhibit an interaction style that encourages new contributors to work productively within community norms. One interviewee used a personal example to illustrate how this interaction style may be exhibited when interacting with a new contributor who appears to be engaged in disruptive editing: Well, they get the standard warning template(s) that all vandals get, but I’ll also add a personal note underneath the template. Nothing too cuddly or kind -they're disrupting the encyclopedia, after all -- but “I’ll grant you, that was hilarious...but please stop before I have to report you to the admins” is the type of language I'd use. What I don’t do is advise them on contributing in a constructive way -- I figure that even if they have the means they pretty much need to decide that on their own, as I did. – Bill

104

K. Derthick et al.

As this quote suggests, the ways that an admin candidate has interacted with newbies is scrutinized to ensure that the candidate is oriented more toward enculturating newcomers than in executing the rules of Wikipedia. Interpreting such past interactions, however, is not easy. A strong candidate for adminship must know the system rules and norms, but also find productive ways of interacting with people: If they’ve nominated a new editor’s bad page for deletion, [that action] is often the subject of scrutiny at RfA. – Dylan While much of the sensemaking that reviewers engage in while considering admin candidates focuses on arriving at a collective assessment of how productive their interaction styles have been, there is a related consideration about the network of editors within which the candidate is interacting. 3.2 Social Networks in Sensemaking When developing shared understandings in their deliberations about admin candidates, Wikipedians consider social networks. Participants described how they conceptualized the nature of social bonds, how those bonds are built, and how those social bonds influence the way they come to understand what others are saying. What Constitutes the Social Network? The creation of a social network in a system like Wikipedia is about establishing social bonds. While some participants talked about meeting “random” people who just happened to be interested in similar topics, others didn’t always see it that way. Some form of homophily emerges when people show up at a topic and begin to express similar ideas or when people come to a Wikipedia policy page and begin to discuss a policy decision. Of the many interactions participants might have, some begin the formation of social bonds, and some people become “friends.” When asked about how he defined friendship on Wikipedia, one interviewee explained: I have real-world friends I interact with on Wikipedia, but by the standard of interactions on-wiki, I wouldn’t call them “wiki-friends”. “Friendship” online is held to a lower standard than the real world, so I suppose you could classify a wiki-friend as someone you regularly associate with in a common topic or area of interest, and agree with, or have a rapport with. – Dylan In Dylan’s view, friendships in Wikipedia are based in shared interests, having rapport and seeing things generally the same way. Another participant elaborated on this point to include collaboration on projects and shared work as characteristic of building a social bond: Every editor has a group of friends they talk with, collaborate with, create projects with, and so forth. My group consists of editors with whom I enjoy speaking, people I agree with on policies, people whose talk pages I watch, people I trust to take the right course of action on matters dealing with policy. – Manny The comment by Manny reinforces many of the common conceptions of how the social network is built—the idea that repeated interactions and the enjoyment of those interactions are important. Additionally, as his quote indicates, bonds form around having similar ideas about Wikipedia policies. Manny mentions that agreeing on

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

105

policy results in trust with policy decisions. Manny also mentions using a watchlist to watch for changes in other users talk pages. Watching is an important part of building a stronger social bond. Dylan explained the watchlist concept at length: I should point out that a lot of Wikipedia relationships are defined by who watchlists what pages. That is, Person A has Person B’s user talk page on his or her watchlist, and so when Person C comments on Person B’s page, Person A might chime in ... Without watchlists, keeping up to date on communication or changes is reduced to manual refreshing, which inhibits further communication or interaction. So wiki-friends or collaborators often have similar watchlist circles, reinforcing the same types of connections. – Dylan Many of our participants had the everyday conception of the social network as simply individuals who interact frequently with each other. That knowing of one another through interaction can result in two types of judgments. Given that Wikipedia is big, it is unlikely that an observer can determine the complete social network, which may not be well connected. Influence of Social Networks. The potentially disconnected nature of a social network has implications for sensemaking in the RfA process. In that process, first the candidate for administrative privileges is nominated. Other editors can ask questions and comment on the candidate. The social network of the individuals who nominate, support, or oppose a candidate is one aspect of how others involved in the deliberation come to understand what a candidate has done and how to interpret his or her responses to questions. In one interview, Taylor talks about knowing who is nominating and supporting a particular candidate. In the quote Taylor is trying to explain one way to understand a person’s reputation. Taylor says: “Reputation” is hard to qualify or explain, but you start to recognize editors and see how they act and what they say. If you see a candidate that has been nominated by several well-known editors who you have always found to make useful and sensible comments, this will suggest that people whose judgment you trust have a positive opinion about the candidate. This isn’t mentioned in the official criteria, but I’m sure it plays a large part in shaping people’s first impressions. – Taylor Editors find that a candidate’s circle of friends is useful for understanding RfA candidates: In a much more succinct way this same comment was made by Marshall who says “To some degree I look at how other editors whose opinion I usually agree with have voted.” The previous two comments show that both the social network around the candidate and the social network around the reviewer matter. The intersection of these social networks is hopefully the people who are most qualified to judge the many aspects of the candidates’ attributes. While the social network around the candidate and the reviewer are important, Taylor notes that the network is not adequate as the sole basis for a judgment:

106

K. Derthick et al.

This is probably a particularly useful shortcut since Wiki[pedia] is such a large place and the number of people you interact with directly is a tiny fraction of the whole. However, like I say this is a shortcut that may well mislead, so I wouldn’t make a decision based entirely on who supports or opposes, it matters very much what they say and what the candidate says! And of source on what the candidate has done. – Taylor In addition to social network information, activity counts and the contexts for interpreting them are important in the process of deliberating about RfA candidates. 3.3 Considering Counts and Contexts Wikipedia editors often strive to live up to a set of shared values, such as Assume Good Faith, as they sense of the behaviors of other editors, particularly when they are making some decision that has consequences for that editor (e.g., when he or she is an admin candidate). Thinking broadly about an editor, however, can be a difficult information-seeking endeavor considering the vast edit histories that many people have. To address this challenge, participants rely on tools to gain insight into an editor’s history, such as edit counts. Such counts can then be considered in defined distributions across time, articles, or namespaces. Counts thus sometimes perform a thresholding function: I check their edits. I see what they’ve edited, where they’ve edited. Like if you’ve only got a few hundred edits, you’re likely not completely familiar with Wikipedia enough to become admin. Similarly, if you’ve got thousands of edits, but only to a small section, such as only articles about one TV show, or one sports team, or something, you’ve not had enough interaction throughout the entire site to be familiar with it all – Marshall In considerations of RfA candidates, raw edit counts are sometimes applied to specific kinds of work, such as consideration of Articles for Deletion (AfDs): I’ll oppose if they don’t have a track record in AfDs, for example, because I think it’s important for admins to know how to “properly” judge and close AfDs. – Dylan Raw edits counts are used to gauge distributions of work across the system: In general, I like to see a good “mix” in contributions. I generally vote against candidates whose major contributions are to Wikipedia space, instead of articles. – Dylan However, while counts can give a baseline sense of the attention and work of a candidate in the system, our interviewees also noted that attention should be paid to qualities of actions rather than just their counts. This inclination for contextual interpretation of actions is illustrated in Bill’s reflection on his sensemaking habits: An important thing to consider when interacting with any editor -- particularly one who appears to be acting disruptively, like a vandal -- is to look at their contributions. Not all editors do this...frankly, I suspect most don’t. But I always do. Sometimes someone who appears to be a vandal, or a provocateur, or just

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

107

someone who doesn’t have good intentions, is actually acting in perfectly good faith. You can infer this if you are about to warn someone for vandalism but look at their user contributions beforehand and note that, say, up until a day ago this person was doing great work... So perhaps what looks like vandalism, or harsh words, or whatever, to me... perhaps that’s just someone coming across the wrong way. – Bill Unfortunately, existing tools that Wikipedians use do not assist in the iterative exploration between summations and contextualized action. Even when data is provided that goes beyond numbers, it can be difficult for editors to evaluate whether the data is part of a pattern that accurately characterizes the editor, or an isolated event. This evaluation is complicated by the fact that other editors are often involved in the process of evaluating the editor in question, providing selected diffs as evidence in the discussion forum where the candidate is being considered. A “diff” is a page in Wikipedia that shows the differences between edited versions of a page. The diffs an editor chooses to provide may not be motivated by a principle as respected as Assume Good Faith, where an editor tries to create a false impression of a pattern in another editor. Taylor illustrates this point with a hypothetical case where one editor attempts to persuade others that a candidate exhibits bad behavior: Imagine somebody says “X is incredibly rude; look at this comment here (link).” You click on the link and it shows somebody saying something that might be interpreted as rude, but you lack all context. It is like seeing everything people say in sound bites rather than conversation. – Taylor Consequently, he elaborates, the work of principled sensemaking about the candidate requires additional work: You need to do a lot of detective work. For instance, you can see when the comment was made and search back to find the conversation in a talkpage’s archives. You can then read the context and perhaps find out that they were being sarcastic, or making a joke to a friend. – Taylor Actions often need to be considered in their context to be appropriately evaluated in a fashion that aligns with the evaluator’s values. But contextualizing actions is difficult in hindsight, particularly when the evaluator was not present when the action was taken. Of course, this qualitative/quantitative tension in large data analysis is a familiar tension for academic researchers.

4 Discussion Through our study we identify themes that are critical to decision making in the Wikipedia Request for Adminship process. The interview data yielded insight into the different understandings of what community participants should be sensitive to when considering the promotion of someone from their ranks into an administrative role. What participants in the RfA process consider important in their decision varies notably from contributor to contributor. Although within Wikipedia contributors can review community-produced lists and commentary about what makes a good

108

K. Derthick et al.

administrator, there is no requirement that contributors consult such materials when participating. As one participant pointed out, even if such lists or commentaries are consulted, participants in the RfA decision weigh all considerations differently. Further, what a given participant values most in RfA considerations shifts over time. The editors who actively participate in RfA cases (the “regulars”) shifts over time, changing the collective sensitivity to different potential considerations. Through an analysis of RfA cases and the tools that are used in the process, we identified techniques that reviewers use to focus the attention of others involved in the deliberation on specific evidence to support their interpretations, and how that evidence can be a source of shared understanding and differing opinions. The deliberation about whether to promote a candidate to adminship is facilitated within the system itself via an asynchronous wiki-based discussion. The deliberation is public and open to participation from the entire community. The candidate is also expected to participate. The candidate answers a set of questions, some of which are standard, some of which are focused on key events in the community and perspectives that the candidate may have about the practices of the community. Following the questions, discussion ensues where participants use evidence to argue for “support” or “oppose” positions. Sensemaking thus occurs as a collaborative effort of the community as they collectively consider the merits of the candidate. Sensemaking is currently an artifact-based process in which selected pieces of evidence available in the system are introduced and interpreted by the deliberators. These interpretations are subject to debate as participants decide which evidence is relevant to their decision and what the evidence means. The deliberative forum is seeded with a set of tools for retrieving the most commonly considered evidence in deliberations. These tools include such things as a counter of past contributions in different segments of the system and a history of the candidate’s status since joining the system. As the central means of generating shared artifacts for consideration, these tools play a significant role in how sensemaking occurs in online systems. One key part of this sensemaking is the shared interpretation of evidence. The technical tools are very effective at identifying “diffs” contributed by the candidate. Isolated contributions rarely tell the entire story. During deliberation, participants use the tools to explain and provide additional details that contextualize interpretations of the candidate’s actions. Some of the more hotly contested deliberations surround discontinuous behavior by the candidate, such as when a candidate has an early history of being a poor member yet changes to be a strong and valuable member.

5 Conclusion and Future Work Our results suggest a number of important requirements for the design of tools to support collaborative decision-making in large-scale online communities. In particular, systems like Wikipedia suffer from their wealth of activity and behavior data. While each action is recorded by the system, the relationships among interrelated actions are difficult to uncover and interpret. Thus, new tools are needed that can help large-scale communities understand and interpret mass interaction data. Acknowledgments. This research was supported by NSF award IIS-0811210.

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

109

References 1. Burke, M., Kraut, R.: Taking up the mop: Identifying future Wikipedia administrators. In: Burnett, M., Costabile, M., Catarci, T., Ruyter, B., Tan, D., Czerwinski, M., Lund, A. (eds.) Proceedings of the Twenty-Sixth SIGCHI Conference on Human Factors in Computing Systems, pp. 3441–3446. ACM Press, New York (2010) 2. Forte, A., Larco, V., Bruckman, A.: Decentralization in Wikipedia Governance. Journal of Management Information Systems 26(1), 49–72 (2009) 3. Heer, J., Agrawala, M.: Design considerations for collaborative visual analytics. Information Visualization Basingstoke- 7(1), 49–62 (2008) 4. Leskovec, J., Huttenlocher, D., Kleinberg, J.M.: Governance in Social Media: A case study of the Wikipedia promotion process. In: Artificial Intelligence, pp. 98–105. AAAI, Menlo Park (2010), http://arxiv.org/abs/1004.3547

Mining Social Relationships in Micro-blogging Systems Qin Gao, Qu Qu, and Xuhui Zhang Department of Industrial Engineering, Tsinghua University. 100084, Beijing, China [email protected]

Abstract. The widespread popularity and vigorous growth of micro-blogging systems provides a fertile source for analyzing social networks and phenomenon. Currently, few data mining tools can deal with unique characteristics of microblogging systems. In this study, we propose an integrate approach for mining user relationships in micro-blogging systems. The approach starts from macroscopic analysis of social networks by grouping users with the method of maximal strongly connected components (MSCC). Following that, a measure of condensation level of groups are calculated to find out the most influential group , and all groups can be ranked according to this measure; then a new algorithm is presented to evaluate the influence of a specific user within a group. The integrated approach is capable to analyze large amount data sets. It is useful for exploring directions of information diffusion and evaluating the scope and the strength of individual user’s influence in micro-blogging systems. Keywords: Social data mining, micro-blogging systems, information diffusion analysis, and graph mining.

1 Introduction Micro-blogging systems, like twitter.com, let users write a brief text about daily life and update immediately through many different ways, including typing online, text messaging, instant messaging (IM), and sending emails. In recent years, microblogging systems become increasingly popular, and the user scale is huge. A report found that in June 2010, nearly 93 million internet users visited Twitter.com, an increase of 109 % from the previous year1. The large volume of user’s social connection information captured by such systems provides many opportunities for mining social networks and phenomenon. In particular, in a micro-blogging system, users need to explicitly indicate whether they want to hear from another user by “following” or not. Knowing such explicit and directed relationships will allow not only structural analysis of social networks but also enables the study of information flow within the networks. Furthermore, networks in micro-blogging systems overlap heavily with social networks in real life. Many users are familiar with their followers 1

http://www.comscore.com/ger/Press_Events/Press_Releases/ 2010/8/Indonesia_Brazil_and_Venezuela_Lead_Global_Surge_in_ Twitter_Usage

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 110–119, 2011. © Springer-Verlag Berlin Heidelberg 2011

Mining Social Relationships in Micro-blogging Systems

111

and followees not only in virtual world but also in real life [1], and new users generally join an existing network by accepting friends’ invitations. This makes social relationship analysis in micro-blogging systems more indicative of real social networks than analysis in other online communities. So far, most research of user relationships analysis studied blog communities, social network sites (SNS), and collaborative tagging systems [e.g., 2-4], and few studies micro-blogging systems. Existing methods used in these studies, such as Social Network Analysis (SNA), have limitations in analyzing huge volume of data sets, and most of them cannot incorporate information flow directions into the analysis. In this paper, we propose an integrate approach to mine social relationships in micro-blogging systems. In addition to high efficiency that is required to process huge data size, our approach also provides a tool for exploring the directions of information dissemination between users and evaluating individual users’ influence in information dissemination. The rest of this paper is organized as follows: The following section describes related studies and motivation. Section 3 introduces the new integrated approach. A case of implementation of our method is presented in Section 4. Section 5 shows our conclusion of the work.

2 Related Work 2.1 Analysis of Online Social Networks The popularity of online social network services makes social relationships mining come again into the limelight with the new communication platform. A lot of researches explore web user relationships in blog, SNS or other web communities from different perspectives with topological analysis [6], link analysis [7] and network evolution [8]. SNA is one of the most influential methods in social relationships analysis. Aligning centrality measures are adopted to discover communities in blogs [2], and betweeness measures are used to extract natural community structures of social networks by dividing the network nodes into densely connected subgroups [3]. Tyler et al. [9] proposed an algorithm based on betweenness and centrality to discover user groups in email networks. Previous studies show that most of SNA researches emphasized binary interaction data, with direct and/or weighted edges, and they focused almost exclusively on very small networks [10]. With the development of internet, especially the rapidly spread of micro-blogging systems, the limitations of SNA become more and more obvious. The size of data set is booming. SNA, however, is hard to process a huge data set. Besides, the traditional SNA is difficult to explore the directions of information transmission in micro-blogging systems. There are a few of other methods of online social relationships mining. To mine a directed social network from an online message board, Matsumura et al. [11] simplify the algorithms of Influence Diffusion Model (IDM) [12] in which the influence of a user is evaluated by propagating terms among couples via messages. Kazienko and Musiał [13] present a new method of personal importance analysis to discover

112

Q. Gao, Q. Qu, and X. Zhang

personal social features in the community of email users based on calculation of the strength of relationships between network members, its dynamic as well as personal position of the nearest neighbor. Similarly with SNA, these methods have difficulties in processing large volume of data and analyzing directed information flow. 2.2 Graph Theory Graphs are widely employed as general data structures in modeling complex systems and networks. Mining frequent subgraph patterns is an effective way to research characters of graphs. There are two basic approaches for pattern analysis: the aprioribased approach and the pattern-growth approach. Apriori-based approach begins with a small size frequent subgraph, and proceeds in a button-up manner by generating candidates with big “size” frequent subgraphs having an extra edge, vertex, or path [14]. The apriori-based approach is based on breadth first search, and they require a good-sized system working space. Different from the the apriori-based approach, the pattern-growth approach adopts depth first search and this approach expand ‘small size’ subgraphs by adding new vertexes or edges. Relational graph is a special graphic structure in which each vertex is unique. Relational graphs are frequently used for modeling biological network, social network, traffic analysis and internet analysis. Dense subgraph is a type of relational graphs, which usually denotes close relationship within a group. CluseCut and Splat are two main algorithms of frequent dense subgraphs mining [14]. Some new methods of analyzing huge directed networks are brought from graph theory. Samudrala [15] used graph theory to discover protein structures. The algorithm denotes each possible conformation of a residue in an amino acid sequence with the notion of a node in a graph. Each node gets a weight based on the degree of interaction between its side-chain atom and local main-chain atoms, and draw edges between pairs of residue conformations/nodes that are consistent with each other [15]. Cai et al. [16] proposed a regression-based graph matrix approach to explore hidden communities in heterogeneous social networks, and they validated the approach with the Iris and Digital Bibliography and Library Project (DBLP) datasets.

3 An Integrate Approach of Social Relationships Mining 3.1 A General Information Diffusion Model Garton et al. [17] suggested a regular social network can be described as a finite set of nodes that are linked with one or more edges, and we build an information diffusion model to explore the user network and information flow in micro-blogging systems. In the model, nodes indicate users or user groups, and directed edges represent information diffusion between two users. Based on this model and existing research related to mining of huge and directed graph data, a new integrate approach based on this model is introduced in the following.

M Mining Social Relationships in Micro-blogging Systems

113

F 1. Information diffusion model Fig.

3.2 Step 1: User Groupin ng by Information Dissemination Relationships Micro-blogging users spon ntaneously form a large number of groups by variious interests or different region ns. For social relationships mining, an ideal grouping is tthat information exchange with hin a group is frequently while information exchange w with members out of a group is as less as possible. Based on information diffusion model, we adopt maximal strongly y connected components (MSCC) to group users in miccroblogging systems. A MSCC in micro-blogging systems can be defined in this w way: Given a directed graph G = (V, E), where V is a finite set of nodes, and E (E V× ×V) is a finite set of directed edges. Nodes denote users, and directed edges exprress information flow. For ∀a V ∀b V if there is at least one path between a andd b, then the directed graph G is i a strongly connected component. And if G would noot a strongly connected compo onent when any node or edge were added to G, it iis a maximal strongly connected d component (MSCC). Based on the definition of o MSCC, we defined a user group as a set of nodes witthin which any two nodes can transfer t information bidirectionally. That is, any two ussers can’t transfer information bidirectionally except for that they belong to the saame MSCC.

∈， ∈，

∈

3.3 Step 2: Group Rankin ng by Information Dissemination Paths between Groups Obviously, the importance of groups in information dissemination is different. For this reason, we propose a method m to rank groups according to their contributionss in information dissemination. Each group is denoted as a a node, that is, the internal relationships within a grooup are masked and only the relationships r among groups are visible. The network oof a micro-blogging system is th hen condensed into a directed acyclic graph G´ (Detailss of proof are showed in Appeendix), and each node of G´is a MSCC. We adopteed a topological sorting algorith hm to rank groups in G´with necessary modifications. T The node without any informaation outflow is deleted from G´and put at the endd of the ranking list. This step iss repeated till all nodes are deleted. In the final ranking llist,

114

Q. Gao, Q. Qu, and X. Zhang

the first one is the strongest influential group in the micro-blogging system. The pseudocode of this algorithm is showed as follows: P<Set> Empty list that will contain sets of nodes in sequence N

Set of nodes with no outside link

Insert all nodes which have no outside link into N while N is non-empty do insert N into P for each node n in N remove n for each node m with a link e from n to m do remove e 3.4 Step 3: User Influence Estimation by the Probability of Information Dissemination In addition to ranking the influence of groups, knowing the strength of a specific user has on his/her followers within a group, especially in the most influential group, would be interesting from a practical point of view. Dijkstra [5] proposed a wellknown algorithm to explore the single-source shortest path problem for a directed graph, and the algorithm is often used for finding costs of the shortest paths from a node to another node. Inspired by Dijkstra's algorithm, we introduce the concept of width and propose a new index using the name of QIndex to estimate the influence of a certain user in a group. In micro-blogging systems, information transmits via subscriptions (by following others) or retweeting by others. We define the number of nodes from the source node to the target along a path as the distance of that path and define the number of different paths connecting the two nodes as the width. The probability that a piece of information reach the target depends on the length and the width of information transmission paths. The shorter the distance and the more paths existing between two nodes, the more probably information can reach the target node. We assume P as the probability that any user retweets a certain update. Therefore, the probability that the target user can receive this information is p= Ʃ i∈N P di, where N represents the set of all paths from the source to the target, and di is the length of path i. According to observation, it is reasonable to infer that P < 0.5. Therefore, the shortest path from the information source to the target node makes the greatest contribution to p. To simplify the problem, we can set a threshold T and if di is larger than T, pi which reflects the probability that information transmits via path i can be considered approximating to zero. Thus we only consider the shortest path within T. Based on the above inference, we propose an algorithm based on Dijkstra's to calculate the influence of a specific node. For a directed graph G = (V, E), V is a finite set of nodes and E is a finite set of edges and E V×V. The information source node is labeled as vs (vs V). The algorithm is described in following steps:

∈

∈

Mining Social Relationships in Micro-blogging Systems

115

1. Set the distance value to zero for initial node vs and to infinity for all other nodes, and assign a width value to one for vs and to zero for all other nodes. Mark all nodes unvisited, and set vs as the current node, noted as nc. 2. Then we need to calculate the distance and width between the current node nc and all unvisited nodes to which it links. Assume the distance from the source node to the current node is dc, and the width of the current node is wc, n' is an unvisited node which is linked by the current node. Before this round of calculation, the distance and width between the source node and n' is d' and w', and the distance between n' and the source node via nc is dc +1. If dc +1 < d' and dc +1 < T, then the distance between the source node and n' will be replaced with dc +1 and w' will be updated with wc at the same time; if dc +1= d' and dc +1 < T, w' will be changed into w' +1. 3. The current node nc will be marked as a visited node when all of its unvisited nodes are calculated. 4. Set the shortest distance node in all unvisited node as current node nc, and repeat step 2. If there isn’t any unvisited nodes in a distance less than T, QIndex of all visited nodes will be calculated and the algorithm will be finished. If width=0, the QIndex is infinity, otherwise, the formula is QIndex=Distance/Width

(1)

The higher the QIndex, the less probably the target node would receive information from the source node. It is very important to set a proper threshold T based on required accuracy, computing resources and other limitations. In the worst case, the running time of QIndex algorithm is O (ǀVǀ2 + ǀEǀ). Threshold T is proportional to time cost and the level of detail about the final result, that is, a larger threshold T costs more time and gets more detailed results, while the smaller T costs less time and gets more roughly results. If T approximates zero, the time cost of QIndex is close to O (ǀVǀ + ǀEǀ).

4 Validation User and usage data from www.digu.com, a popular micro-blogging system in China were collected. Digu.com was established in February, 2009, and the function and user interface design is very similar with twitter.com. We use snowball sampling method for data collection. Snowball sampling which is a nonprobability sampling method with which future subjects are recruited from acquaintances of exiting subjects, just like a rolling snowball. Twenty users were chosen randomly from the public discussion board of digu.com as ‘seeds’, and the information of user and user relationships were collected for each seed (Table 1). Then all the followers and followees of each seed were chosen as ‘seeds’ again. Our data collection system sends a request to digu.com every 10 seconds, and the whole data collection lasts two weeks (March 23, 2010 - April 7, 2010). Finally we collected 332,122 users and 11,160,822 inter-user connections. Because of computing resource constraints, a smaller sample includes 2,556 users with 35,510

116

Q. Gao, Q. Qu, and X. Zhang

inter-connections were used in further computation. There are 60 users’ relationships changed during data collection. However, the lost data only account for 2.34% of all users so that the effect is limited on the final result. Table 1. A data collection example

Item ID User name Nick name Location Gender Self-introduction Address Homepage Information Privacy The Number of Followees The Number of followers The Number of updates Folloee Follower

11528569 ququjoy Qu Beijing 1 (1-male, 2-female, 3-private) From Chongqing http://pic.minicloud.com.cn/file/default/SIGN_24x24.png http://digu.com/ququjoy false(false-information disclosure, true-information 2 2 7 digu, robot xabcdefg, flyinglin456

Fig. 1. MSCC of the biggest group

Through calculating MSCC, the biggest group which contains 1,426 users is found. Then we randomly choose a user named yoohee1221_to find out the top 5 users most influenced by him. We set T=5, and the result of QIndex calculation shows yoohee1221_ has strong influence on his direct followers, e.g., classyuan, which is not a surprising result. However, although some users don’t follow yoohee1221_ directly, e.g., liuxinwu, QIndex shows that they are as strongly influenced by yoohee1221_ as those directly linked to yoohee1221_ (Table 2.). Result shows using MSCC and QIndex to mine user relationships of micro-blogging systems is effective.

Mining Social Relationships in Micro-blogging Systems

117

Table 2. Result of Qindex

Users classyuan gambol liuxinwu xujun99663 dan123 chervun tuniu harliger zxb888 topidea yuanjuan WDM123 shaun

Distance 1 1 2 3 4 4 4 4 4 4 4 4 4

Width 1 1 2 2 2 2 2 2 2 2 2 2 2

QIndex 1 1 1 1.5 2 2 2 2 2 2 2 2 2

5 Conclusion In this paper, we propose a new approach to mine user relationships in microblogging systems. First, the new method of user grouping by maximal strongly connected components (MSCC) is introduced for social network structure analysis. Second, by condensing the graph and sorting with a topological algorithm, groups are ranked according to their influence on other groups. Third, a new algorithm that inspired by Dijkstra's algorithm is presented to assess the influence of individual users on others. The integrate approach can be applied to explore directions of information diffusion in micro-blogging systems and discover opinion leaders, and the results can serve various purposes, including product advertisement, policy advocacy, viral marketing and other information diffusion applications in micro-blogging systems.

References 1. Java, A., Song, X., Finin, T., Tseng, B.: Why We Twitter: Understanding Microblogging Usage and Communities. In: The Joint 9th WEBKDD and 1st SNA-KDD Workshop 2007, pp. 56–65. ACM, New York (2007) 2. Chin, A., Chignell, M.: A Social Hypertext Model for Finding Community in Blogs. In: 17th Conference on Hypertext and Hypermedia, pp. 11–22. ACM, New York (2006) 3. Newman, M.E.J., Girvan, M.: Finding and Evaluating Community Structure in Networks. J. Phys. Rev. 69(2), 26113 (2004) 4. Girvan, M.: Community Structure in Social and Biological Networks. PNAS 99(12), 7821–7826 (2002) 5. Dijkstra, E.W.: A Note on Two Problems in Connexion with Graphs. J. Num. Math. 1(1), 269–271 (1959)

118

Q. Gao, Q. Qu, and X. Zhang

6. Ahn, Y., Han, S., Kwak, H., Moon, S., Jeong, H.: Analysis of topological Characteristics of Huge Online Social Networking Services. In: 16th International Conference on World Wide Web, pp. 835–844. ACM, New York (2007) 7. Hsu, W.H., Lancaster, J., Paradesi, M.S.R., Weninger, T.: Structural Link Analysis from User Profiles and Friends Networks: A Feature Construction Approach. In: ICWSM 2007, pp. 75–80. ACM, New York (2007) 8. Golder, S., Wilkinson, D., Huberman, B.: Rhythms of Social Interaction: Messaging within a Massive Online Network. J. Com. and Tech. 2007, 41–66 (2007) 9. Tyler, J., Wilkinson, D., Huberman, B.: E-Mail as Spectroscopy: Automated Discovery of Community Structure within Organizations. J. Info. Soc. 21(2), 43–53 (2005) 10. McCallum, A., Wang, X., Corrada-Emmanuel, A.: Topic and Role Discovery in Social Networks with Experiments on Enron and Academic Email. J. Arti. Inte. Res. 30, 249–272 (2007) 11. Matsumura, N., Goldberg, D., Llorà, X.: Mining Directed Social Network from Message Board. In: 14th International Conference on World Wide Web, pp. 1092–1093. ACM, New York (2005) 12. Matsumura, N.: Topic Diffusion in a Community. In: Ohsawa, Y., McBurney, P. (eds.) Chance Discovery, pp. 84–97. Springer, Heidelberg (2003) 13. Kazienko, P., Musiał, K.: Mining Personal Social Features in the Community of Email Users. In: Geffert, V., Karhumäki, J., Bertoni, A., Preneel, B., Návrat, P., Bieliková, M. (eds.) SOFSEM 2008. LNCS, vol. 4910, pp. 708–719. Springer, Heidelberg (2008) 14. Han, J., Kamber, M.: Data Mining: Concepts and Techniques, 2nd edn. Morgan Kaufmann, San Francisco (2006) 15. Samudrala, R., Moult, J.: A Graph-theoretic Algorithm for Comparative Modeling of Protein Structure. J. Mol. Biol. 279(1), 287–302 (1998) 16. Cai, D., Shao, Z., He, X.F., Yan, X.F., Han, J.W.: Mining Hidden Community in Heterogeneous Social Networks. In: The 3rd International Workshop on Link Discovery, pp. 1–26. ACM, New York (2005) 17. Garton, L., Haythorntwaite, C., Wellman, B.: Studying Online Social Networks. JCMC 3(1), http://www.ascusc.org/jcmc/vol3/issue1/garton.html

Appendix Lemma 1. Given , and is a directed acyclic graph.

,

, where

is the condensation of G. G'

Proof. Suppose that there is a cycle c v, e in G and ∀p, q v. Then we get p and q are reachable from each other. Thus c is a strong connected subgraph of G , which is contradictory to the condition that G is the condensation of G. Therefore, G is a directed acyclic graph. Lemma 2. Given a directed acyclic graph any outside link from node n.

,

. If

, there will not be

Proof. Suppose the size of set V is L and nq has an outside link to nq. Ifn has not any outside link, the lemma will be proved; if nq has an outside link to nr, it is easy to prove that r≠p≠q for the reason that G is a directed acyclic graph. Continue this iteration until nr has no outside link, then the lemma will be proved. After L times

Mining Social Relationships in Micro-blogging Systems

119

iterations, nL will be get. We hypothesize there is an outside link from nL to ni. Because G is a directed acyclic graph, the inference that i≠1, …, L and the hypothesis that the size of V is L become contradictory. Therefore, there is at least one node in G which has no outside link.

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations Nirupama Dharmavaram Sreenivasan, Chei Sian Lee, and Dion Hoe-Lian Goh Division of Information Studies, Wee Kim Wee School of Communication and Information, Nanyang Technological University, 31 Nanyang Link, Singapore 637718 [email protected], {leecs,ashlgoh}@ntu.edu.sg

Abstract. Microblogs have received considerable attention in crisis communication due to its real-time nature. Despite this, research has not adequately addressed how users make use of information amongst the vast amount of real-time content available in a crisis. Hence, this study aims to understand information use in crisis situations by employing microblogs. Taylor’s [4] information use environment (IUE) model provides theoretical background for this study. According to this model, there are eight classes of information use. We focus on the IUE surrounding the 2010 Icelandic volcano eruption. Our study uses Twitter to analyze users’ postings related to this volcano. The types of postings were ascertained using content analysis. Our findings suggest enlightenment to be the largest category of information use. Other categories such as status messages, problem understanding and factual data were also reported. Further, humour not previously identified by Taylor [4] emerged as a substantial class of information use. Keywords: Microblogs, Twitter, real-time communication, information use, crisis communication.

1 Introduction Microblogs have become a popular Web 2.0 tool because of their valuable source of user-generated content. Recently, microblogs have received much attention in the field of crisis communication due to their real-time nature of communication [1, 2]. For instance, past research by Hughes and Palen [2] compares the behaviour of microblog users in mass convergence and emergency events from more general use. Study by Starbird, Palen, Hughes, and Vieweg [1] describes the mechanisms of information production and distribution on microblogs during a critical situation. However, owing to the sheer amount of real-time information potentially available in a crisis, it is often not known how users make use of such information. Moreover, the types of information requested or shared by users on microblogs during a crisis situation are also not known. Hence, understanding information use in crisis through the lens of microblog communication is an important research area, but to our knowledge, there are limited studies that have been conducted. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 120–129, 2011. © Springer-Verlag Berlin Heidelberg 2011

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

121

Thus there is a need for scholarly research to assess information use in microblogs to investigate how users make use of the heterogeneous and distributed information available during crisis situations. The need arises because it is not known how users use the diverse information from microblogs to address their problems in a crisis situation. For example, when a volcano eruption occurs, people create many microblog postings related to the event. This enables users to detect and monitor the volcano occurrence in an up-to-date manner, to address their needs, by simply observing the posts. It is important to understand how users make use of the vast amounts of information available in crisis to make it meaningful for them to understand the scale and extent of the crisis. The link in the chain of such microblog communication in crisis, from real-time detection of events, to dissemination on this platform depends on understanding information use [3]. Therefore the objective of this study is to ascertain the role played by microblogs to support information use in crisis situations. In order to understand information use in a particular setting such as a crisis in microblogs, we use Taylor’s [4] information use environment (IUE) model in our study

2 Literature Review According to Taylor [4], information use is seen as a process where information sought or received by users is put to use by making sense of it so as to address their problems. Taylor [4] looked at the user and the uses of information, and the contexts within which those users deem information that will be useful to them at particular times. Taylor [4] associated information use to be an integral part of users’ everyday life where they construct and process information. Similarly, a context is a setting where users disseminate, seek and make use of information. This classical model proposed by Taylor [4] has been referred as information use environments (IUE). IUE places emphasis on the set of elements that affect the flow and use of information into, within, and out this context. IUE also determines the criteria by which the users judge the value of information. Taylor’s [4] model foreshadowed much of the current contextually-focused research by recognizing four influences on information use namely: individuals involved, constraints and opportunities provided by the context, problems faced and attitudes and approaches to problem resolution. This study makes use of Taylor [4] as a theoretical framework to develop a holistic understanding of the context in which users resort to information use in microblogs. A well-known IUE study by Davenport [5] also highlighted the importance of context in studying information use. However, we chose Taylor’s [4] seminal work as it is one of the pioneering studies done on IUE and has been incorporated in several contextual studies by Kuhlthau [6], Choo [7] and others. Taylor [4] emphasized IUEs to be the bridge between users and their environments whereby the context affects the availability, access to, and use of information. Additionally, Taylor [4] identified eight classes of information use by individuals to resolve their problems: enlightenment, problem understanding, instrumental, factual, confirmational, projective, motivational and personal. This paper focuses on these classes of information use to categorize the types of information used during a crisis situation. Our study describes a crisis as an

122

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

IUE to understand this context within which the affected users make choices about what information will be useful to them. Crisis may be seen as an interruption in the economic and social lives of individuals in the form of various ranges of situations [8]. For example, crises can be of various types such as natural disasters, human-made crises including terrorism attacks, epidemics such as spread of H1N1, nuclear crisis, environmental crises such as oil spill and many more. As crises present highly complex information environments, management of information before, during and after such emergency situations becomes important to influence the way in which such situations can be managed [9]. Crises events in the past have also demonstrated the need to understand how individuals, government agencies, public relation professionals and other nongovernment agencies collect, organize, share, manage and disseminate information during such situations [1]. From an information perspective, crisis situations are demanding owing to the various information challenges that can surface. Some of the problems that individuals face during crises include information overload, lack of relevant information and changing information needs at different stages of a disaster [8]. Other information challenges include integration and coordination of the information from diverse actors involved, information dissemination at the right time via multiple channels and evaluation of the trust in the information sources [8]. Given this background, our study aims to investigate the types of information potentially used by users in crisis on microblogs. As much as its ability to allow fast exchange of information between users, this platform also differs from other social media because of its ability to broadcast concise real-time updates that can be easily accessed using text messaging via mobile communication devices, Web and other third-party applications. With this ability to send messages using mobile devices and also be able to easily broadcast them to a wider audience, they seem to be fit for use during crisis events. Recently, microblogs have received much attention in the field of crisis communication owing to its growing ubiquity, ability to rapidly communicate and cross-platform accessibility [1, 3]. Despite the fact that these studies have suggested that microblogs are effective in handling emergency communications, limited work has been done to understand how users make use of the vast amounts of real-time information on microblogs during crisis to support their needs. Identifying users’ information usage on microblogs in crisis is important for several reasons. For example, microblogs users may turn to this medium in crisis to learn about real-time events, seek information, share information or communicate with affected agencies [9]. Owing to the vast amount of information available in crisis, users often have difficulty in the early phases to search and make use of relevant information to solve their problems. Even when they begin with a direction in mind, many may become confused and uncertain about how to proceed after a short period of time [1]. Hence, this research gap motivates our study. Our goal is then to examine whether microblogs can facilitate information use. We do this by uncovering categories of information use on microblogs during a crisis situation. Our study shares similarities with prior work on microblogs on event-based user behaviour [2] and usefulness of microblogs in emergencies [9], among others, in that we aim to investigate how users adopt microblogs to communicate with each other during crisis. Nevertheless, there are distinct differences between this study and

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

123

past work that warrant the current research. For example, the study by Hughes and Palen [2] compared features of user behaviour in Twitter during major events .In contrast, our study investigated microblog postings in a crisis situation to understand the potential uses of information. Next, Mills, Chen, Lee, and Rao [9] examined the usefulness of Twitter as a medium for emergency response. On the other hand, our study drew from Taylor’s [4] classes of information use to examine the categories of potential uses in microblogs during a crisis. With this stimulus, our study aims to explore the following research question: RQ: How do users employ microblogs to support information use in a crisis?

3 Data Collection To investigate the research question, Twitter was chosen because it provides a realtime perspective of what people are thinking and doing. According to recent statistics from comScore, Twitter has grown from 37.3 million unique visitors in May 2009 to 90.2 million in May 2010 [10]. Twitter was launched on July 13, 2006 and allows users to post tweets or updates to a network of other users. Tweets are text-based, up to 140 characters in length and can be posted via multiple means such as Twitter.com, text messaging, instant messaging, mobile devices or from other third party clients. Users can indicate if their tweets are public or private. By being public, tweets would appear in reverse chronological order on the public timeline on Twitter’s home page or on the user’s Twitter page where all the updates are aggregated into a single list. Users can follow and read each other’s tweets without giving mutual permission. By being private, tweets are visible only to a user’s followers. Followers are a class of users who receive each other’s updates on their home page. Our paper reports on the use of Twitter during the 2010 Icelandic volcanic ash eruption that forms the IUE for our study. This eruption that began late March 2010 caused a large scale disruption for air travel across several airports in Europe. Due to the volcanic ash, many airports across Europe were shut down and thousands of travellers all over the world experienced flight delays and cancellations. This large scale crisis affected the airline industry, by costing them approximately 130 million pounds a day [11]. Based on this crisis, we characterize the categories of information use during and following the crisis. We chose the IUE surrounding the volcano in Iceland for analysis, as it provided an adequate coverage of tweets that was required for examining information use closely in a critical situation. In our study, tweets are defined as posts made by any Twitter user. The collection dates for user tweets posted during the crisis were from April 13, 2010 to May 5, 2010. The data collection period was determined based on the timeline when the Icelandic volcanic ash eruption was at its peak [12]. Queries were submitted to the Twitter search facility and the search terms used corresponded to name of the Icelandic volcano ‘Eyjafjallajokull’, ‘volcanic ash’ and hashtag ‘#icelandvolcano‘. We used hashtags (i.e. ‘#’) to search for user tweets that relate to specific events during this crisis. For the entire data collection period, we harvested a total of 63435 tweets. From this, we sampled 15% of tweets per day over the entire time period after eliminating duplicates and non-English tweets, resulting in 9641 tweets.

124

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

4 Findings and Analysis 4.1 Coding Scheme This study adopted qualitative content analysis as it uses a naturalistic paradigm to interpret meaning from text data content [13]. Hsieh and Shannon [14] defined this research method where content of text data was subjectively interpreted through a systematic process of identifying themes or patterns. This qualitative content analysis approach is appropriate for our study to inspect the content of the postings to examine the various types of information that users used during crisis situations. To enable this, coding schemes were developed to facilitate content analysis for the user tweet content. We adopted the coding strategy from Taylor’s [4] eight classes of information use. After this, we used the entire sample of tweets (N=9641) to check if Taylor’s [4] categories of information use could be coded to the data. An iterative coding process was used to consolidate and finalize these codes either through combining or omitting existing codes that were adapted from Taylor [4] or developing additional codes if any. Each category then became a label for assigning meaning to the descriptive or inferential information found in the tweet content to categorize them based on Taylor’s [4] eight classes of information use. For example, tweets from users providing contextual information to better understand the situation were categorized as Enlightenment while tweet content providing a detailed explanation of specific problems were categorized as Problem understanding. If the tweet content did not correspond to any of the eight information use categories of Taylor [4], then additional categories were formulated. After finalizing the categories, all tweets in our sample were coded. Each tweet could be coded into more than one content category. The details of all categories used in the coding schemes for tweets are shown in Table 1. The coding consistency was measured through an assessment of inter-coder agreement to ensure that all coding issues related to the definitions of categories were discussed and resolved. Two coders were used in the study. Both were graduate students who were familiar with microblogging. An inter-rater reliability analysis using Krippendorf’s alpha was performed to determine the agreement among the two coders. The software package, SPSS was used to compute this statistic. For the entire tweet sample, the inter-rater reliability for the coders was found to have Krippendorf’s alpha value of 0.82, indicating strong agreement [15]. 4.2 Results In terms of supporting information use in crisis among users, our results in Table 1 show that microblog communication was mainly used for enlightenment, to provide personal updates, for problem understanding and reporting of factual data. In particular, our analysis indicated that microblogs was used primarily to understand the contextual information of the critical situation (23.5 % of the user tweets) so as to make sense of it. Our findings showed higher percentages for this category as users were perhaps making use of the background information to get an idea of the scale of crisis. For example, a user tweet was used to provide background information on the rise in travel costs due to volcanic eruption in Iceland (“News post: Volcano air,

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations Table 1. IUE categories N=9641 Category Enlightenment

Personal

Problem understanding

Factual

Motivational

Projective

Instrumental

Confirmational

Description

Example

Contextual information to Flight disruptions in better understand the situation Europe get even worse: Thick drifts of volcanic ash blanketed…. http://bit.ly/8YuCz2 Status messages of users Woke up to the news that there was an earthquake in China & a volcano in Iceland. Explanation of particular Iceland farmers try problems to save herds from ash: http://bit.ly/93Cr8N #msnbc Precise data (e.g. numbers, 800 flee as Iceland figures) volcano rumbles http://bit.ly/blzK2p Different perspectives on the Iceland Volcano same problem Was God's Response To British Ad Ban http://tinyurl.com/2c 8kj7s Predicting or forecasting Iceland volcano crisis will hit European airlines worse than U.S., experts say. http://bit.ly/bUWum V Providing instructions on how Stranded by a to handle certain situations volcano? Handy survival guide! http://bit.ly/aug00k Information verification @andrewqh Is that the name of the other glacier volcano in Iceland?

Non-IUE categories Humour

Expressing humour

Irrelevant

Spam

% 23.5

14.7

12.2

11.6

5.3

3.7

3.5

3.1

Iceland, good luck 8.1 and please turn off your volcano before I fly to Europe in two weeks! Silly Kissers…. 2.6

125

126

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

tourism costs may top $3.5 billion in Europe - The Age http://cli.gs/qYs6j”). Next, microblogs were used to update their status messages in the form of what they were facing currently in this crisis. One example, was a user tweeting about his disrupted flight journey during the volcano eruption (“Ready for an 11 hrs flight over that nasty volcano, hope it will let me come back home when I`m supposed to”). Hence our observation on the top two frequently occurring categories supports the notion that microblogs are used for disseminating critical updates and first-person accounts of the disasters [9]. The third highest category of information use was related to in-depth understanding of the problems pertaining to the current situation. For example, a user tweet highlights the problems that stranded travellers are facing due to volcano (“Airlines appeal to passengers to give up seats to travellers stranded by the volcano ash crisis http://is.gd/bGH63”). This user tweet is categorized as problem understanding as it provides a better comprehension of specific problems such as flight delays in the event of air travel disruption because of which airlines are rebooking the tickets of stranded travellers. The fourth category of information use was related to reporting of factual data. One example was a user tweet that reported factual volcano data in the form of numbers (“Eyjafjallajökull: activity still the same - plume height 3-6km, wind now blowing ash to the NW, Strombolian explosions continue #volcano”). Additionally, we found that other categories of information use related to different perspectives on the same problem (5.3%), forecasts (3.7%), instructions to handle certain situations (3.5%) and information verification (3.1%). These showed relatively lower percentages. This was perhaps because a majority of the users were broadcasting real-time information related to the crisis situation instead of concentrating on specific topics. It was also found that humour (8.1%) was a substantial category of information use that was not identified by [4] as shown in Table 1.

5 Discussion On examining the information use categories in user tweets, our results indicate that the main themes of information use could be grouped into topics such as sharing contextual information, providing detailed information on problems, reporting of daily routines and facts. These results support the categorization adopted by past studies that have examined the nature of microblog information during crisis [1, 2]. We found that these categories from our findings were consistent with the broadcastbased information sharing categories outlined previously by Hughes and Palen [2]. Categories such as personal updates were similar to the tweets corresponding to firsthand information synthesis identified by [1]. At the same time, our study also highlighted that users not only share useful updates about the crisis situation but also resort to humour to reduce their stress levels. On a different note, fewer tweets were found related to categories such as asking questions for the purposes of information verification (3.1%) and instructions to handle certain situations (3.5%). This may be attributed to the high information demands of crisis situations. Owing to the rise in demand, users may broadcast real-

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

127

time information to increase their awareness about the situation rather than directing specific questions [2]. Another reason could be that the tweets that contain background information related to the critical situation may also contain instructions and information verification questions in the form of pointers to other sources. For example, user tweets such as “AccuWeather.com - Weather Video - Volcanic Ashes Disrupt Travel: Should travel insurers pay up after volcano disruption?... http://bit.ly/aQC6dD” suggest that the URL contained in the tweet encompasses the contextual information along with the questions pertaining to a specific topic such as travel insurance. Next, our findings on the information use categories reported that users expressed different perspectives on the same problems in their tweets (5.3%). These findings suggest that users looked at the problems that they were facing because of crisis in different ways. For example, user tweets such as as “News Feature: What's God trying to tell us with the volcano in Iceland... http://bit.ly/bko3Ao” indicate that users were looking for spiritual support to manage the volcanic crisis. Another interesting finding from our study showed that humour not identified by [4] was found to be a substantial class of information use (8.1%). The appearance of humour as an emotional manifestation could be attributed to the way in which people vent their stress or frustration during crisis [16]. For example, humour was used to relieve stress while facing problems of being stranded in airports for prolonged periods owing to the volcanic eruption. User tweets such as “Its earth day today...well earth already celebrated its day with fireworks we'd seen in Iceland's volcanic eruption! - stranded at heathrow” indicate the emergence of humour. This finding highlights the importance of microblogs in crisis by suggesting that emotions are used as a coping mechanism to overcome problems.

6 Implications, Limitations, and Future Research Arising from our findings, the following implications for research and practice may be derived. First, this study contributes to our improved understanding of crisis as an IUE. Our findings indicate that crisis influences the way different users make use of real-time information found on microblogs. For example, our findings demonstrate that many users share contextual information about the crisis to understand the situation and be well-informed. At the same time, other users reported on their firsthand experiences of the critical situation by providing insight and varying perspectives to the same problems. These findings imply that users want to take initiative in providing different types of information about the crisis situation to raise awareness and at the same time assist the affected individuals and agencies. If they do not have first-hand knowledge of the critical situation, they rely on other media sources by providing links to pass on information on microblogs. Second, our study contributes to the extant literature on crisis informatics by exploring the role microblogs play to facilitate emergency communication. Our study sought to examine the types of information that users make use of in crisis situations. Moreover, our findings on the emergence of humour as a substantial class of information use during crisis imply that users circulate light-hearted chatter amongst each other to spread cheer and hope. For example, by sharing jokes on Icelandic

128

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

volcano eruption in their tweets, users not only use humour to help each other cope with the crisis but also instill optimism to overcome from the after-effects. This finding implies that in times of crisis, users are hopeful that by being optimistic and cheerful, they can overcome their problems better in crisis. Next, our findings on how users make use of information from the vast amount of real-time information available in crisis have implications for crisis mitigation and risk communication agencies. They can track these postings to take steps to manage the situation and supplement with required information if needed. Doing so may help them improve their overall process of emergency warning, response and recovery procedures. Another implication that can be derived from our findings is that users make use of real-time information on microblogs extensively during crisis. This implies that under such critical situations, it is possible that false and inaccurate information gets circulated that may prove devastating to affected individuals. Our study suggests that crisis mitigation agencies should moderate the microblog postings or set up official microblog accounts during a crisis situation to disseminate accurate and up-to-date information. However, our study has certain limitations that can be addressed in future research. First, we examined only one microblogging site, Twitter, for our analysis. Users of other microblogging services might differ in their information use that may require separate investigations. Second, only one crisis situation was used in our analysis, although a major one. Using more than one crisis situation for investigation, would allow better generalizability of our results. In addition, we only captured Twitter posts from April 13, 2010 to May 5, 2010. It is possible that tweets in a different time period such as the day the crisis was declared over, may yield other varied and useful results not uncovered in this study. However, this limitation does not affect the accuracy of the data used in our study as we chose the timeframe where the volcanic ash eruption was at its peak ensuring an adequate number of tweets for analysis. Finally, future studies could extend our work by examining information use on a different crisis situation or an event to compare and contrast our results. Future research can also focus on exploring the categories of information use that are most important to satisfy user needs during a crisis situation. Our analysis of user-generated content from microblogs can be supplemented with a mixed methods approach using ethnographic studies, focus groups and interviews with Twitter and non-Twitter users to further understand the potential of this new medium for crisis communications. Acknowledgements. This work was supported by Nanyang Technological University Academic Research Fund Program (Tier 1), under research Grant No. 59/09.

References 1. Starbird, K., Palen, L., Hughes, A.L., Vieweg, S.: Chatter on The Red: What hazards threat reveals about the social life of microblogged information. In: Proceedings of the 2010 ACM Conference on Computer Supported Cooperative Work, pp. 241–250. ACM, New York (2010) 2. Hughes, A.L., Palen, L.: Twitter adoption and use in mass convergence and emergency events. International Journal of Emergency Management 6(3/4), 248–260 (2009)

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

129

3. Li, J., Rao, H.R.: Twitter as a rapid response news service: An exploration in the context of the 2008 China Earthquake. The Electronic Journal on Information Systems in Developing Countries 42(4), 1–22 (2010) 4. Taylor, R.S.: Information use environments. In: Dervin, B., Voigt, M.J. (eds.) Progress in communication sciences X, pp. 217–255. Ablex Publishing Corp., Norwood (1991) 5. Davenport, T.H.: Information ecology: Mastering the information and knowledge environment. Oxford University Press, New York (1997) 6. Kuhlthau, C.C.: Seeking meaning, 2nd edn. Libraries Unlimited, Westport (2003) 7. Choo, C.W.: The knowing organization: how organizations use information to construct meaning, create knowledge, and make decisions, 2nd edn. Oxford University Press, New York (2006) 8. Coombs, W.T.: Ongoing Crisis Communication: Planning, Managing, and Responding, 2nd edn. Thousand Oaks, CA (2007) 9. Mills, A., Chen, R., Lee, J., Rao, H.R.: WEB 2.0 emergency applications: How useful can Twitter be for emergency. Journal of Information Privacy & Security 5(3), 3–26 (2009) 10. Latest comScore stats show Twitter growth is still strong, http://techcrunch.com/2010/06/18/latest-comscore-stats-showtwitter-growth-is-still-strong/ 11. Ash cloud costing airlines £130m a day, http://www.guardian.co.uk/business/2010/apr/16/icelandvolcano-airline-industry-iata 12. Met Office, http://www.metoffice.gov.uk/corporate/pressoffice/2010/ volcano.html 13. Cavanagh, S.: Content analysis: concepts, methods and applications. Nurse Researcher 4(3), 5–16 (1997) 14. Hsieh, H.F., Shannon, S.E.: Three Approaches to Qualitative Content Analysis. Qualitative Health Research 15(9), 1277–1288 (2005) 15. Krippendorf, K.: Content Analysis. Thousand Oaks, CA (2004) 16. Kendra, J.M., Wachtendorf, T.: Reconsidering Convergence and Converger Legitimacy in Response to the World Trade Center Disaster. Research in Social Problems and Public Policy 11, 97–122 (2003)

Impact of Blog Design Features on Blogging Satisfaction: An Impression Management Perspective Wee-Kek Tan and Hock-Hai Teo Dept. of Information Systems, National University of Singapore, 21 Lower Kent Ridge Road, Singapore 119077, Republic of Singapore {tanwk,teohh}@comp.nus.edu.sg

Abstract. Trends observed among bloggers worldwide suggest that selfpresentation is a main motivation to blog and a greater majority of bloggers do so with the aid of third party hosting services and tools. Thus, providing tools that are designed to meet the usability requirement of bloggers with respect to self-presentation may increase blogging satisfaction. Grounded on the theory of impression management, we propose that a blogger actively engages in impression management with content, functional and aesthetic blog design features. This can lead the blogger to obtain a perceived confirmation from the readers of his or her identity, i.e., perceived identity verification. Perceived identity verification may heighten blogging satisfaction since it helps the blogger to achieve the goal of self-presentation. Results obtained from an online survey, which was preceded by a focus group discussion, provide support for our conjecture. Theoretical and practical implications are also discussed. Keywords: Blog, blog design features, satisfaction, self-presentation, theory of impression management.

1 Introduction Blogging is a highly popular online activity. According to survey statistics from the State of the Blogosphere 2009 [1], the majority of bloggers on the Internet are casual bloggers who seek to express their life, and to share their expertise and experience. Beyond personal motivations, there is also a growing trend in using blog to make money and to exert political influence. These wide-ranging impacts are being wielded by the majority of bloggers using third party hosting services and tools [1]. It is therefore important to develop an in-depth understanding of how blog tools may be designed to meet the usability requirements of bloggers [2]. Since the majority of bloggers use their blog to express their life and personality in the form of online journal [3], i.e. self-presentation, it makes sense to provide design features that support this main motivation. To this extent, functional features to aid in blogging and the ability to customize the blog’s aesthetic elements are two key factors affecting bloggers’ choice of third party tools [1]. Moreover, most bloggers consider personal satisfaction as the key success measure of their blogs [1]. This research thus A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 130–139, 2011. © Springer-Verlag Berlin Heidelberg 2011

Impact of Blog Design Features on Blogging Satisfaction

131

examines how content, functional and aesthetic design features of blog, which aid the blogger in achieving self-presentation, may affect blogging satisfaction. The findings from this research can potentially help researchers and practitioners to design effective blogging tools. Besides blog hosting service providers such as Blogger, social networking sites such as Friendster that incorporate blogging as a pertinent feature may also benefit from this research.

2 Theoretical Background Bloggers typically express their identity on their blogs in the form of personal information such as age and location, emoticons, language styles, occupations, and hobbies. Over a period of time, it is possible for a blogger to develop virtual identity [4]. One fundamental theoretical notion underpinning this process is that individuals proactively present their identity to others using props as explained by the theory of impression management. 2.1 Theory of Impression Management The impressions that individuals leave on other people typically affect how they are perceived, evaluated and treated [5]. Consequently, individuals are inclined to control the impressions that others form of them, i.e., impression management. According to the theory of impression management, this process draws a metaphoric similarity with theatrical performance [6]. In a performance that takes place in a public setting, the actor prepares the required props and scenery in the back-stage before performing with the props and scenery on-stage. In a similar fashion, an individual makes a conspicuous effort to acquire knowledge, skill and props so that s/he can use them to project certain impression upon other people. 2.2 Impression Management on the Web Researchers have found that personal home pages are being used by their authors to publish information on topics deemed important to them together with their social contexts [7]. To this extent, personal home page is an inclusive self-presentation tool that explicates the identity of its author [8]. Indeed, the author of a personal website makes a conspicuous effort to adorn his or her web pages with identity claims that present a coherent online identity [9]. The identity claims could include textual and non-textual contents reflecting the interests, preferences and values of the author, i.e., the props [6]. For instance, aesthetic elements (e.g., image, video and sound) and functional elements (e.g., hyperlink to website related to the author’s interest). 2.3 Blog Design Features for Impression Management Within the context of blog, self-presentation may be applied by bloggers through the creation or usage of blog template and manipulation of the underlying Hypertext Markup Language (HTML) code [10]. Our present study, however, goes beyond the use of basic website elements such as image and hyperlink to examine how

132

W.-K. Tan and H.-H. Teo

sophisticated functional and aesthetic blog design features provided by third party service providers can be used for self-presentation. We shall term this as impression management with blog design features. Functional features such as blogger profile and blogroll are required for bloggers to share their life and experiences with readers. Aesthetic features such as blog header and background enhance the visual appeal of the blog and sustain readers’ interest. In addition, the most important characteristic of blog is the posting of journal entries in a reverse chronological order by the author. Most third party blog tools aim to automate this process in the simplest way possible [2]. For this reason, we consider the content feature as another important dimension of blog design. Content feature aids the blogger in posting both textual and non-textual materials that reflect his or her identity over and above what the functional features can accomplish. In gist, content, functional and aesthetic design features collectively make up the props required by the blogger to present his or her identity to the readers. Fig. 1 depicts sample screenshots of the three categories of blog design features currently available on third party blog service providers.

Fig. 1. Screenshots of blog design features

2.4 Presentational Expectancies Although the theory of impression management explains how bloggers could use the various blog design features as props to present their identities, there is a non-trivial caveat that needs to be taken into consideration. Researchers have noted that social anxiety could develop if an individual is motivated to make a desired impression to a target audience but is hesitant that s/he has the capability to do so [11]. That is, an individual’s self-presentation behavior can be affected by presentational outcome expectancy and presentational self-efficacy expectancy [11, 12]. The former refers to the belief that projecting certain impression will lead to the desired goal and the latter refers to the belief on one’s ability to present that impression.

Impact of Blog Design Features on Blogging Satisfaction

133

3 Research Model and Hypotheses Development Drawing on the theory of impression management as the theoretical lens, we put forth our research model (see Fig. 2) to understand how the provision of various blog design features may fulfill the usability requirement of bloggers in term of impression management.

Fig. 2. Research model

3.1 Impression Management Process and Its Impacts In a technology-mediated setting, tools that facilitate rich social interaction can aid in achieving shared understanding among the communicating parties [13], which is helpful for identity communication. In particular, personal website has often been heralded as a new form of identity communication [14]. Prior research has suggested that it is possible for personal website to communicate nonverbal cues through the use of hyperlinks, emotions, animations and other technological conventions [15]. Collectively, these website elements form the props for individuals to present themselves to the public using the personal websites as the performing stage [6]. In a similar fashion, bloggers are known to present themselves through their blogs [10]. A blogger actively using the various content, functional and aesthetic design features can put across his or her desired impression in a consistent and focused manner to readers of their blog. This can give the blogger a perceived confirmation from the readers of his or her identity, i.e., perceived identity verification [16]. We thus hypothesize that: H1: Impression management with blog design features is positively related to perceived identity verification. In an online community, members whose identities are verified and recognized by others will feel that they are better understood and will be treated in the desired way [16]. Thus, communications among members should have lesser misunderstandings and conflicts. Furthermore, members will be inclined towards increasing their attraction to other community members and commitment to the online community [17]. In a real world setting, identity verification can lead to a satisfying interpersonal relationship because of the resulting shared understanding forged through the verification process [18].

134

W.-K. Tan and H.-H. Teo

In sum, identity verification should lead to contentment with an individual access to community resources, i.e., satisfaction [16]. In our context, we define satisfaction as whether the blogger is contented with the blogging experience to express one’s life and to share one’s expertise and experience with the readers. Accordingly, we hypothesize that: H2: Perceived identity verification is positively related to satisfaction. We also propose a second direct causal pathway between impression management and perceived identity verification. In an online community, identity communication can often help members to connect to others who share similar interests, attitudes and experiences [19]. People with similar personality are more likely to communicate and build relationships [20]. Impression management aids the blogger to present their identity to the readers and thus attracts those who share similar identity. It thus aid bloggers to achieve their goals of expressing themselves and share their experiences and expertise with similar others. Indeed, it is possible for bloggers and their readers to form a community of similar interest [3]. Satisfaction may be viewed as a positive affective arousal on the part of the blogger towards his or her blogging experience [21]. It may be induced in the presence of an advancement factor that brings the blogger closer to the attainment of the self-presentation goal [22]. We posit that impression management with the aid of content, functional and aesthetic design features constitute this advancement factor. We thus hypothesize that: H3: Impression management with blog design features is positively related to satisfaction. 3.2 Antecedents of the Impression Management Process Outcome and self-efficacy expectancies have been widely studied in the extant literature on individual’s use of information technology (e.g., [23]). Both computer outcome and self-efficacy expectancies are known to reduce computer-related anxiety and increase computer usage [23]. Within the online context, Internet self-efficacy expectancy has also been found to enhance individuals’ perception of the ease of use of Internet technology and consequently higher behavioral intention and actual usage behavior [24]. In this study, we adapt the original conceptualization of presentational outcome and self-efficacy expectancies [12, 25] to the context of the blogger using blog design features to manage one’s impression. Presumably, high degree of presentational expectancies should reduce a blogger’s anxiety towards the use of the various blog design features to manage his or her impression and consequently increase their propensity to do so [12]. Accordingly, we put forth our final two hypotheses: H4: Presentational outcome expectancy is positively related to impression management with blog design features. H5: Presentational self-efficacy expectancy is positively related to impression management with blog design features.

Impact of Blog Design Features on Blogging Satisfaction

135

4 Research Methodology 4.1 Research Design This study primarily adopted a quantitative survey design to empirically validate the research model depicted in Fig. 2. Specifically, an online survey technique was used for two reasons: 1) an online survey would have been more convenient for bloggers to participate; and 2) online survey allowed us to administer visual stimuli for the various blog design features to enhance the questionnaire. In addition, we conducted an initial focus group discussion with a group of experienced local bloggers to help us gather insights on two issues. First, we wanted to determine whether the majority of local bloggers do indeed blog to express their life, and share their expertise and experience, as reported by the State of the Blogosphere 2009 [1]. Second, we wanted to determine the blog design features that the bloggers would most likely use for impression management. This was to aid us in preparing a representative set of features for the visual stimuli of the survey questionnaire. 4.2 Focus Group The focus group was conducted with ten bloggers over two sessions of five bloggers each. They were recruited through an online invitation posted on the website of a popular online community for local bloggers. The proceedings of the discussion were audio recorded and transcribed subsequently. In the first part, the participants were asked to share their motivations for blogging and the types of contents posted on their blog. Consistent with the findings of the State of the Blogosphere 2009, six out of ten participants explicitly mentioned selfpresentation as well as experience and expertise sharing as the main motivation for blogging: “I use my blog as a diary that is opened to others, especially my friends and family…; I write just about anything that happened to me on my blog…” Three other participants indirectly hinted at self-presentation as their blogging motivation: “I mostly gossip about what happened in the office…; I give my takes on current affairs, well sometimes I grumbled about bread and butter issues…” In the second part, we explained to the participants the purpose of the focus group and also gave them an overview of content, functional and aesthetic blog design features. We then asked the participants to discuss how these features may be used for self-presentation. Most participants highlighted the content feature as the most useful and obvious one in expressing their identity: “I feel that the content of the blog is usually detailed enough to show my identity…” Aesthetic design features were also given the thumbs up: “I make extensive changes to my blog template to reflect my favorite color, hobby and idols…” Functional design features drew mixed responses. Most participants, however, pointed out that certain features were useful: “I placed many links to other blogs and websites associated with my idols… I guess that counts as some form of social affiliation…; the short blogger profile says everything about me…” Towards the end, the participants were asked to rank the various features in descending order of suitability for self-presentation. Various functional and aesthetic design features were identified from the extant literature (e.g., [26]) and given to the participants. Those features with a ranking score greater than the mean were used in the subsequent survey questionnaire as visual stimuli (see Table 1).

136

W.-K. Tan and H.-H. Teo Table 1. Selected functional and aesthetic design features used in the online survey

Functional Blogger Profile Photo Albums Tagboard Links to Non-blog Websites Blogroll Badge

Mean Std. Dev. Aesthetic 18.0 3.300 Post Content 16.7 3.129 Blog Header 14.9 5.971 Content Area 14.8 2.860 Post Title 14.6 3.627 Blog Background 13.0 5.477 Primary Navigation

Mean Std. Dev. 10.7 1.947 9.9 2.514 9.0 1.700 8.5 1.900 7.4 2.989 6.1 3.381

4.3 Operationalization of Constructs The theoretical constructs shown in the research model (see Fig. 2) were measured with instrument scales either adapted from validated scales or self-developed based on established definitions from the extant literature. Except for satisfaction, which is a seven-point semantic differential scale, the remaining constructs used seven-point Likert scales. The scale measuring content impression was adapted from [27]. The scales for functional impression and aesthetic impression were self-developed based on the content impression scale. Presentational outcome expectancy and presentational selfefficacy expectancy were measured using scales adapted from [12] and [29]. Satisfaction was measured with a scale adapted from [28]. For the scale measuring perceived identity verification, we followed Ma and Agarwal’s prescription [16]. 4.4 Survey Design and Procedure The online survey was conducted over a one month period and participants were recruited via an open invitation posted on the same online community. All participants were checked to ensure that they had an existing personal blog that was actively maintained for at least the past 12 months leading up to the survey. The survey proceeded according to the following sequence. Participants first registered their particular and also undergone a simple validation of their blog. They then read the survey instructions before proceeding to read the description of each

Fig. 3. Screenshots taken from actual survey website

Impact of Blog Design Features on Blogging Satisfaction

137

category of blog design features and view the associated visual stimuli. The visual stimuli were spread over three web pages, one per category of features. For content, the participants were shown a sample screenshot of a blog post editor taken from Blogger together with the preview of the post. For the functional and aesthetic features, the participants were shown a skeleton layout of a blog with the various features marked up (see Fig. 3). Participants must view all features in each category before proceeding to the next category. The survey website ensured that they spent at least five minutes on each category to adequately understand the various features. The participants then proceeded to complete the main survey and a short demographic questionnaire. Most of the participants took less than 30 minutes to complete the entire survey.

5 Results 5.1 Descriptive Statistics There were altogether 261 participants who registered. After removing the invalid responses, the final useable sample size was 244. There were 113 male and 131 female participants. The majority of the participants were aged from 21-30 (67.2%) and were locals (82.4%). On the average, the participants had 3.0 years of blogging experience (δ = 2.232) and 66.4% of them blog on a daily or weekly basis. 5.2 Validity and Reliability of Instrument Scales The data were analyzed with partial least square (PLS) based structural equation modeling. The internal consistency, convergent validity and discriminant validity of the measurement model were assessed to be acceptable. The instrument scales were also deemed to be reliable after assessing their Cronbach Alpha scores. 5.3 Hypotheses Testing We then proceeded to test the structural model with the bootstrapping procedure. The standardized path coefficient, standard error and statistical significance are shown in Table 2. The R2 for perceived identity verification is 0.203, satisfaction is 0.237, content impression is 0.381, functional impression is 0.227 and aesthetic impression is 0.233. Table 2. Summary of hypotheses testing results Hypothesis Path Std. Error Sig. H1 CI → PIV 0.253 0.068 *** FI → PIV 0.067 0.077 n.s. AI → PIV 0.231 0.074 *** H2 PIV → SAT 0.202 0.066 ** H3 CI → SAT 0.243 0.068 *** FI → SAT 0.169 0.067 ** AI → SAT 0.026 0.072 n.s.

Hypothesis Path Std. Error Sig. H4 POE → CI 0.257 0.070 *** POE → FI 0.247 0.071 *** POE → AI 0.230 0.081 ** H5 PSEE → CI 0.436 0.074 *** PSEE → FI 0.293 0.071 *** PSEE → AI 0.315 0.072 *** * p < 0.05 ** p < 0.01 *** p < 0.001

138

W.-K. Tan and H.-H. Teo

6 Conclusion The results from the survey generally support our proposition that a blogger treats one’s blog as a performing stage to actively present one’s identity to a public audience of readers [6, 9]. This is accomplished with the aid of the various blog design features as props. Through this process of impression management, a blogger can indeed achieve the goal of expressing one’s life and sharing one’s expertise and experience. Consequently, the blogger derives personal satisfaction with the blogging process. The main theoretical contribution of this study is providing a plausible explanation as to why and how the majority of bloggers seek to express their life, and to share their expertise and experience. By adopting a process model, instead of a variance model, we provide a rich insight into the antecedent and consequential factors involved in the impression management process. This model may be adapted to explain why and how individuals use social computing technologies that have the capability to reflect one’s identity, e.g., social networking sites and virtual worlds. Practically, our research findings emphasize the importance for third party blog hosting service providers to make available a well-balanced basket of content, functional and aesthetic blog design features to their bloggers. This will help the majority of the bloggers in their impression management endeavor and thus leading to greater satisfaction. Presumably, this will improve the acquisition and retention of new and existing bloggers.

References 1. Technorati State of the Blogosphere (2009), http://technorati.com/blogging/feature/state-of-theblogosphere-2009 2. Blood, R.: How Blogging Software Reshapes the Online Community. Communications of the ACM 47(12), 53–55 (2004) 3. Nardi, B.A., Schiano, D.J., Gumbrecht, M., Swarth, L.: Why We Blog. Communications of the ACM 47(12), 41–46 (2004) 4. Moon, J., Li, J.P., Sanders, G.L.: The Role of Virtual Social Identity through Blog Use in Social Life. In: 12th AMCIS, Mexico, pp. 4085–4095 (2006) 5. Leary, M.R., Kowalski, R.M.: Impression Management: A Literature Review and TwoComponent Model. Psychological Bulletin 107(1), 34–47 (1990) 6. Goffman, E.: The Presentation of Self in Everyday Life. Doubleday, New York (1959) 7. Wynn, E., Katz, J.E.: Hyperbole over Cyberspace: Self-Presentation and Social Boundaries in Internet Home Pages and Discourse. The Information Society 13(4), 297– 327 (1997) 8. Schau, H.J., Gilly, M.C.: We are What We Post? Self Presentation in Personal Web Space. J. Consumer Research 30(3), 385–404 (2003) 9. Vazire, S., Gosling, S.D.: e-Perceptions: Personality Impressions Based on Personal Websites. J. Personality and Social Psychology 87(1), 123–132 (2004) 10. Bortree, D.S.: Presentation of Self on the Web: An Ethnographic Study of Teenage Girls’ Weblogs. Education, Communication & Information 5(1), 25–39 (2005)

Impact of Blog Design Features on Blogging Satisfaction

139

11. Leary, M.R.: Understanding Social Anxiety: Social, Personality, Clinical Perspectives. Sage, Beverly Hills (1983) 12. Maddux, J.E., Norton, L.W., Leary, M.R.: Cognitive Components of Social Anxiety: An Investigation of the Integration of Self-Presentation Theory and Self-Efficacy Theory. J. Social and Clinical Psychology 6(2), 180–190 (1988) 13. Preece, J., Maloney-Krichmar, D.: Online Communities. In: Jacko, J., Sears, A. (eds.) Handbook of Human-Computer Interaction, pp. 596–620. Lawrence Erlbaum Associates Inc., New Jersey (2003) 14. Walker, K.: It’s Difficult to Hide It: The Presentation of Self on Internet Home Pages. Qualitative Sociology 23(1), 99–120 (2000) 15. Walther, J.B.: Computer-Mediated Communication: Impersonal, Interpersonal, and Hyperpersonal Interaction. Communication Research 23, 3–43 (1996) 16. Ma, M., Agarwal, R.: Through a Glass Darkly: Information Technology Design, Identity Verification, and Knowledge Contribution in Online Communities. Information Systems Research 18(1), 42–67 (2007) 17. Swann, W.B., Polzer, J.T., Seyle, D.C., Ko, S.J.: Finding Value in Diversity: Verification of Personal and Social Self-Views in Diverse Groups. Academy of Management Review 29(1), 9–27 (2004) 18. De La Ronde, C., Swann, W.B.: Partner Verification: Restoring Shattered Images of Our Intimates. J. Personality and Social Psychology 75(2), 374–382 (1998) 19. Jensen, C., Davis, J., Farnham, S.: Finding Others Online: Reputation Systems for Social Online Spaces. In: SIGCHI Conference on Human Factors in Computing Systems, Minnesota, pp. 447–454 (2002) 20. Newcomb, T.M.: The Acquaintance Process. Holt, Rinehart and Winston, New York (1961) 21. Briggs, R.O., Reinig, B.A., de Vreede, G.J.: Meeting Satisfaction for TechnologySupported Groups. Small Group Research 37(6), 585–611 (2006) 22. Reinig, B.A.: Towards an Understanding of Satisfaction with the Process and Outcomes of Teamwork. J. Management Information Systems 19(4), 65–83 (2003) 23. Compeau, D.R., Higgins, C.A.: Computer Self-Efficacy: Development of a Measure and Initial Test. MIS Quarterly 19(2), 189–211 (1995) 24. Hsu, M.-H., Chiu, C.-M.: Internet Self-Efficacy and Electronic Service Acceptance. Decision Support Systems 38, 369–381 (2004) 25. Leary, M.R.: Understanding Social Anxiety: Social, Personality, Clinical Perspectives. Sage, Beverly Hills (1983) 26. Stone, B.: Blogging: Genius Strategies for Instant Web Content. New Riders Publishing, Indiana (2003) 27. Callero, P.L.: Role-Identity Salience. Social Psychology Quarterly 48(3), 203–215 (1985) 28. Bhattacherjee, A.: Understanding Information Systems Continuance: An ExpectationConfirmation Model. MIS Quarterly 25(3), 351–370 (2001)

An Exploratory Study of Navigating Wikipedia Semantically: Model and Application I-Chin Wu1, Yi-Sheng Lin1, and Che-Hung Liu2 1 Department of Information Management, Fu-Jen Catholic University No.510, Zhongzheng Rd., Xinzhuang Dist., New Taipei City 24205 Taiwan 2 Department of Business and Management, National University of Tainan 33, Sec. 2, Su-Lin St. Tainan, 700, Taiwan [email protected], [email protected]

Abstract. Due to the popularity of link-based applications like Wikipedia, one of the most important issues in online research is how to alleviate information overload on the World Wide Web (WWW) and facilitate effective information-seeking. To address the problem, we propose a semantically-based navigation application that is based on the theories and techniques of link mining, semantic relatedness analysis and text summarization. Our goal is to develop an application that assists users in efficiently finding the related subtopics for a seed query and then quickly checking the content of articles. We establish a topic network by analyzing the internal links of Wikipedia and applying the Normalized Google Distance algorithm in order to quantify the strength of the semantic relationships between articles via key terms. To help users explore and read topic-related articles, we propose a SNA-based summarization approach to summarize articles. To visualize the topic network more efficiently, we develop a semantically-based WikiMap to help users navigate Wikipedia effectively. Keywords: Navigation, Normalized Google Distance, Semantically-based, SNA-based summary, Wikipedia.

1

Introduction

With the ubiquity of the Internet and Web 2.0 technologies, the WWW has become the main source of information and knowledge in the modern era. The top sites on the Web, as ordered by Alexa traffic rank, are Google, Facebook, YouTube, Yahoo, Live, Baidu, Wikipedia, Blogger, MSN, and Tencent (October, 2010). Wikipedia is the most popular web-based, free-content encyclopedia web site out of the top 10 Web sites. The statistical data from Wikipedia shows that in 2008, the site welcomed 684 million visitors; there were more than 91,000 contributors working on more than 16 million articles. Because of the popularity of applications like Wikipedia, the number of articles in Wikipedia is constantly expanding. As we know, more and more people regard Wikipedia is an efficient means to find needed knowledge, such as searching definitions of terminologies, exploring articles on related topics, and so on. Basically, Wikipedia A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 140–149, 2011. © Springer-Verlag Berlin Heidelberg 2011

An Exploratory Study of Navigating Wikipedia Semantically

141

users browse content in the traditional manner (i.e., by following hyperlinks) when searching for information. However, users may unconsciously change their search goals or get lost when exploring or retrieving information in Wikipedia. In order to make research more effective for the vast number of Wikipedia users, it is highly important to develop effective search or navigation tools to guide users to find and organize needed information or topics. Generally, users invest a great deal of time browsing by following links or searching for specific information. Because of the rapid growth in the volume of information on the WWW, web mining and information retrieval are regarded as key techniques for finding desired information. Web mining tries to extract potentially useful implicit information, link structures and patterns from information units or activities on the WWW. There are three types of web mining techniques: web content mining, web structure mining, and web usage mining. The main difference between web pages and static text documents is that the former contain content as well as link information, and metadata [1][9]. Web content mining exploits IR and artificial intelligence (AI) techniques to extract, mine and analyze information from web pages. Generally, web content mining strategies can be divided into those implicitly that mine information or knowledge from the content of documents, and those designed to improve the search results, i.e., the information retrieved by search engines. IR technology relies primarily on content analysis techniques, but Web pages are usually noisy and contain various types of content, such as text, images, and multimedia. To resolve this problem, some researchers have exploited the hyperlink structure, which provides hyperlink information for a collection of web pages, and proposed ranking algorithms to rank search results. Analyzing the hyperlink structure between WWW pages to support user search activities has attracted a great deal of attention in recent years. Since the link structure encodes a considerable number of latent human judgments, link mining and analysis techniques are employed by commercial search engines, e.g., the PageRank algorithm [2] used by the Google search engine is one of the most well-known link-based algorithms. Almpanidisa and Kotropoulos (2007) [1] proposed a topical information resource discovery algorithm that implements a focused or topic-driven crawler by combining text and link analysis techniques. Their results show that, in the initial stage, the content-based and link-based algorithm does not need a lot of data and that it outperforms comparable methods. In Wikipedia, a topic may contain many articles; thus, it is difficult for users to locate articles relevant to the topic simply by following the hyperlinks in the articles. To address this problem, we propose a semantically-based navigation system that is based on the theories and techniques of link mining, semantic relatedness analysis, social network analysis (SNA) and text summarization. Specifically, we employ a Link Strength (LS) measure to establish a preliminary topic network by analyzing Wikipedia’s internal links [13]. Our goal is to find the specific topic or related subtopics for a seed query (topic) to construct a preliminary internal link-based network. Moreover, we refine the Normalized Google Distance algorithm [3] in order to quantify the strength of the semantic relationships between articles via key terms, and filter out articles that do not have strong semantic relationships. Our preliminary evaluation results demonstrate the effectiveness of applying semantic analysis in an internal link-based network. To help users search for information, we apply centralitybased and cohesive measures in SNA to summarize multiple articles. The measures

142

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

are k-clique and degree centrality, which identify the sub topics of the seed query and key articles of the sub topics respectively. Then, when the user clicks on a topic node that he/she wants to explore, an SNA-based summary is presented on the interface. To more efficiently visualize the semantically-based topic network, an interface as it would appear on PCs and mobile devices is developed to help users navigate Wikipedia effectively.

2 Basic Concepts 2.1 Semantic Relatedness Analysis: Normalized Google Distance Humans acquire the meaning of words and the relationship among words according to their background knowledge. For example, many humans could answer the question of how “polar bears” or “automobiles” are related to “global warming.” However, it is difficult for computers to make judgments regarding the semantic relationships between keywords. How to find the correct tasks to allow computers to automatically extract the meaning of words has recently gained a great deal of attention among the natural language processing and artificial intelligence communities [3] . Basically, there are mainly three measures to estimate the semantic relatedness of different words, which are thesaurus-based, corpus-based and Wikipeida-based measures [5]. Most of the research methods rely on long-term and labor-intensive efforts to construct the semantic relationships among words. Recently, an automatic algorithm has been proposed using the search results counts from the Google search engine, i.e. the normalized Google distance (NGD) algorithm [3]. The NGD algorithm can detect the semantic relationship among terms using the Google search engine. Based on reports from WorldWideWebSize.com, the Google search engine indexes 23.6 billion pages (December, 2010). Thus, the WWW is the largest corpus that can be used to analyze the semantic relationships among words. The main idea of the NGD is to understand the relationships between any two terms according to the number of search results, i.e., the number of return pages. Vitanyi and Cilibrasi (2007) [3] provided a statistical index based on Google page counts, showing the logical distance of a pair of terms called NGD. When the value gets lower, it implies there is a closer relation between two terms. Eq. (1) is the normalized Google distance based on conditional probability. That is, the probability p(x|y) is defined by p(x|y)=p(x,y)/p(y) and . In the NGD equation, y means the probability, as the number of words that Google searched. The result number, by which it is divided, is M (total numbers of document indexed by Google search). p(x, y) is the probability of queries when searching two terms at the same time. Furthermore, because the conditional probabilities are independent of M, we use frequency, i.e., number of search pages, instead of conditional probability to derive the Eq. (2). D( x, y ) =

NGD( x, y ) =

max{log 1

), log 1 )} p( x | y ) p( y | x) 1 1 max{log ), log )} p ( x) p( y)

max{log f ( x), log f ( y )} − log f ( x, y) log M − min{log f ( x), log f ( y)}

(1)

(2)

An Exploratory Study of Navigating Wikipedia Semantically

143

2.2 Automatic Summarization Techniques With the fierce growth of information on the Web, it is important to extract the key abstract content from information sources. Therefore, text summarization techniques and tools are useful to help users find needed information and make decision quickly. Basically, the aim of text summarization is to extract key sentences from one or more documents to represent the meaning of the target document(s). Generally, text summarization can be divided into three phases: analyzing the source text, determining the salient points and synthesizing an appropriate output [7]. For the number of documents, it could be classified as single-document summarization and multi-document summarization. Single-document summarization is generated from the content of one document by different methods that aim to reduce the redundant information. Forsyth and Rada (1996) [6] researched the approach of computing the TF-IDF weight of a document to figure out important terms. Based on their research, they found that sentences are composed from many terms; thus, they attempt to extract important sentences using signature words of the document. Teufel and Moens (1997) [11] employed five heuristic methods: cue, location sentence length, thematic word method, and title methods to extract important sentences from document training sets. They examine the effectiveness of each method and then integrated five methods to analyze the tradeoff between methods. The experiment results show that cue phrases (i.e. in summary, in conclusion, in short, therefore or proper nouns) are the strongest single heuristic and the combination of 5 heuristics will lead to the best performance. Hovy and Lin (1997) [8] mainly considered the important position of sentences (i.e. the first sentence of a document or a phrase), called sentence position, to generate the summarization directly.

3 The System Framework With the emergence of Web 2.0 technologies, social web sites (i.e., social networking websites and micro-blogging services) provide unprecedented opportunities for sharing user-generated content. Wikipedia, one of the most famous collaborative projects on the Web, has become an extremely popular reference database for people seeking information or knowledge. The process for generating a semantically-based topic network is illustrated in Fig. 1. In the following, we describe the modules of the framework. Article pre-processing module. The proposed framework retrieves all link-related articles within three degrees. Besides extracting the hyperlinks from each article, the module stores the content of the article, including the title and its associated links, in XML format. Basically, the Wikipedia is written in wiki markup language which is based on the XML web language composed of wiki-recognized title, content, and most of all containing huge numbers of out-links and in-links. In this phase, we extract and indentify information we want to use from those special symbols of wiki language. For example, the symbols “[” and “]” are used for out-links, while the symbols “[[” and “]]” present the in-links.

144

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

Fig. 1. The process for generating an SNA-based summaries interface

Internal link-based and semantic network analysis module. This module utilizes the proposed link strength (LS) measure to search for topics related to the user’s seed query. It filters out unimportant articles and finds possible subgroups around the seed query. If the value or strength of a link is less than a specified threshold, the system will remove the link from the internal link-based network. Then, the system executes semantic analysis based on the NGD algorithm to filter noise nodes from the initial network. SNA analysis module. This module applies the social network analysis (SNA) indicators defined in social network theory [12] to recognize the role of articles in the topic network based on the results of the previous stage. Basically, we mainly apply centrality and cohesive measures in SNA, i.e., k-cliques and degree centrality, to identify subtopics and key articles of subtopics to help users effectively search for information in Wikipedia. SNA-based summaries generation module. The aim of this step is to generate the summaries of articles based on the SNA analysis results, namely, the interface will present different summarization results based on the topologies of the network. To help users search for information, we apply SNA indicators to summarize single and multiple articles. Then, when the user clicks on a topic node that he/she wants to explore, an SNA-based summary is presented on the interface to help the user quickly read the related articles.

4 Incorporating Semantic Analysis into the Internal Link-Based Network 4.1 Internal Link Analysis by the LS Measure We use the term “article” to denote an entry in Wikipedia rather than a page on the WWW, and the term “node” to denote a word in an article with a hyperlink to another

An Exploratory Study of Navigating Wikipedia Semantically

145

article. The link strength (LS), which indicates the degree of closeness between two articles, is determined by considering the type and frequency of the links between the articles. Our goal is to find the specific topic or related subtopics for a seed query. An article may have three types of links: in-links, out-links, and reciprocal bidirectional or multi-directional links. Two articles linked to each other by the same nodes have a bidirectional link. In addition to the relationships between articles, the frequency of the links (the link frequency) between the nodes is determined by the LS measure, denoted by ζ, which is calculated as follows: ζ(ai , aj) =(f_bi(ai , aj))w1 + w2×f_in(ai , aj)+w3×f_out(ai , aj),

(3)

where f_in(ai , aj) denotes the frequency of in-links from aj to ai; f_out(ai , aj) denotes the frequency of out-links from ai to aj ; and f_bi(ai , aj) denotes the frequency of bidirectional links between ai and aj. Details of setting the relative weights of in-links and out-links and the threshold of the LS value is shown in our recently work [13]. 4.2

Semantic Relatedness Analysis by the NGD Algorithm

For filtering articles with low semantic relatedness in the initial internal link-based network, we conduct further tests for semantic relatedness via the key terms of each article. As we know, it is a difficult task to make judgments of the semantic relationships between keywords via computers. Recently, an automatic algorithm has been proposed using the number of search results returned by the Google search engine, i.e. normalized Google distance (NGD) algorithm, as we introduced in Section 2.1. In our research, we defined the titles of the articles in Wikipedia as the nodes in the NGD algorithm. We can find out the relevancy strength of two web pages by calculating and analyzing the titles of these two pages. A value will be given to the distance of the relationship using the NGD algorithm. Notably, the lower the value is, the higher the semantic relationship is between articles. We filter topics that have a distance value higher than the threshold aim in order to remove noise. The formula of the algorithm is given in Equation (2). Table 1. NGD threshold value of the seed query of “History of personal computers” Title of the Article 74181(ALU) BASIC Bluetooth Burroughs_Corporation Laser_diode Left-handedness Pixar PlayStation_2 The_Walt_Disney_Company Video_game Wii_Remote

Judgment(Irrelevant)

NGD Value

2 2 2 3 2 3 2 3 3 3 5

0.464 0.876 0.971 0.444 1.000 1.000 0.896 0.736 0.421 0.836 0.421

Average: 0.695

146

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

Testing of NGD threshold. Based on Evangelista and Kjos-Hanssen’s research (2006) [4], the expected value of the NGD threshold will be around 0.7. Accordingly, we invited experts from the information management department to determine if the nodes in the topic network are relevant to the seed query of “History of the personal computer”. Table 1 shows the results of the unrelated nodes’ titles with the seed query, derived using the judgment of the experts. The results shows that the average NGD value of all those unrelated nodes is 0.695, which is similar to the execrated results from previous research, i.e., the NGD value is 0.7 [4]. In our the other test, the average NGD value of all those unrelated nodes for the seed query of “knowledge management” was 0.732. Thus, we set the value of the threshold as 0.7 to construct the final topic network toward the seed query. Similarity Calculation. Notably, we also conduct a similarity calculation using the cosine measure to investigate the cohesiveness of each network. Table 2 shows that after applying NGD filtering, the networks will have a higher cosine value than those for which semantic analysis has not been conducted. Table 2. Similarity comparison between IIN and IIN with NGD Queries History of PCs Star Trek Abraham Lincoln Knowledge Management

Internal-link based network (IIN) 0.248 0.149 0.103 0.261

IIN With NGD Analysis 0.240 0.275 0.148 0.291

5 SNA-Based Summaries for the Topic Network 5.1 Process for Generating Summarization Summaries can be divided into three types based on purpose: indicative summaries, informative summaries and critical summaries. In our research context, we generate informative summaries to help users to explore the topics in Wikipedia, with the aid of shorter versions of the original articles. The information summary provides a highly reduced list of the important content of the document, which users could even use to replace the original document. The 30% of a text that is important always represents 80% to 90% of the article’s original focus. Fig. 2 shows the two phases needed to generate the SNA-based summaries, i.e., the analysis phase and the synthesis phase, in our research context. The steps for generating summarization proceeded as follows: (1) Pre-process the source articles; (2) parse articles into terms; (3) select the terms based on the feature set; (4) weight the terms based the role of the articles in the SNA; (5) Calculate sentence scores based on the feature set; and (6) generate summaries from top-N sentences and then resort them based on the sentences’ original order in the articles. We consider the roles of articles in the social network. Accordingly, we then apply different summarization strategies based on the role of articles in the social network. First, the cohesive article (CA) will be identified based on the formula of k-cliques given in the social network analysis (SNA). The analysis results help us label the sub-

An Exploratory Study of Navigating Wikipedia Semantically

147

topics of the nodes in the network. Second, we will find the hub article (HA) of each sub-topic. The HA is analyzed based on the formula of degree centrality given in SNA. This seeks to help the user explore the main topic of each sub-topic. Similarly, we also present a guided summarization of the hub article and its associated articles.

Fig. 2. Process for generating SNA-based summaries

5.2 Identify Sub-topics and Extract the Summaries for the Topics Radev et al. (2004) [10] presented the MEDA algorithm to generate the summaries of multiple multi-documents. The algorithm adopted three features to select the top sentences which are centroid value, positional value, and first-sentence overlap. Similar to previous researchers, we consider the characteristics of the articles in Wikipedia in order to extract key terms and then select the top sentences for extracting summaries of the topic or sub topic. In this work, we proposed three features, which are first sentence, weighted first-sentence overlap and cue phrases to select key terms. Particularly, we incorporated the concept of the position value of a sentence into the concept of the first-sentence overlap to generate the score of each sentence, i.e. weighted first-sentence overlap, wfi, as shown in Eq.(4). Basically, position value, Pi, means the position of the ith sentence in the document, and n is the number of sentences of the target article. Furthermore, the candidate terms can be selected from these sentences, i.e., sentences with high value of wfi., as shown in Eq. (5). The final key terms are selected based on different relative importance of the first sentence Eq.(6).

S1 ,

weighted first-sentence overlap ak , and cue phrases,

CPk

,as shown in

wf i = Sim( f1 , f i ) × Pi where Pi =

n − i +1 n

(4)

n

a k = ∑ wf si × S i

(5)

a'k = λ × top( ak ) + (1 − λ ) × ( S1 + CPk )

(6)

i=2

148

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

6 The Illustrative Example and Application The following table shows the results of the article summarization for the seed query “knowledge management.” We adjusted the relative importance of the weights for the weighted first-sentence overlap, first sentence, and cue phrases based on the Eq. (6). We asked two experts to select the sentences which are highly relevant and useful to assist novices understand the subject of “knowledge management.” There are 20 distinct sentences are selected from the article for we set the value of λ at 0.1, 0.5, and 0.9 respectively, based the Eq. (6). Both of experts agree that the meaningful and useful sentences will be extracted when we set λ at 0.5. Table 3 shows the results of the top-5 sentences selected from the article for we set the value of λ at 0.5. The sentences with bold type denote the overlapping sentences among three different settings of weights. Finally, we present the interface for the semantically-based WikiMap as it would appear on a mobile device and personal computer, as shown in Figure 3(a) and 3(b) respectively. In the future, the effectiveness of the proposed applications will be evaluated in the real setting. Table 3. Article summarization for the seed query “knowledge management” The weight of weighted first-sentence overlap is set to 0.5 1. Knowledge Management efforts typically focus on organizational objectives such as improved performance, competitive advantage, innovation, the sharing of lessons learned, integration and continuous improvement of the organization. 2. Knowledge Management (KM) comprises a range of strategies and practices used in an organization to identify, create, represent, distribute, and enable adoption of insights and experiences. 3. More recently, other fields have started contributing to KM research; these include information and media, computer science, public health, and public policy. 4. Many large companies and non-profit organizations have resources dedicated to internal KM efforts, often as a part of their business strategy, information technology, or human resource management departments (Addicott, McGivern & Ferlie 2006). 5. KM efforts overlap with organizational learning, and may be distinguished from that by a greater focus on the management of knowledge as a strategic asset and a focus on encouraging the sharing of knowledge. Note: Available at: http://en.wikipedia.org/wiki/Knowledge_management (accessed 30 January 2011).

Fig. 3(a). Semantically-based WikiMap Fig. 3(b). Semantically-based Presented in the Mobile Device Presented in the PC

WikiMap

An Exploratory Study of Navigating Wikipedia Semantically

149

7 Conclusion and the Future Work In this present study, we proposed an SNA-based summarization model and the development of search or navigation applications to guide users to find and organize needed information or topics within Wikipedia. We employed the NGD algorithm in the proposed LS measure to quantify the strength of the semantic relationships between articles in the topic network. Our preliminary evaluation results demonstrate the effectiveness of applying semantic analysis in an internal link-based network. To help users quickly read topically related articles, we proposed SNA-based summarization to present single article’s summaries in a newly developed interface. This study also presents the interface for the semantically-based WikiMap as it would appear on PCs and mobile devices. In the future, we will evaluate the precision and accuracy of the summaries for the articles based on the proposed methods. Moreover, the effectiveness of the proposed applications will be evaluated in the real setting. Acknowledgments. This research was supported by the National Science Council and Fu-Jen Catholic University of Taiwan under the Grant No. 99-2410-H-030-047-MY3 & No.409931074078, respectively.

References 1. Almpanidisa, G., Kotropoulos, C., Pitas, I.: Combining Text and Link Analysis for Focused Crawling—An Application for Vertical Search Engines. Information Systems 32(6), 886– 908 (2007) 2. Brin, S., Page, L.: The Anatomy of a Large-scale Hypertextual Web Search Engine. Computer Network 30(1-7), 107–117 (1998) 3. Cilibrasi, R., Vitányi, P.: The Google Similarity Distance. In: IEEE Transactions on Knowledge and Data Engineering, pp. 370–383. IEEE Press, New York (2007) 4. Evangelista, A., Kjos-Hanssen, B.: Google Distance between Words. Frontiers in Undergraduate Research. University of Connecticut (2006) 5. Finkelstein, L., Gabrilovich, Y.M., Rivlin, E., Solan, Z., Wolfman, G., Ruppin, E.: Placing search in context: the concept revisited. ACM Transactions on Information Systems (TOIS) 20(1), 406–414 (2001) 6. Forsyth, R., Rada, R.: Adding an Edge in Machine Learning: Applications in Expert Systems and Information Retrieval. Ellis Horwood Ltd., pp.198-212 (1986) 7. Hahn, U., Mani, I.: The Challenges of Automatic Summarization. Journal of IEEE Computer 33(11), 29–36 (2000) 8. Hovy, E., Lin, C.Y.: Identifying Topic by Position. In: Proceedings of the 5th Conference on Applied Natural Language Processing (ANLP), Washington, DC, pp. 283–290 (1997) 9. Liu (ed.): Web Data Mining: Exploring Hyperlinks, Contents and Usage Data. Springer, New York (2007) 10. Radev, D.R., Jing, H., Stys, M., Tam, D.: Centroid-based Summarization of Multiple Documents. Information Processing and Management 40(6), 919–938 (2004) 11. Teufel, S., Moens, M.: Sentence extraction as a classification task. In: Proceedings of the Workshop on Intelligent Scalable Summarization. ACL/EACL Conference, Madrid, Spain, pp. 58–65 (1999) 12. Wasserman, S., Faust, K. (eds.): Social Network Analysis: Methods and Applications. Cambridge University Press, UK (1994) 13. Wu, I.-C., Wu, C.-Y.: Using Internal Link and Social Network Analysis to Support Searches in Wikipedia: A Model and Its Evaluation. Journal of Information Science 37(2), 189–207 (2011)

Productivity Improvement by Using Social-Annotations about Design Intent in CAD Modelling Process Gerardo Alducin-Quintero1, Manuel Contero1, Jorge Martín-Gutiérrez2 David A. Guerra-Zubiaga3, and Michael D. Johnson4 1 2

13BH, Universitat Politècnica de València, Camino de Vera s/n, 46022 Valencia, Spain Universidad de La Laguna, Avda. Astrofísico Fco. Sánchez s/n, 38206 La Laguna, Spain 3 CIDESI Texas, Caldwell Waston Bldg., 1700 Research Pway, College Station, Texas, USA 4 Dept. of Engineering Technology and Industrial Distribution. Texas A&M University, 3367 TAMU, College Station, Texas, USA [email protected], [email protected], [email protected], [email protected], [email protected]

Abstract. This paper focuses on the New Product Development Process (NPDP) area to contribute to increasing the productivity of CAD users by means of an improved design intent communication using a social-annotation technique. Design teams operate in a similar way to an online social network, and CAD models are not just only a 3D geometry representation, as they reflect the result of an specific modeling strategy, that usually constrains the future capability for modification and reuse of the existing 3D model. Considering this context, this work tries to assess the impact of annotations in the engineering change process in order to determine its influence in the user performance during this process. Preliminary experimental results obtained from several experiments with Spanish CAD students indicate that it is possible to reduce the time needed to perform engineering changes in existing models into a 10-20% range, if those CAD models provide annotations explaining the original design intent. Keywords: New Product Development Process, CAD, Social-Annotations, Design Intent, Engineering Change Process.

1 Introduction The New Product Development (NPD) process is a key activity for the enterprise survival and competitiveness. Empowering the role of design and shortening the development cycle of new products are two well-known strategies to improve this process. In this context, advanced CAD systems are an important tool to support the NPD process, providing a key advantage to improve organizational effectiveness [1]. This paper focuses on this area in order to contribute to increasing the productivity of CAD users by enhancing design intent communication by means of a socialannotation technique, inspired on the idea that design teams operate as an online social network. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 153–161, 2011. © Springer-Verlag Berlin Heidelberg 2011

154

G. Alducin-Quintero et al.

Design intent governs the relationships between modeling features in a part, and between parts in assemblies. It is related to the behavior of the CAD system when a modification is performed on a CAD model. CAD users should analyze the geometry of the part to be generated, and plan what is the most efficient sequence of modeling features, in order to assure that future modifications in the design are managed by the CAD system in a determined expected way. Usually good CAD models are associated to change flexibility and robust variation capabilities [2] [3] [4]. Two kinds of knowledge are related to the design intent concept: declarative knowledge that is knowledge of facts (knowing that or knowing what) and procedural knowledge that is knowledge of how to do things (knowing how) [5]. This means that a CAD model is not just only a 3D geometry representation, but also this model stores the know-how about the strategy used to build it. Capture, share and transfer this complex knowledge is a key element to improve the NPD process. Through this work is attempted to demonstrate the importance of going beyond the capture of design intent in a traditional way, based on non-explicit information associated to the feature types used in the CAD model and their interrelationship. This is aligned to Guerra-Zubiaga [6] work trying to structure different knowledge types to support manufacturing and design decisions. This work proposal includes an innovative solution by using design annotations relative to the CAD modeling strategy, specifically, about the decision making process during the 3D model construction. Design annotations give an important clue about how and why a CAD feature was used to build a CAD model by using just a few words. This research activity is directed to assess the impact of these annotations in the engineering change process and in second place in analyzing the dynamics of the social annotation process. This paper is centered in the first aspect, presenting a pilot study that has been conducted using CAD models related to the automotive industry. These CAD models represent parts with different geometry complexity levels. For each part, two different CAD models were created, with and without annotations. These models were modified by two classes of engineering students (one mechanical engineering group at the undergraduate level, and other belonging to a master degree on CAD/CAM/CIM). Both classes were given the same written instructions where it was explained a series of engineering changes to be performed by modifying the geometry of the CAD models. For each class a control and an experimental group were randomly organized. Experimental groups had to change the models that contained annotations about the design strategy used by the original author of the model, while the control groups had not available this additional information. Preliminary results indicate that it is possible to reduce the time needed to perform engineering changes in existing models into a 10-20% range by using the annotation technique. It was determined the basic knowledge-mapping and examined the problem-solving process employed by the participants in the modification of constraint-based CAD models [7]. The initial hypothesis was confirmed: the experimental group performed the design changes using less time. Additionally it was noticed that certain patterns of behavior of participants coincided with previous research works like: [1] [3] [8]. In the next point, is provided all the details about the experimental design followed to analyze the influence of annotations on the modification of CAD models. Then

Productivity Improvement by Using Social-Annotations

155

results are presented, explaining the limitations and main conclusions of this preliminary study and giving some orientation about future works about this topic.

2 Hypothesis and Methods 2.1 Hypothesis Definition This piece of research work pursues to explore how design annotations influence in the user performance during the engineering change process. This is performed before analyzing the dynamics of the social annotation process, because it is needed to assure that the availability of these annotations provides an added value to the CAD models. This research hypothesis that CAD operators, using annotated models where original design intent is made explicit, are more efficient dealing with CAD model modifications. In this context, efficiency is related to the time used by CAD users when they have to perform a change in a 3D geometric model to accomplish an engineering change order. The author’s perspective is: if the design intent knowledge is made explicit by means of these annotations, the corresponding CAD models will be created with better semantic quality, following the product data quality model by Contero et al. [2]. Author´s vision is that this annotation process can be performed following the behavior of social networks, where knowledge associated to the CAD modeling process is made explicit by the collaborative annotations performed by design engineers. 2.2 Experimental Stage The experimental stage was divided in two phases in order to test the hypothesis: phase I, called as “undergraduate study”, and phase II, named as “postgraduate study”. All the participants were given the same written explanation and modeling tasks with accompanying figures, and they were asked to perform a series of engineering changes that consisted in geometric modifications on the original 3D CAD models. The parts were created thinking that the references used for the creation of each feature impacted the ability for later modification and edition of the geometry. This includes capturing design intent during geometry creation. Specifically, in the compound parts (e.g. automotive components) some references (e.g. datum planes or sketches) were used to separate in groups of CAD features that define each area of the part. The used models are divided in two sets: simple geometry (low amount of CAD features used to build it) and compound geometry (composed by a high amount of CAD features). As simple geometry parts were used a housing of gear box and the part used by Johnson in [9]. As complex geometries were used a component of a PC fan housing base (from Brigham Young University’s NX CAD Manual) and an automotive radiator collector (from a Spanish company). In both cases, the same type of geometry modifications was required: simple changes (e.g. modify the height of one element that is controlled by an expression) and complex changes (e.g. create a copy of a group of features with strong parametric parent-child relationships). Two variants for each CAD model were created: one without any annotation information

156

G. Alducin-Quintero et al.

explaining design intent (named regular CAD model) and other one with explicit textual annotations about the design intent (named annotated CAD model). The content of texts in the annotated CAD models provided information about: • Location. This kind of text comment allows identifying the set of CAD features that define a geometric element of the part (indicating the first and last feature of a specific element). • Design Intent. This commentary allows understanding the intention of the author of the model (e.g. this profile was used to create the main body of the connection). • Modification Procedures. This comment gives information about both recommended steps for making modifications and procedures that must be avoided (e.g. recommendation, don’t use mirror copy feature). 2.3 Undergraduate Experience Context. Firstly, the undergraduate students group (composed by thirty students and denoted as the “undergraduate group”) was split into two groups of fifteen participants. Students were enrolled in a CAD course from a Mechanical Engineering degree in La Laguna University (Spain). The course was based on Autodesk Inventor. The difference between the experimental group and control group was in the CAD model files that they received in order to conduct the study. The experimental group received CAD files with annotations, by using the Inventor Engineer Notebook functionality, that allows to create CAD annotations that support both text and images. The control group received the same CAD files, but stripped of any annotation information. The observed variable was the time expended in performing each modification requested to the participants. Students were given 50 minutes as maximum time to complete all the required modifications. They controlled the time, writing down the initial and final time for each requested geometric modification. Afterwards this time list was rechecked using the time stored into the participant’s CAD files in order to avoid inconsistencies. Availability of numerous groups of CAD users with a similar knowledge and expertise level in modelling is a very limited, outside the academic world. CAD students provide an interesting study population, whose behaviors can be extrapolated to the professional and industrial world. With respect to the sample size, Polkinghorne [10] and Meyer & Booker [11] recommend a number between five and twenty designers for an exploratory phenomenological study, which is verified by the present experience. Participant students showed homogenous basic skills as Inventor users, and completed the modeling exercises as requested. Knowledge Mapping Tasks. In both groups, participants showed a similar amount of declarative knowledge (knowledge about CAD commands) but they had a lack of procedural knowledge (knowledge of how to apply the commands to achieve a goal) due to their short experience. The high number of create-erase contiguous events was an interesting behavior pattern exhibited by the undergraduate group. This is related to the fact that they only knew the most basic commands of the CAD system, so they didn’t waste time trying to use complex commands or trying to find them in the CAD user interface. The majority of the participants that completed the exercises used the most simple and direct solution.

Productivity Improvement by Using Social-Annotations

157

2.4 Postgraduate Experience Context. The postgraduate students group (composed by eighteen students and denoted as the “postgraduate group”) was divided into two groups of nine participants. Students were enrolled in an advanced CAD course from the CAD/CAM/CIM Master degree at Universitat Politècnica de València (Spain). The course was based on Siemens NX CAD system. The experimental group received CAD files with annotations. The control group received the same CAD files, but stripped of any annotation information. The observed variable was the time expended in performing each modification requested to the participants. Students were given 60 minutes as maximum time to complete all the required modifications. They controlled the time, writing down the initial and final time for each requested geometric modification. Afterwards this time list was rechecked using the time stored into the participant’s CAD files in the same way that during the undergraduate experience. At the beginning of this research work, was supposed that the impact of the annotations depends on several factors. One of these factors is the user knowledge and expertise. Taking into account that postgraduate students have a more extensive experience with real design problems, exercises used in this experience, have been taken from real industrial CAD models. That’s the reason that in this second experimental phase was tried to organize a postgraduate students group with real design experience in the industry. This means, that they knew several CAD systems, with a medium level of expertise, being one of these systems Siemens NX. However, this group was observed to have a less homogenous behavior that the undergraduate group, due to the different background of participants. Knowledge Mapping Tasks. It was detected that the participants’ expertise influences their performance during the decision-making process to complete an engineering change request. For example if they must create a copy of a geometric element and they know that in other CAD software there is a command to perform a mirror-copy operation, they invest time to find this CAD function, although perhaps it does not exist in Siemens NX or it is not implemented in the same way than in other CAD system. This behavior is related to the fact that the annotations are easy to visualize and their content is relevant to perform the design change in a specific way. For example, if the annotations had contained direct recommendation to perform the specific design change, this expert reaction could have been avoided.

Fig. 1. Simple Engineering Change Request

158

G. Alducin-Quintero et al.

Fig. 2. Complex Engineering Change Request

3 Results 3.1 General Results and Statistical Analysis In the first experimental stage (undergraduate group) only fourteen students completed the engineering changes requested in exercises one and two, as it is seen in Table 1. The average time for those completing the exercise one was 37 minutes. For this exercise the experimental group (with annotations) there were six students and their average time was 28 minutes, while the control group (without annotations) there were eight students and their average time was 37 minutes. The average time for completing engineering changes in the exercise two was eight minutes. In this case the experimental group had an average of five minutes and the average time of the control group was eight minutes. In both cases the control group had an average completion time that was between 26-37% greater than the experimental group. During the second experimental stage (postgraduate group) the results were: for the first exercise only 16 students completed the engineering change requested with a average time of 49 minutes. Specifically the experimental group (eight students) had an average time of 43 minutes and the control group (eight students) had 52 minutes. In the case of the second exercise only 14 students completed the geometric changes requested with an average time of 50 minutes. The experimental group (seven students) had an average time of 47 minutes and for the control group (seven students) was 54 minutes. In both cases the control group had an average completion time that was between 13-17% greater than the experimental group. Table 1. Experimental Phase I: Undergraduate Group Group Total Students Participating Exercise One Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation Exercise Two Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation

Control 15

Experimental 15

8 37 7.305

6 28 8.262

8 8 3.583

10 5 2.846

Productivity Improvement by Using Social-Annotations

159

Table 2. Experimental Phase II: Postgraduate Group Group Total Students Participating Exercise One Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation Exercise Two Number Completing Exercise Average Time for Exercise Four (min.) Standard Deviation

Control 10

Experimental 9

8 52 21.715

8 43 20.584

7 54 17.321

7 47 17.153

Table 3. Statistical t Analysis for Hypothesis Exercise / Phase E1/Ph I E2/Ph I E1/Ph II E2/Ph II

Time (minutes) Control Experimental 37 28 8 5 54 43 54 47

T -2.243 -2.165 -0.993 -0.729

Significance 0.045 0.046 0.338 0.480

While the results in the first phase (undergraduate group) showed that there is difference between control and experimental groups, the results were further analyzed to detect any statistically significant differences. The results of those statistical tests are shown in the table 3. This table shows the relevant statistic and one-tailed probability. Although in all the cases, experimental group (who’s received additional information through annotations) showed a better performance, this means they used less time than the control group. But only there is a statistically significant difference with the undergraduate group experience. The authors relied on that is because the undergraduate group has more homogenous understanding about design and CAD software than other groups [12].

4 Limitations and Future Work One of the most important limitations of this research work relates to the sample size of the CAD users that participated in the experiences. The sample size was small as this pilot study, being exploratory in nature, was intended to give us suggestions for more extensive studies in the future, where larger number of participants would allow to perform a more robust statistical analysis. Other factor that can be considered a limitation is the users’ expertise. The undergraduate group students showed low level CAD skills (they were enrolled in a basic level CAD course) but a homogenous background. This was a positive factor to obtain a result concordant with the initial hypothesis. The postgraduate group of students had a different profile. The majority of them had a real industry experience in design, showing a heterogeneous professional background with different CAD skills. This heterogeneity could have let to the lack of more conclusive results.

160

G. Alducin-Quintero et al.

Fig. 3. Impact of Design Intent Annotations

Future work will be focused on mitigating some of these limitations mentioned above and improving the analysis of the results. The first step would be to find a larger group of students for participating in the experiments. They should offer a similar experience and expertise. Second step would be the redesign of experiments applying all the experience obtained during the previous study.

5 Conclusions Product design can be considered a social process involving groups of individuals (i.e., design teams) attempting to meet complex product needs through sharing varied expertise and knowledge types [13]. It is a creative process involving networks of people with negotiations and interactions between them; the product design itself is dynamic and evolving [14, 15]. This work explored the effects that design intent annotations had during the engineering change process and assessed the possible relationship between annotations’ content and the designer’s knowledge related to an expertise level. The findings presented in this paper show that the productivity impact through CAD annotations depends of several factors such as geometry among others. It can be shown that if a simple design change is performed on a simple geometry the annotation impact is low or null. However, if a complex change is required on complex geometry the annotation impact is significant. In other words, the annotation impact depends on how it can be structures according to the annotation’s content and the user’s knowledge as shown in figure 3.

Productivity Improvement by Using Social-Annotations

161

It has been shown the need of explore new knowledge infrastructure to support better design decisions, not depending on a particular computer design platform through product design development. Acknowledgement. The authors acknowledge the support received by UPV, CIDESITEXAS, Texas A&M University, and Universidad de la Laguna. The authors also acknowledge the support by Tecnológico de Monterrey through the Research Chair in Nanotechnology and the valuable support of Oscar Martínez and Germán Manacera on the experimental performance of this piece of work.

References 1. Anderl, R., Mendgen, R.: Parametric design and its impact on solid modeling applications. In: Third ACM Symposium on Solid Modelling and Applications, pp. 1–12. ACM, New York (1995) 2. Vila, C., Contero, M., Company, P.: Extended modeling, a tool for cooperative design. In: 6th International Conference on Concurrent Enterprising, Tolouse France (2000) 3. Anderl, R., Mendgen, R.: Analyzing and optimizing constraint-structures in complex parametric CAD models. In: Bruderlin, B., Roller, D. (eds.) Geometric Constraint Solving and Applications, pp. 58–81. Springer, Berlin (1998) 4. Hartman, N.W.: The development of expertise in the use of constraint-based CAD tools. Eng. Design Graph. J. 68, 14–26 (2004) 5. Rynne, A.: AC 2007-2132: Cognitive modelling strategies for optimum design intent in parametric modelling (PM). American Society for Engineering Education (2007) 6. Guerra-Zubiaga, D.A.: A Model to enable Knowledge Maintenance: Supporting Manufacturing Decisions. VDM Verlag, Germany (2009); ISBN 978-3639209983 7. Hartman, N.W.: Defining expertise in the use of constraint-based CAD tools by examining practicing professionals. Eng. Design Graph. J. 68, 6–15 (2005) 8. Bhavnani, S.K., Garrett, J.H.J., Shaw, D.S.: Leading indicators of CAD experience: Paper presented at the Proceedings of Computer-Aided Architectural Design Futures CAAD Futures 1993, pp. 313–334 (1993) 9. Johnson, M.D., Diwakaran, R.P.: Assessing the effect of incentive on computer-aided design intent. In: The ASME 2009 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, San Diego, California, pp. 523–532 (2009) 10. Polkinghorne, D.E.: Phenomenological research methods. In: Valle, R.S., Halling, S. (eds.) Existential-Phenomenological Perspectives in Psychology, pp. 41–60. Plenum, New York (1989) 11. Meyer, M.A., Booker, J.M.: Eliciting and analyzing expert judgment: A practical guide. Academic Press, San Diego (1991) 12. Reddy, J.M., Finger, S., Konda, S., Subrahmanian, E.: Design as Building and Reusing Artifact Theories: Understanding and Supporting Growth of Design Knowledge. In: The Design Productivity Debate. Springer, Heidelberg (1998) 13. Bucciarelli, L.L.: Designing engineers. MIT Press, Cambridge (1994) 14. May, P., Ehrlich, H.C., Steinke, T.: Mapping Sociotechnical Networks in the Making. In: Bowker, G., Star, S.L., Turner, W., Gasser, L. (eds.) Beyond the Great Divide,Technical Systems and Co-operative Work, Lawrence Erlbaum, Mahwah (1997) 15. Bhavnani, S.K., Garrett, J.H.: Leading Indicators of CAD Experience. In: Flemming, U., Van Wyk, S. (eds.) CAAD Futures 1993, pp. 313–334. Elsevier Science Publishers, Netherlands (1993)

Handshake: A Case Study for Exploring Business Networking for the Enterprise, Inside and Out Laurie E. Damianos, Donna L. Cuomo, and Stan Drozdetski The MITRE Corporation, 202 Burlington Road, Bedford, MA 01730, USA {laurie,dcuomo,drozdetski}@mitre.org

Abstract. MITRE has developed and launched Handshake, a social networking platform for business networking use. We took a unique approach in providing a single MITRE-hosted platform for connecting, collaborating, and networking with internal employees and external partners. The business networking prototype also serves as a research platform for building and deploying other social capabilities and exploring the value of social media for the enterprise. Since its initial release, Handshake has grown to support over 4300 users (~1000 are external participants) and 450 groups. Feedback from our early adopters has been very positive, with compelling stories on how they are able to engage our sponsors, mission users, and other partners in ways not possible with previous tools. We discuss the initial results of our study here and provide examples of how Handshake is able to meet our corporate objective of bringing the larger community to bear on problems of national importance. Keywords: social networking, business networking, social software, social computing, web 2.0, social media, collaboration, collaborative environments, collaborative computing, computer-supported cooperative work, evaluation, business value, web-based interaction, group and organization interfaces, Elgg.

1 Introduction Social networking tools (e.g., Twitter, Linked-In, Facebook, Ning, MySpace) have been proliferating on the internet. We describe a social networking tool as a webbased service which allows users to construct a profile, explicitly declare a connection with other users, view and browse their connections and those of others [1]. Many of these tools can also be used for personal publications, status updates, broadcasting, and direct messaging. Because of the social nature of the tools, they can provide situation awareness of one’s personal network as long as people frequent these sites and make contributions. Statistics indicate that many people visit these sites at least monthly, with frequent users visiting several times a day. According to [2], there are almost 600 million Facebook users worldwide. Most social networking sites primarily support pre-existing social relations [3]; in fact, Facebook users search for people they already know more than they browse for people to meet [1], [3]. While humans can recognize up to 2000 connections [5], they can maintain only 150 [5], [6]. Indeed, the average number of Facebook friends is 130 A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 162–171, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

163

[7], [8]. On average, people have 10-15 strong ties [5] (e.g., close confidantes, people who can be relied upon in an emergency, and people with multiple shared interests [9]), and those networks tend to reinforce beliefs rather than introduce new ideas. Weak tie networks provide access to a variety of ideas and experiences and expand the scale and the scope of a given network [9]. Both strong ties and weak ties can be sustained on social networking tools [3], [10], and latent ties (not socially activated) can be converted to weak ties on these sites [3], [11]. Individuals can use their social capital to gain access to resources from other members of their network [3]. Having connections to individuals outside one’s immediate network can introduce non-redundant information, new perspectives, and other benefits (e.g., employment connections), and therefore weak ties help build social capital [12], [13], [14], [15], [16]. In fact, a MITRE-Babson study on the impacts of technology-mediated networks showed a positive relationship between the number of diverse connections in someone’s network and their innovative behavior [17]. Indeed, social software has become “crucial in supporting and strengthening collaboration and nonroutine work in businesses” [18]. As early as November 2006, approximately 22,000 organizations had Facebook directories [19]. Gartner explains that “The demand from workers is escalating, and they can and are turning to the consumer Internet if their corporate technology provider isn’t offering a solution” [18]. To meet this demand, some organizations have started deploying social networking capabilities, but many are still in the proof-of-concept or early stages [20]. Those sites that have been instantiated are primarily for internal enterprise use only [21], [22], and the Burton Group reports that these organizations are faced with many non-technical issues such as “business case, metrics, policies and controls, roles and responsibilities, employee participation models, and cultural dynamics” [20]. In this paper, we describe our own unique approach of designing and deploying a single social networking platform for the enterprise that supports both internal and corss-organizational business networking.

2 Case Study The MITRE Corporation [23] is a not-for-profit organization with expertise in systems engineering, information technology, operational concepts, and enterprise modernization. In addition to managing several Federally Funded Research and Development Centers (FFRDCs), MITRE supports its own independent technology research and application development for solving sponsors’ near-term and future problems. MITRE has approximately 7000 scientists, engineers, and support specialists distributed across many locations and working on hundreds of different projects for various sponsors in numerous domains. MITRE employees have traditionally relied on email, telephone, and face-to-face meetings to communicate and collaborate with external partners. Artifacts distributed via email get lost or conflicts occur when different copies of materials are edited. Telephone conversations are ephemeral, and face-to-face meetings involve travel and scheduling issues. The company also uses an external Microsoft SharePoint site as a secure document repository, but the strict authentication requirements, the lack of

164

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

lightweight functionality, and the limited social aspects of the tool make it difficult for users to connect socially and collaborate. Forming ad hoc group workspaces around topical areas is also difficult with existing tools. In addition to these problems, individuals want to be able to establish relationships with external parties and manage those contacts as people move from job to job. The recent popularity of tools like LinkedIn and Facebook make connecting and tracking possible, but they are not trusted environments for conducting business and are fraught with overwhelming amounts of personal and extraneous information [24]. As a result, it is currently difficult to capture the business networks and relationships that MITRE staff have with people from other organizations. To address these and other collaboration issues, a research team at MITRE embarked upon building a trusted environment for MITRE and its partners to connect, collaborate, and share new information. Our goals were to improve MITRE’s ability to establish and maintain relationships, form groups and facilitate multiorganizational collaboration around topics, leverage expertise from industry, and bring broader segments of the community to bear on important sponsor problems. In addition, we wanted to enhance the individual – and enterprise – situational awareness around relationships, activities, topics, and communities. A specific focus of this project was how to address and mitigate the risks of this new capability. We needed to ensure that the social networking platform – supporting both internal and external access – and its data-level access model were both secure and usable. From a research perspective, we were also interested in exploring and assessing the value of an enterprise networking tool for business use as well as identifying and evaluating emerging business models enabled by social networking tools. Could the enterprise move away from supporting teams to leveraging networks and make the enterprise, as a whole, more efficient?

3 Approach Before selecting and deploying a social networking tool, we started by gathering requirements and identifying business need use cases. We conducted work practice studies by interviewing and observing 18 MITRE employees who were engaged in cross-organizational projects and teams with sponsors, vendors, contractors, consultants, academia, and other FFRDCs. We documented work practices, assessed needs that were not being met, identified breakdowns and ineffective use of existing technology, and collected additional requirements for our prototype tool. The team also explored existing social networking tools and products – including commercial technology, open source tools, and hosted platforms – to understand what functionality and options existed. We used this data, along with extensive criteria, to select an appropriate tool that would support a broad spectrum of research efforts. Once we had a research platform selected, we built upon it and deployed it outside our firewall. We identified teams and partners to support as we piloted the tool and used their feedback to make improvements. We leveraged these pilots to study risks and costs, deployment mechanics, use and usage issues, integration, and cultural and political dynamics. Throughout this process, we worked with several teams to create a

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

165

framework for evaluating social media and a multi-fold plan for assessing the value of business networking for the enterprise and for the individual.

4 Research Platform 4.1 Tool Selection and Implementation We wanted to create a trusted environment with fine-grained access control and a secure, yet easy, invitation model, all of which had to meet our information and operational security requirements. We required a single platform to allow our employees to engage with each other internally but also to connect and collaborate externally without having to go to another site. The latter proved far more difficult, as no commercial tool supported the single-platform approach to span a firewall. For user adoption, the tool had to be easy to use and require little to no training. Ideally, we wanted it to be similar to popular social tools like Facebook that many of our employees were already familiar with. We were also looking for a low cost solution: an inexpensive tool with no licensing costs to accommodate an unknown and unlimited number of external participants. Finally, in order to support our experiments and allow us to use the tool as a research platform, the tool had to be fully customizable, extensible, measurable, and easy to integrate with other applications. After an extensive market survey, we selected the open source Elgg [25] platform as the basis for our prototype. Elgg’s fine-grained access control model suited our unique requirements for sharing both inside the company and outside, and the ability to host it on our own servers helped meet our information security requirements. Elgg also provided enough out-of-the-box functionality and many of the key features that we wanted. In addition, Elgg was flexible enough to customize to our environment and our needs; its plug-in model allowed for extensibility without modifying the core, and its social graph API made it easy to integrate with other applications and services. An advantage of using open source software for research was that it gave us access to the community of developers; we were able to leverage the contributions of others and share our own plug-ins and enhancements. Through this exchange, we developed a good working relationship with core Elgg developers who were responsive to our needs and concerns. Finally, using an open source platform required no software purchase and no licensing costs, resulting in a solution with no initial investment. Using an agile method of iterative and incremental development, we designed and implemented a working prototype. We focused on creating a trusted environment: we tightened but simplified the permissions model, added new access control levels, encrypted communication via HTTPS, removed public access, and made the invitation mechanism more secure. We also worked on enterprise enhancements: improving the usability and integrating with single sign-on, active directory, email, our intranet portal, and other prototypes. Lastly, we wrote an event listener to log activity for evaluation and research. Our business networking research platform, named Handshake, was first prereleased internally to the corporation and then later made available outside the firewall. Initially, we invited and nurtured several pilot teams to build their communities on Handshake and then used them to drive our early requirements, but

166

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

our user base grew rapidly through word-of-mouth. We encouraged feedback from our users and were able to address concerns quickly. Our agile development methodology allowed us to implement, test, and deploy new features rapidly. 4.2 Handshake Overview and MITRE Customizations Handshake [26] is open to all MITRE employees and invited external participants for connecting, communicating, and collaborating in a trusted environment. Handshake members can form explicit relationships with other Handshake members, create their own profile, and establish groups. MITRE staff must explicitly invite external participants to join Handshake, and these external partners can in turn establish connections with other Handshake members and join groups. All content in Handshake has its own access control settings, so that content owners can regulate access. Awareness of activity on Handshake is provided through activity streams and customizable email or site alerts. The rest of this section describes some of the MITRE-built features and specific customizations for our environment. Access. MITRE employees access Handshake by single sign-on and MITREid (based on OpenID) from inside or outside the firewall. Handshake is closed to the general public, but MITRE employees may invite external partners into Handshake via an authenticated email-based invitation model. External members currently may not invite others into Handshake. Security Features & Permissions. Handshake utilizes a MITRE identity model to distinguish MITRE employees from other Handshake members. MITRE members are labeled as such and visually distinctive. All content in Handshake – groups, profile fields, files, wiki pages, discussion topics, blog posts, bookmarks, and images – has its own access control settings. Content owners may choose to make content accessible to all Handshake members, their connections, members of a specific group, MITRE only, or MITRE members of a specific group. Members may also create their own customized access control lists. To help prevent sharing of information to the wrong audience, all content is clearly marked with an appropriate padlock icon, either open or locked. This icon is colored to indicate the access level. Our model for group permissions restricts content within a specific group to that group’s permission level, ensuring that material is not inadvertently exposed beyond the membership of the group. As an added precaution, MITRE members see a banner as a reminder to share information appropriate to that audience. Profiles. Handshake automatically populates profiles of MITRE employees with their name, email address, and phone number; external participants are required to provide their name and organizational affiliation. All Handshake members may optionally upload a profile picture and add their job title, website, hyperlinked professional and personal interests, education, and a bio. Each profile field is access-controlled so that members can choose what to share with whom. Groups. MITRE employees are able to create their own groups around a project, topic, community of practice or interest, affiliation, hosted conference or recurring meeting, etc. Groups can be open or moderated and can be made private or accessible to larger communities. Elgg’s out-of-the-box group capability includes tools such as

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

167

discussion forums, file repositories, wiki-like pages, blogs, photo albums, and a message board. We added our own generic HTML widget so that users can post announcements, provide links to resources and contacts, or create hyperlinked lists. We have also added a group activity feed, a group tag cloud, and a group wiki tag cloud. A simple customization feature allows owners to disable group tools that are not needed. To allow for more flexibility in group management, we have implemented a multiple co-owner feature, the ability to invite members into Handshake and a specific group simultaneously, sub-group functionality, and a group restore feature (for accidental group deletion). Browsing & Searching. MITRE developers improved the metadata and tag search capability across all content, groups, and users. This enhanced search was incorporated and released to the community in Elgg version 1.7. An autocomplete feature in the search box saves users time by allowing them to select from their established connections or group memberships. Further enhancements allow for browsing of groups by membership level, recent activity, recent visits, alphabet, and creation date. We also dynamically sort users’ groups so that they can quickly access those they visit most often. Finally, we have implemented tag clouds for personal content, group content, and group wiki pages to assist in browsing and navigation. Awareness. Handshake provides widgets on each user’s home page that display recent activity – status updates, file uploads, blog postings, comments on content, new group creation, etc. Members can choose to view their own activity feed, the activity of their connections, activity of all Handshake members, and activity across their groups. Each Handshake group also has its own activity feed to keep members up to date and bring new members up to speed quickly (this MITRE plug-in has been released back to the Elgg open source community). Users can view items in the activity feed only if they are permitted access to the actual content. By default, Handshake users also receive customized email alerts to activity that occurs within their groups. The alerts contain textual content and hyperlinks to both the user who performed the action and to the group where the activity occurred. Users may elect to receive their alerts as a single daily digest, turn off notifications from specific groups, or not receive alerts at all. The email alerts provide users with lightweight access to activity occurring in Handshake. In the near future, users will be able to respond to discussion topics directly from their email. Integration & App Development. In addition to our current work of integrating Handshake with email, we have been coordinating with other teams at MITRE who are building plug-ins, applications, and services for Handshake. Recently, work has been done to integrate status updates and activity feeds with our enterprise intranet. Other teams at MITRE are working on mobile applications to Handshake, a contact recommender system for automatically suggesting connections based on users’ digital footprints [27], and an “introduction” tool allowing users to introduce people to each other. Another research group is investigating automatic profile generation. Future Customizations. We would like to improve the user experience within Handshake by enhancing navigation, browsing, searching and filtering for users, groups, and content. We would also like to implement the ability for group owners to customize groups and create their own widgets. In order to entice people to visit

168

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

Handshake frequently, we are building a “Pulse of Handshake” feature where visitors can view dynamically updated content and learn what is happening, what is new, and what is getting a lot of attention. We are designing an extended permissions model around the concept of affiliated partner organizations and trusted partners; external participants belonging to an organization with a contractual relationship with MITRE will have special rights and permissions in Handshake. We are also working on implementing a lifecycle model for user management as users move from organization to organization and want to stay connected to Handshake. Finally, we will continue to explore integration with other socially enabled services and applications in the enterprise.

5 Evaluation Since the initial release in August 2009, Handshake has grown to support over 4300 users (~1000 are external participants) and 450 groups and communities. We are in the formative stages of a follow-on research project that will allow us to study adoption, motivation, incentives, and facilitation in a longitudinal evaluation of socially-enabled business models. Social media follows a long-tail model where contributions from each individual are small, but there are many individuals. Longtail effects grow rapidly with population size, and impacts are seen in large populations [28]. To conduct a proper assessment, we need time for the technology to be adopted, understood, and work practices adapted. To that extent, we have instrumented Handshake with an event logging functionality and are looking at the relationships and interactions among Handshake users, their contributions, group activity, and “silent” participation (often involving sharing Handshake content outside the tool). We are planning to employ a “use case” evaluation methodology where individual users and groups will start by stating their business goals. We will provide an orientation to the group members, and we will follow up at a later time to assess which factors facilitated or hindered the attainment of their goals. We will also extract lessons learned as a way of identifying techniques to speed adoption, remove barriers to achieving value, and institutionalize the best practices. In the meantime, we have, however, been impressed with the creativity of the MITRE staff in adopting these new ways of working. We have collected anecdotal stories of business value achieved via new socially enabled collaboration techniques from our early adopter groups. For example, a chief scientist was approached by the head of a newly formed association which brings together organizations from across the country. The two needed an easy way to share and comment on policy and related issues affecting their centers nationwide. Today, Handshake is one of their primary means of communication and collaboration: "Handshake is very simple and intuitive, and it's helping people to connect in ways they've found difficult to do using traditional methods." Other staff members are turning to Handshake for some novel problem-solving activities. For example, a MITRE senior principal scientist is helping to envision the company's workforce of the future and how it can best be supported. Using a Handshake group called “MITRE 3.0”, he reached out to over 1,000 employees to

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

169

engage them in envisioning their future needs and work practices. "A social networking platform is uniquely suited to engaging a larger swath of users on various problems. The degree of engagement is high and the interactions cut across users working in various locations and domain areas," noted the group owner. Because of low barrier to entry and wide participation, discussions on Handshake are flattening the traditional notions of company hierarchy. In one discussion thread, a user lamented that MITRE was no longer hosting show-and-tell exhibits for internal research projects. Other members from across the company contributed ideas and later decided to host a science fair. This idea arose spontaneously as a grassroots effort without any corporate encouragement. However, managers tracking the discussion offered funding and communications assistance. The science fair was wellattended, and provided an opportunity for participants to meet face-to-face. In a similar vein, the corporate communications group created the MITRE Alumni group to help identify MITRE employees in historical photograph archives, from as far back as the 1950s and 1960s. Monthly, photographs are shared in Handshake where alumni and current employees identify the subjects and the event. Handshake groups are also proving valuable for supporting technical exchange meetings, multi-organizational working groups, MITRE-sponsor project teams, selfhelp groups, new employees, and other virtual communities such as the teleworker community. For many of these Handshake communities, user profiles provide a unique way of helping members who do not have face-to-face access get to know each other better. Finally, the activity rivers for network and group activities provide a unique situational awareness view not available in our other collaboration tools. One of our users noted that Handshake was an “ecosystem” for collaboration that is leading people to discover information they were unaware of.

6 Lessons Learned Handshake provides a trusted way to connect and collaborate across organizations, geographical locations, and time. By maintaining control over the information exchanges and the membership model, MITRE can support a safe environment for its employees while inviting selected outside partners. The flexible but extensive permissions model allows users to choose what information they want to share with whom, bringing just the right subset of members to bear on the issue being discussed or the artifact being created. In creating Handshake and watching it grow, we have learned – and continue to learn – several valuable lessons. It is important not to “reinvent the wheel.” We did not want to build our own “Facebook for the enterprise” but instead focused on how we could utilize this new technology to extend the ways MITRE works and to leverage the broader community to provide better recommendations and products to our customers. We realized we would have to deal with both the early adopters and those reluctant to change their ways. The early adopters quickly grasped the power of the new platform, eagerly adopted it so they could reach a broader swath of users and partners, and pushed us to provide more capabilities quickly. We also had to market, educate and explain the new paradigm to staff who were reluctant to change their existing work practices. We spent a significant amount of time working with our

170

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

internal corporation communications department to capture and publish the early success stories so these new ways of collaborating could be more easily understood and emulated. We also recognized that our users were not going to abandon their existing tool suites, so we integrated the new capabilities with tools they were already using. We observed, for example, that near-real-time notification via email alerts was one of the major contributions to the adoption of Handshake. Letting users employ channels they were most familiar with to access their business network and communities was a plus for everyone. We made sure that the tools were lightweight and required no training to use, but we did provide education (such as webinars) for new concepts. We understood the implications of culture and tried to design for the digital natives, the digital immigrants, and the digital tourists. We developed a policy for both content and user behavior. Since security was important, we built the security features early and made the permissions model easy to understand. Lastly, we learned that measuring the value of social media in the enterprise cannot be done in isolation with a handful of pilot users. The long-tail effect of establishing connections and maintaining peripheral awareness may only be realized over an extended period of time with many cross-organizational members. Acknowledgments. The authors would like to thank all contributors to the Handshake project: Jesse Ciancetta, Lena Fried, Michael Jett, Austin Kennedy, Ernie Kim, Francine Lalooses, Jon Maul, Ed Overly, Tom Read, and Justin Richer. Handshake was funded, in part, by the MITRE Innovation Program, MITRE’s internal research and development program.

References 1. Boyd, D., Ellison, N.B.: Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication 13(1), 210–230 (2007) 2. Slideshare: Facebook Demographics (January 2011), http://www.slideshare.net/amover/facebook-demographics2011?from=ss_embed 3. Ellison, N.B., Steinfield, C., Lampe, C.: The Benefits of Facebook “Friends:” Social Capital and College Students’ Use of Online Social Network Sites. Journal of ComputerMediated Communication 12(4), article 1 (2007) 4. Wildbit: Social Networks Research Report (2005), http://wildbit.com/wildbit-sn-report.pdf 5. Dunbar, R.: Grooming, Gossip, and the Evolution of Language. Harvard University Press, Cambridge (1996) 6. Dunbar, R.: You’ve Got to Have (150) Friends. In: The New York Times, The Opinion Pages (2010), http://www.nytimes.com/2010/12/26/opinion/26dunbar.html 7. Bodnar, Kip: The Ultimate List: 100+ Facebook Statistics [Infographics], http://blog.hubspot.com/blog/tabid/6307/bid/6128/TheUltimate-List-100-Facebook-Statistics-Infographics.aspx 8. Kissmetrics: Facebook Statistics, http://blog.kissmetrics.com/facebook-statistics/ 9. Donath, J.: Signals in Social Supernets. Journal of Computer-Mediated Communcation, 13(1), article 12 (2007)

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

171

10. Boyd, D.: Friends, Friendsters, and MySpace Top 8: Writing Community Into Being on Social Network Sites. First Monday 11, 12 (2006), http://firstmonday.org/issues/issue11_12/boyd/index.html 11. Haythornthwaite, C.: Social Networks and Internet Connectivity Effects. Information, Communication & Society 8(2), 125–147 (2005) 12. Putnam, R.: Bowling Alone: The Collapse and Revival of American Community. Simon & Schuster, New York (2000) 13. Granovetter, M.: The Strength of Weak Ties: A Network Theory Revisited. In: Marsden/Lin, Social Structures and Network Analysis, pp. 105–130 (1982) 14. Donath, J., Boyd, D.: Public Displays of Connection. BT Technology Journal 22(4), 71–82 (2004) 15. Wellman, B., Haase, A.Q., Witte, J., Hampton, K.: Does the Internet Increase, Decrease, or Supplement Social Capital? Social Networks, Participation, and Community Commitment. American Behavioral Scientist 45(3), 436–455 (2001) 16. Resnick, P.: Beyond Bowling Together: SocioTechnical Capital. In: Resnick, P., Carroll, J.M. (eds.) HCI in the New Millenium, ch.29, pp. 247–272. Addison-Wesley, Reading (2002) 17. Parise, S., Iyer, B., Cuomo, D., Donaldson, W.: MITRE Corporation: Using Social Technologies to Get Connected (2011), http://www.iveybusinessjournal.com/mitrecorporation-using-social-technologies-to-get-connected 18. Gartner, Inc.: Organizations Face 5 Major Challenges Regarding Social Software (2008), http://software.tekrati.com/research/10150/ 19. Smith, J.: Updated Lists of All Companies and Regions on Facebook (2006), http://www.insidefacebook.com/2006/11/15 20. Gotta, M.: Field Research Study: Social Networking within the Enterprise (2009), http://www.burtongroup.com 21. Millen, D., Feinberg, J., Kerr, B.: Dogear: Social Bookmarking in the Enterprise. In: CHI Proceedings, pp. 111–120 (2006) 22. Gaudin, S.: Cisco’s Quad Takes Social Networking to the Enterprise. Computerworld (2010), http://www.computerworld.com/s/article/9177766/Cisco_s_Quad_ takes_social_networking_to_the_enterprise 23. The MITRE Corporation, http://www.mitre.org 24. Curry, R., Kiddle, C., Simmonds, R.: Social Networking and Scientific Gateways. In: Proceedings of the 5th Grid Computing Environments Workshop (2009) 25. Elgg open source social networking platform, http://www.elgg.org/ 26. Cuomo, D.: Say Hello to Handshake (2010), http://www.mitre.org/work/info_tech/software_collaboration/ 27. Gertner, A., Gaimari, R., Richer, J., Bartee, T.: Contact Recommendations from Aggregated On-Line Activity. Submitted to Conference on User Modeling, Adaptation and Personalization (2011) 28. Brenner, R.: Adopting Social Media: Projects, Pilots, and Politics. In: Cutter Consortium Business-IT Strategies Advisory Service Executive Update, vol. 12(13) (2009)

“Your Team Cohesion is Low”: A Systematic Study of the Effects of Social Network Feedback on Mediated Activity Luciano Gamberini, Francesco Martino, Anna Spagnolli, Roberto Baù, and Michela Ferron Human Technology Lab, Department of General Psychology, University of Padova Via Venezia, 8 – 35131, Padova (Italy) {luciano.gamberini,francesco.martino,anna.spagnolli}@unipd.it, {bau.roberto.htlab,michela.ferron}@gmail.com

Abstract. Collaborative mediated environments compete to provide visitors with social feedback, whose actual effects on visitors’ behavior is poorly known. This study considers feedback based on Social Network Analysis (SNA) and assesses whether this feedback is able to affect user activity in an online collaborative game. The results show that SNA feedback is able to modify group activity beyond a mere novelty effect, especially on the dimensions of the behavior covered by the feedback itself. The results also point to the possible role of task type in accounting for the feedback effect on behavior. Keywords: Social Network Analysis, Social Feedback, Computer Mediated Communication, Online Collaborative Games.

1 Introduction The availability of digital technologies to communicate and collaborate has continuously progressed over the last three decades [1]. A recent result of this development is social networking services implemented on websites and in other mediated environments, which turn users’ network from a latent, hidden infrastructure into an explicit, visible social object with its own value. Models and algorithms, elaborated by Social Network Analysis (henceforth, SNA) [2],1 are usually adopted to extract the structure and properties of a social network; several applications use them to make interactions visible or create social statistics [3], namely as a sort of feedback [4]; however, the effect of this information on user behaviors has seldom been evaluated (with the exception of a few studies; e.g., Morris [5]). The current study aimed to fill this gap by investigating, systematically, the effect of SNA feedback on specific dimensions of mediated activity. 1

For instance, the popular blogging service Technorati provides information on the “popularity” of a blog by calculating the number of links connected to it.

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 172–181, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

173

Previous studies using SNA-based feedback have found that they are effective in sustaining and increasing user activities [6,7], especially when information is perceived as correct and trustworthy [8]. The current study extends these results by adding two feedback types (density and centralization). These added SNA indices are at the group level (i.e., describing the status of the whole network with respect to a certain dimension) and literature has reported differences when feedback is administered at the group versus individual level [9,10]. Thus they are worth a specific investigation. The first hypothesis served as a preliminarily check to determine whether the newly added conditions could trigger the same effect observed by Martino et al. [6], namely an increase in the number of messages exchanged in the group. The explanation for that effect was that feedback on communication made more relevant the interactional dimension of the activity and then encouraged participants to invest more in this dimension by increasing their behavioral engagement. The number of messages exchanged in the group was taken as an overall, generic measure of this mutual behavioral engagement. H1: The provision of SNA feedback on communication increases the number of messages sent from every player to the other team members with respect to control groups with no feedback provision. The second hypothesis is related to the stability of this effect across subsequent activity sessions, discarding the possibility that the feedback effect is merely due to its novelty. In addition, Martino et al. [6] found that feedback, based on reciprocity, had less stable effects than did feedback based on centrality. As an explanation, in [6] it was hypothesized that some fatigue emerged from the previous session and made subtler aspects of communication (such as those needed in ensuring symmetry between all exchanges) more difficult to control. Thus, in the current study we tested if this explanation would hold true for group feedback regarding communication symmetry: H2: Feedback density increases communication activity in all game sessions, whereas feedback on centralization has a smaller or no effect in the final game sessions. The third aspect considered was the specificity of the feedback effect because, in the literature, it has been assumed but not directly investigated. In fact, there are only a few studies that have systematically evaluated the effect of augmenting groupmediated interaction via the provision of some type of feedback about the communication activity [11,12,13]. These studies have tested the effectiveness of such feedback on the aspects of group processes covered by the feedback itself (which might increase or decrease according to user’s position in the group), somehow presupposing that feedback worked by affecting the specific areas of behavior on which it provided information. In the current study, this assumption was tested directly, since effectiveness and stability were assessed on several dimensions of communication with respect to the specific information provided by the feedback. H3: A specific dimension of communication is always affected by the feedback on that same dimension. More specifically, the value of degree centrality increases after

174

L. Gamberini et al.

the provision of degree-centrality feedback; the value of reciprocity increases after provision of reciprocity feedback; the value of density increases after provision of density based feedback; the value of centralization increases after provision of centralization feedback.

2 Method Participants were grouped into teams that played an on-line treasure hunting game (based on the Crossfire platform) for four consecutive sessions. Participants communicated via dyadic chat (through Skype®2) and their message exchanges were logged and measured. Prior to each new game session, participants in the feedback conditions viewed an SNA index that measured a property of the communication activity of the group from the previous game session. Type of feedback varied between subjects and could cover Centrality, Reciprocity, Density, or Centralization; no feedback was provided in the Control Condition. The two new feedback conditions added in the current study involved 80 voluntary participants with similar characteristics of [6] (age M = 23.05, SD = 2.71; 37 men, 43 women). They were recruited by asking university students met at the university premises; people who agreed to participate where then randomly assigned to the different teams. The overall design was composed by these new conditions in addition to the conditions of the previous study. In total, the design involved 200 participants (86 men and 114 women, age: M = 23.66, SD = 4.14) who were distributed among five between subjects feedback conditions. Each condition included four teams (40 participants). 2.1 Feedback The types of feedback displayed are shown in Figure 13. “Centrality” describes how extensively a social actor is involved in social relationships with other actors in the communication network [2]. In the current investigation, centrality was visualized as a network; nodes represented actors and lines represented the ties between actors. Nodes with a higher centrality occupied a more central position, had a wider diameter, and a darker tone of green (Figure 1a). “Reciprocity” is the tendency of an actor to reciprocate the action of other actor and can be considered as a measure of cooperation [14]. The event in which this symmetry was measured was the initiation of a dyad communication exchange. This event was called “Thread Starting Request” (TSR), where threads were temporally 2

3

We choose to use Skype because it is more intuitive than the communication system embedded into Crossfire. Some informal pilot sessions were conducted to determine whether participants could easily use these two programs together and no difficulty was observed (nor were any difficulties complained about during the actual experimental sessions). Only dyadic communication was allowed in order to make the recipient of the message and then the analysis of the ties in the network more straightforward (of course, participants were allowed to open more chats simultaneously to keep in touch with more than one teammate at a time). While all indices are mentioned in this paper with the name they have in SNA literature, in this study some were re-named with more intuitive labels to better convey their meaning to common users.

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

175

bounded sequences of messages on a single topic [6,15] and a TSR was a message that potentially started a new thread. Participants reciprocated if both individuals in a dyadic chat attempted to start a new thread of messages at some point during the session. The representation of this index (Figure 1b) consisted of a graduated scale; two hands were joined when reciprocity between a certain actor and the rest of the group was at a maximum and the hands were separated when there was no reciprocity.

Fig. 1. Examples of the four types of feedback used in the various feedback conditions: (a) degree centrality, (b) reciprocity, (c) density, (d) and centralization

“Density” measures network cohesion [2]. The representation used in our study (Figure 1c) consisted of a graduated meter with an arrow pointing to the position on the meter that corresponded to the index value. “Centralization” measures the tendency of a network to be more or less centralized around a few actors [2]. The representation used in our study was similar to the one adopted for the density feedback, except for the two icons representing extreme values (Figure 1d). Formulas for calculating indices follow Wasserman & Faust [2], with the only exception of centralization (where the final value was achieved by subtracting the value obtained in the classic formula from 1, which maximized the index value when the network was decentralized). All index values varied from 0 to 1. 2.2 Task and Procedure Each participant was represented by an avatar in a large team-shared virtual world. The world contained hidden goblets (the treasure), different locations participants were free to explore, clues to help to find the goblets, and virtual food to keep the avatars alive. Participants acted in the virtual environment using a mouse and keyboard. To present the activity as a cooperative task, participants were informed that the winner would be the team that found the most goblets for all sessions. Participants who belonged to the same team met in a computer room. After signing an informed consent, participants received instructions on the game and played a 10minute training session to familiarize themselves with the game and the controls. Each experimental session lasted 20 minutes with a 25-minute break in between sessions. Feedback was displayed on the participant’s monitor at the beginning of

176

L. Gamberini et al.

each new session for the feedback conditions; a standard explanation of the feedback visualization and the index value was provided to all participants; no evaluation was implied in the explanation that could encourage the achievement of a specific index value. 2.3 Data Collection and Analysis The data collected were obtained from the chats that occurred during the game sessions. Of these data, the overall number of messages exchanged was calculated as well as Degree Centrality, Reciprocity, Density, and Centralization. The statistic procedure used to test the hypotheses varied according to the nature of the dependent variable; namely whether it measured data at the group or individual level. The former case applied to centralization and density and a mixed design ANOVA was performed, with Condition and Session as main factors. Individual indices included the number of messages, degree centrality, and reciprocity. For these measures, Multi Level Models (MLM) were adopted to account for the non-independency of scores for each individual in the group [16]. These models were implemented using SPSS Mixed Procedure [17], with Session and Condition as fixed effects and group belongingness as a random effect. Effect size was calculated using the d coefficient, as suggested by Williams et al [18]. Post-hoc analyses were conducted by comparing 95% confidence intervals4 for the mean value of groups in each condition and along the four gaming sessions.

3 Results 3.1 Effect on the Number of Messages Exchanged It was assumed that the number of messages exchanged in the group would measure the general feedback effect on communication activity. The results are illustrated in Figure 2. The analysis yielded significant results for the main effects of the factors

Fig. 2. Number of messages exchanged in each experimental condition in the four sessions of game. Error bars correspond to 95% Confidence Intervals. 4

Confidence Intervals were calculated using SPSS EMMEANS procedure for MIXED MODELS (SPSS v. 16).

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

177

“Condition,” F(4, 192.281) = 6.572, p < .001, d = .49 and “Session,” F(3, 222.75) = 10.331, p < .001, d = 0.47 and for the interaction between Session and Condition, F(12, 222.775) = 3.673, p < .001, d = .30. During the first session, the comparisons of estimated marginal means did not reveal any significant differences among conditions. In the second and third sessions, namely after the feedback provision, fewer messages were exchanged in the control condition compared to the feedback conditions; feedback conditions did not significantly differ. In the fourth session, the number of messages remained higher in the centrality and density conditions; however, it did not differ from the control group in the centralization and reciprocity conditions. 3.2 Effects on Degree Centrality Results for degree centrality are displayed in Figure 3. The analysis yielded significant results for the main effects for “Condition,” F(4, 194.431) = 19.764, p < .001, d = .22 and “Session”, F(3, 249.264) = 52.928, p < .001, d = .36 and an interaction effect for “Session” and “Condition,” F(12, 249.264) = 2.013, p = .024, d = .07.

Fig. 3. Mean values of degree centrality in each experimental condition and in the four sessions of the game. Error bars correspond to 95% Confidence Intervals.

During Session 1, all feedback conditions except for reciprocity were higher than in the control condition. From the second session onwards, all feedback conditions differed from the control condition. The initial difference with control, due to coincidence, was maintained steadily during all sessions. However, given that groups in the feedback condition differed from the controls since the first session, it is not possible to attribute the differences in the subsequent sessions to feedback alone. 3.3 Effects on Reciprocity The effects of all types of feedback on reciprocity are shown in Figure 4. The analysis yielded significant results for the main effects of “Condition,” F(4, 186.873) = 15.673, p < .001, d = .28 and “Session,” F(3, 290.570) = 11.140, p < .001, d = .67 and for the interaction between “Session” and “Condition,” F(12, 290.570) = 2.462, p = .004, d = .25. Comparisons of estimated marginal means showed no significant difference

178

L. Gamberini et al.

during the first session. In the second and third sessions, no differences were found between any feedback condition and, excluding Centralization, all conditions yielded a higher reciprocity value than did the control condition. Differences among feedback and control conditions remained in the fourth session, where reciprocity scores were also higher with reciprocity feedback compared to the other types of feedback. Centrality and density feedback were effective but less so than reciprocity feedback in the fourth session. Feedback specificity seems to have played a role here. In addition, since reciprocity was effective on reciprocity feedback in the fourth session, this aspect of communication was not affected by fatigue as hypothesized in [6]. This notion influenced H2 and will be discussed in the conclusion.

Fig. 4. Reciprocity values in each experimental condition and in the four sessions of the game. Error bars correspond to 95% Confidence Intervals.

3.4 Effect on Density Regarding the effect on density, a graphical representation of the mean values of density and how they varied across the different conditions is displayed in Figure 5. This analysis yielded significant results for the main effects of “Condition,” F(4, 15) = 5.985, p = .004, η2 = 0.51 and “Session,” F(3, 45) = 18.280, p < .001, η2 = 0.61. There was no significant interaction between “Session” and “Condition,” F(12, 45) = 2.462, ns, η2 = 0.068. As was the case for degree centrality, it is not possible to attribute any subsequent difference to the feedback and was not possible to test this hypothesis.

Fig. 5. Density values for each condition in the four sessions of the game. Error bars correspond to standard errors.

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

179

3.5 Effect on Centralization The results for centralization are displayed in Figure 6. The analysis yielded significant results for the main effects for “Session,” F(3, 45) = 4.826, p = .005, η2 = 0.20. There were no significant main effects for “Condition,” F(4, 15) =.583, ns, η2 = 0.12 and no significant interaction between “Session” and “Condition,” F(12, 45) = .943, ns, η2 = 0.16. Therefore, it seems that this value decreased from the first to the third session, regardless of the presence of feedback.

Fig. 6. Centralization values for each condition across the four sessions of the game. Error bars correspond to standard errors.

4 Discussion and Conclusion Finding the proper cue to influence how users communicate in a mediated environment has always concerned scholars in human-computer interactions [20]. The study reported here examined the effects of a class of augmenting cues (i.e., SNA feedback) on group communication. The effect on the number of messages exchanged (H1) was confirmed for all types of feedback. Stability (H2) was also confirmed: the effect persisted until the third session and afterwards decreased for reciprocity and centralization feedback. However, the explanation for this decrease is not as we originally hypothesized. Specifically, reciprocity feedback did not lose its effectiveness on reciprocity behavior, only on the general amount of messages exchanged. Thus, reciprocity cannot be considered as generally prone to fatigue as we thought in H2. Rather, it is possible that reciprocity feedback decreased its effect on the exchange of messages because it is not a generic dimension of communication, but a specific one; it is more related to message quantity than to message direction. In fact, feedback based on message quantity (centrality and density) was able to maintain longer effects. This result suggests that specificity affected (H3) reciprocity behavior, as hypothesized, but it also affected message quantity. Specificity effects were not found on centralization, degree centrality, or density. Since there was an initial difference due to sampling error in degree centrality and density among conditions, despite homogeneity in demographics, we can only affirm that differences found in the first session for these indices were maintained consistently during all sessions, without knowing if they were due to the feedback or to other

180

L. Gamberini et al.

differences between the groups. Regarding centralization, there may by other reasons for the lack of effect, which was probably connected to the nature of the task. Centralization concerns the level of hierarchization in the communication structure of the group. Because the goal of communication in this study was to circulate information, participants were unable to predictable who found this information; therefore, centralizing communication was not beneficial to the task. This finding can explain why centralization decreased, regardless of feedback, and suggests that the task can play an important role in defining what feedback information is used and how. In conclusion, each feedback affected at least some dimensions of communication and did so steadily. Therefore, SNA-based feedback can profitably be integrated in electronic communication systems to support user awareness of group interactions. The specificity of the feedback, with respect to the targeted behavior, plays a role in determining its effectiveness. This finding has important consequences in designing a feedback service. In case that service is not customized on a specific activity, it is recommended that designers provide a repertoire of different feedback types that users can chose from based on the needs of their activities. Finally, further studies are needed to examine the role of the task in mediating the feedback effect. Acknowledgments. The study reported here is partially funded by the PASION project (Psychologically Augmented Social Interaction over Networks, reference number 27654 PASION, EU IST program). All authors contributed equally to this work.

References 1. Schmidt, K.: Cooperative work and coordinative practices. Springer, London (2011) 2. Wasserman, S., Faust, K.: Social Network Analysis. Theory and methods. Sage, New York (1994) 3. Isbell, C.L.I., Kearns, M., Singh, S., Shelton, C.R., Stone, P., Kormannn, D.: Cobot in LamdaMOO: An adaptive social statistics agent. Auton. Agents and Multi-Agent Syst. 13, 327–354 (2006) 4. Kluger, A., DeNisi, A.: The effects of feedback interventions on performance: a historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychol. Bulletin 119, 254–284 (1996) 5. Morris, M.E.: Social networks as health feedback displays. IEEE Internet Computing 9, 29–37 (2005) 6. Martino, F., Baù, R., Spagnolli, A., Gamberini, L.: Presence in the age of social networks: Augmenting mediated environments with feedback on group activity. Virtual Real 13, 183–194 (2009) 7. Kirman, B., Lawson, S., Linehan, C., Martino, F., Gamberini, L., Gaggioli, A.: Improving Social Game Engagement on Facebook through Enhanced Socio-Contextual Information. In: Proceedings of ACM CHI 2010, pp. 1753–1756. ACM Press, New York (2010) 8. Gamberini, L., Martino, F., Scarpetta, F., Spoto, A., Spagnolli, A.: Unveiling the structure: Effects of social feedback on communication activity in online multiplayer videogames. In: Schuler, D. (ed.) HCII 2007 and OCSC 2007. LNCS, vol. 4564, pp. 334–341. Springer, Heidelberg (2007)

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

181

9. DeShon, R.P., Kozlowski, S.W.J., Schmidt, A.M., Milner, K.R., Wiechmann, D.: A multiple-goal, multilevel model of feedback effects on the regulation of individual and team performance. J. Appl. Psychol. 89, 1035–1056 (2004) 10. Ilgen, D.R., Fisher, C., Taylor, M.S.: Consequences of individual feedback on behavior in organizations. J. Appl. Psychol. 64, 349–371 (1979) 11. DiMicco, J.M., Hollenbach, K.J., Pandolfo, A., Bender, W.: The impact of increased awareness while face-to-face. Human-Computer Interaction 22, 47–96 (2007) 12. Losada, M., Sanchez, P., Noble, E.E.: Collaborative technology and group process feedback: their impact on interactive sequences in meetings. In: Proceedings of the 1990 ACM Conference on Computer-Supported Cooperative Work, pp. 53–64. ACM Press, New York (1990) 13. Zumbach, J., Schonemann, J., Reimann, P.: Analyzing and supporting collaboration in cooperative computer-mediated communication. Paper presented at Learning 2005: the Next 10 Years! Taipei, Taiwan (May-June 2005) 14. Mui, L., Mohtashemi, M., Halberstadt, A.: A Computational Model of Trust and Reputation. In: Proceedings of the 35th Hawaii International Conference on System Science, p. 122. IEEE Computer Society, Washington (2002) 15. Yates, J., Orlinowski, W.J., Wörner, S.L.: Virtual organizing: using thread to coordinate distributed work. Working papers, pp. 4320–4323. MIT, Sloan School of Management (2003) 16. Freeman, L.C.: Centrality in Social Networks: I. Conceptual clarification. Soc. Netw. 1, 215–239 (1979) 17. Kenny, D.A., Mannetti, L., Pierro, A., Livi, S., Kashy, D.A.: The statistical analysis of data from small groups. J. Pers. Soc. Psychol. 83, 126–137 (2002) 18. Peugh, J.L., Enders, C.K.: Using the SPSS mixed procedure to fit cross-sectional and longitudinal multilevel models. Educ. Psychol. Meas. 65, 717–741 (1999) 19. Williams, D., Caplan, S., Xiong, L.: Can you hear me now? The impact of voice in an online gaming community. Hum. Commun. Res. 33, 427–499 (2007) 20. Oulasvirta, A., Petit, R., Raento, M., Tiitta, S.: Interpreting and acting on mobile awareness cues. Hum.-Comput. Interact 22, 97–135 (2007)

E-Business Solutions in the Cable TV Industry Viorica Harrison and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected]

Abstract. This paper aims at developing implementable e-business solutions for the cable TV industry to accelerate the information technology and information system adoptions; and thereby, increase value for the operations in the industry. Specifically, an electronic value chain is developed to show how to improve value by using information technology and information system. Then, a set of implementable solution items are developed based on this electronic value chain model. Finally, recommendations are provided on how to implement these solution items. Keywords: Value chain, cable TV industry, electronic business.

1

Introduction

Although many large cable TV providers are fully utilizing the Internet and transforming their business operations to include complex e-Infrastructures and IT applications, other providers are slow to move to e-business solutions because of high costs, risks, business reengineering, and failure to recognize the future value shifts toward online television viewing [2]. Cable TV industry providers that neglect to exploit e-business solutions are not aligning their information systems (IS) with their business and organizational strategies. Since IT is integrated in nearly all business aspects, as technology develops, companies must change IS quickly to meet market demands and compete effectively. Cable TV providers, therefore, should use scalable e-Infrastructures for e-business solutions to increase time-to-market speed, expand business functionality, lower costs, improve customer satisfaction, and increase revenues[3] [4] . As more customers migrate to the Internet and wireless devices for their television viewing, the cable TV industry will encounter considerable reformation. Cable TV providers that do not adopt sufficient information technology applications and solutions in their supply chain will encounter massive revenue losses, similar to what the music industry experienced at the turn of the 21st century, before iPod and iTunes [2]. As it stands now, the abundance of technology and Internet programming are already forcing cable TV providers to offer premium online TV content to their existing customers to keep them from accessing online shows through free Internet sites, such as YouTube and Hulu [1]. Since the US cable TV market is a mature and heavily saturated industry, increased free television electronic content and online A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 182–187, 2011. © Springer-Verlag Berlin Heidelberg 2011

E-Business Solutions in the Cable TV Industry

183

viewing could lead to decreased national cable TV industry growth and increased industry competition [8]. Comparing with other industries, IT adoptions in the cable TV industry are still slow. Increased online television viewing may bring about many operational challenges resulting from increased consumer broadband traffic, limited provider bandwidth, and network system bottlenecks, which will cause significant delays of econtent delivery [9][10]. The purpose of this paper is to develop an electronic value chain model for the cable TV industry, decompose this model into a set of implementable solutions for increasing value for the cable TV industry, and provide recommendations for e-business solution implementations. The remaining of this paper is organized as follows: Section 2 presents a value chain model for the cable TV industry. Section 3 develops a set of e-business solutions based on the value chain model. Section 4 provides discussions, recommendations, and conclusions.

2

Electronic Value Chain Model

Fig. 1 presents a value chain model based on the utilization of IT in the cable TV industry. This model was developed based on Michael Porter’s value chain with five primary activities (inbound logistics, operations, outbound logistics, marketing and sales, and services) [12]. Inbound logistics involves receipt of programming content, equipment, inventory storage and management. Cable TV operators can receive, track, and manage video inventories, such as video-on-demand VoD and pay-per-view PPV, online by processing incoming video libraries in batches [2]. For live content delivery, real-time streaming is used and, therefore, the content simultaneously goes through inbound and outbound logistics to reach subscribers. For equipment inventories, cable TV providers can use just-in-time (JIT) inventory systems in order to achieve efficient inbound logistics. For example, a barcode strip can be assigned to each individual piece of television, Internet, and telephone equipment to facilitate easy e-tracking of inventory issued to technicians and contractors for installation. At the end of the day, scanners can be used to return the bar-coded inventory items to the warehouse to be placed back in stock. Operations in the cable TV industry include in-house content production and packaging, content management and protection for both television and Internet outlets, and facility operations. Cable providers package television programming received from inbound logistics into two main tiers, basic and digital. The digital packages consist of premium content and include access to VoD, PPV, DVRs, and HDTV services [1]. To remain viable and earn profits, content library management and protection becomes increasingly important, especially when incorporating ondemand and pay-per-view options with Internet TV. Also, since many subscribers utilize the Internet to manage their cable TV accounts, offering secure e-payment systems and protecting users’ stored personal data by using encryption is critical for cable TV operators. The incorporation of electronic data interchange (EDI) into facility operations assists in creating more efficient supply chain relationships. EDI technology “enables

184

V. Harrison and J. Wei

the transmission of routine business documents having a standard format from computer to computer over telephone or direct leased lines” [6]. Cable TV providers utilize EDI to improve communication between programming networks and advertising agencies to manage advertising spot schedules and build an e-infrastructure for e-business [7]. Outbound logistics in the cable TV industry involves video, data, and telephone service delivery to subscribed users. Increasingly, cable providers have been using Web-based technologies to deliver television and advertising content over the Internet. Today, e-logistics, including e-service distribution, e-content delivery, and online VoD and PPV order dispatch enable automated distribution of movies, sport and concert events, to the end-user via the Internet. Also, due to e-logistics, viewers are able to actively interact with their service providers and receive products and services at faster speeds. Marketing and sales efforts can also be enhanced by the use of technology. For the cable TV industry, e-marketing includes e-advertising, e-sales, web-based press room portals, online newsletters and e-mail usage. Besides Web portals, e-mail selling and e-advertising is very common in the cable TV industry. An investigation of the top twenty companies’ web pages revealed that each company does an excellent job in eadvertising their own products and services via their own Web pages. Customer service represents the last function within the primary activities of the value chain. Although firms are mostly concerned with cost savings and quality, time savings can ultimately be the deciding factor between keeping a customer and loosing a customer to the competition [11]. Today, cable operators must meet customers’ busy on-the-go lifestyles by providing fast, efficient, and satisfactory customer service. Utilization of technology, such as customer relationship management (CRM) software, online chat, online account management, e-troubleshooting of program and video errors, e-order confirmation of purchased programs, e-feedback, e-testing of equipment, and e-mail will increase customer satisfaction, acquisition and retention.

Fig. 1. IT Adopted E- Value Chain Model for the Cable TV Industry

E-Business Solutions in the Cable TV Industry

3

185

E-Business Solutions

E-business solutions are IT applications that are utilized in an industry’s e-value chain. This analysis paper focuses on five categories of e-business solutions: B2B (business-to-business), B2C (business-to-consumer), C2B (consumer-to-business), B2G (business-to-government), and B2I (business-to-internal). Since the cable TV industry provides video, Internet, and telephone services to business customers, governmental customers, and residential customers, some ebusiness solutions are utilized in more than just one category. For instance, ecommunication and company websites are used in both B2B and B2G categories, while more specific e-communication applications, like e-mail alerts, exist in B2C relationships. Research shows that the majority of e-commerce happens in the B2B and B2C categories since government has been somewhat slower at adopting ebusiness solutions [5]. However, presently, government e-commerce is increasing and companies can greatly benefit from incorporating e-government into their ecommerce business models. The focus of this research paper is to analyze e-business solutions present in the cable TV industry’s e-value chain model. Table 1 lists 54 e-business application items employed by the cable TV industry, summarized by e-business categories. Table 1. E-Business Solutions for the Cable TV Industry 12 Items B2B E-Business Portal

24 Items B2C 3 Items C2B E-News Portal Online Customer Feedback

7 Items B2G Cable in the Classroom

8 Items B2I E-Knowledge management and Sharing System E-Payment System Electronic Online Job E-Payment E-Human Program Guide Search/Application System Resource Management (HRM) E-Procurement Interactive E-Parental Controls E-Public Access E-mail Television Television (ITV) E-Communication E-Mail Alerts E-Military Internet Access Electronic Data Online Intranet Interchange (EDI) Advertising E-911 Service E-search Online Account Electronic EManagement Teams Communication Online Cable Ad E-Customer Employee EInventory Help / Support Company Training and EManagement Website Learning Company Website Online Security Digital Technology Online Online Clearinghouse Shopping Mall

186

V. Harrison and J. Wei Table 1. (continued)

Extranet Interface with Suppliers and Partners Online Partner Ads Wireless E-mail

Online Equipment Testing/Upgrade Online Data Storage Online Multilingual Services E-Interactive FAQ’s Video-OnDemand (VoD) Pay-Per-View (PPV) Online Customization Tools Online Outlet Locator E-Video Error Repair Live Online Chat Online Games/Music Online 3D Content

4

Discussions, Recommendations and Conclusions

The cable TV industry has gone through many technological changes in the last ten years. The introduction of innovative technologies, such as interactive television, VoD, PPV, DVR, online TV, and lately, 3D TV, has revolutionalized the cable business into a complex and dynamic industry. The cable TV industry value is shifting from traditional broadband viewing to online and wireless television viewing access. Consumers’ desires and trends toward on-the-go flexible programming and Internet information access are forcing companies to innovate and “get in front of the [technology] change or the consumers [will] threaten to leave them behind” [2]. As such, many cable TV companies currently employ various e-business solutions to improve efficiency, reduce waste, and increase information flow between activities in the value-chain with the sole end goal—to satisfy the ultimate customer. In this paper, the e- value chain model was developed to illustrate how various applications could increase traditional value creation through the use of IS technology [12]. Further, e-business solutions were analyzed at each step of the e- value chain to better illustrate linkages and relationships between various activities in the e-value

E-Business Solutions in the Cable TV Industry

187

chain, e-business applications, and five e-business groups (B2B, B2C, C2B, B2G and B2I). Clearly, if cable TV operators desire to achieve strategic competitive advantages vis-à-vis their rivals, they need to adopt as many practically useful e-business solutions as possible. Of course, adding random e-business solutions without considering value enhancement, operations efficiency, and cost reductions is not recommended. However, in light of the research and analysis provided in this paper, it is evident that adopting less valuable e-business solutions can lower a company’s status in the market place. A prime example is Armstrong Cable Services, which is ranked twentieth by NCTA in terms of US market share, and which also has the lowest e-business solutions implementation rate. Future research will focus on data collection for the top companies in the cable TV industry to see the implementation patterns of these e-business solutions; and thereby, provide suggestions and guidelines on how to improve value for the cable TV industry in general.

References 1. Amobi, T.N., Kolb, E.B.: Broadcasting, Cable & Satellite. Standard & Poor’s Industry Surveys.from Standard and Poor’s NetAdvantage (retrieved, March 15, 2010) 2. Berman, S., Duffy, N., Shipnuck, L.: The end of television as we know it. A Future Industry Perspective. IBM Institute for Business. [Value Study] (2006) 3. Cable, Internet & Telephone Providers in the US: 51322. IBISWorld Industry Report. from IBISWorld (retrieved February 13, 2010) 4. Dess, G., Lumpkin, J., Eisner, A.: Strategic Management, 4th edn. McGraw-Hill, New York (2008) 5. Gordon, L.: Business-To-Government E-commerce Procurement: Online Centers, Databases, and Tools. AllBusiness. A D&B Company (2000), from http://www.allbusiness.com/technology/software-servicesapplications-internet/10616612-1.html (retrieved April 10, 2010) 6. Krajewski, L., Ritzman, L.: Operations Management: Processes and Value Chain. Prentice-Hall, New Jersey (2006) 7. Larson, M.: Charging hard on EDI: CAB hopes to have nets, agencies on same e-based adsales system next year (cable TV). MEDIAWEEK, 14, 45. p.5(1) (2004) 8. McClellan, S., Morrissey, B.: The future of TV: flexible, interactive and measurable, TV will work a lot like, well, the Web. MEDIAWEEK 18(41), 20(9) (2008) 9. Pearlson, K., Saunders, C.: Managing & Using Information Systems. John Wiley & Sons, Chichester (2006) 10. Turban, E.: Electronic Commerce – A Managerial Perspective. Prentice-Hall, New Jersey (2006) 11. Walters, D., Lancaster, D.: Implementing value strategy through the value chain. Management Decision 38(3), 160–178 (2000); from ABI/INFORM Global. (Document ID: 254418631), (retrieved April 7, 2010) 12. Wei, J., Platt, R., Boyd, W., Jasquith, A.: Development of Standardized E-Business Solutions via E-Chain Analysis in the Digital Utility. International Journal of Services and Standards 2(2), 117–136 (2006)

Measuring the Success of On-line Communities in an Enterprise Networking Tool Lester J. Holtzblatt and Laurie E. Damianos The MITRE Corporation, 202 Burlington Road, Bedford, MA 01730, USA {holtzblatt,laurie}@mitre.org

Abstract. We look at self-forming communities on one company’s externallyfacing business networking tool to better understand factors contributing to the success of a community. We propose that there is no single measure of success of a community; success of different kinds of communities depends on a variety of factors, including the perspective of the stakeholders and the type of community established. Keywords: communities, social software, social computing, web 2.0, social media, collaboration, collaborative environments, group and organization interfaces, asynchronous interaction, collaborative computing, computersupported cooperative work, evaluation/methodology, web-based interaction.

1 Introduction In this paper, we discuss an on-going research project evaluating factors indicative of the success of groups formed within Handshake, a business networking prototype deployed at the MITRE Corporation [1]. There is a considerable amount of published literature on measures of the success of on-line groups [2], [3], [4], [5], [6]. However, most studies show that there is no single quantitative measurement that can be used to assess a community’s success [7]. Although many of the typical quantitative measurements can provide useful information on the activity of a community, none of these is necessarily a true measurement of success [8]. Success also depends highly on perspective – the perspective of the community owners, facilitators (if any), and community members, as well as other stakeholders in the community. Communities form for many reasons, but even communities established for similar purposes may have different ways of evaluating their own success [9]. Based on our initial evaluation of Handshake groups, we have similarly found that a variety of factors contribute to and are indicative of a community’s success. Furthermore, we found that different kinds of communities form within Handshake, and that measurements of success appear to depend on the type of community established. In this paper, we will describe the different types of groups established within our business networking tool and factors associated with success for these different types of groups. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 188–196, 2011. © Springer-Verlag Berlin Heidelberg 2011

Measuring the Success of On-line Communities in an Enterprise Networking Tool

189

2 Background and Motivating Problem The MITRE Corporation is a not-for-profit organization with expertise in systems engineering, information technology, operational concepts, and enterprise modernization. In addition to managing several Federally Funded Research and Development Centers, MITRE supports its own independent technology research and application development for solving sponsors' near-term and future problems. MITRE has approximately seven thousand scientists, engineers and support specialists distributed across many locations and working on hundreds of different projects in numerous domains. For solving problems, staff is expected to seek out and rely on the expertise and knowledge of technical and domain experts distributed across the company. As a result, the corporation places a high value on sharing knowledge across individuals, projects, and business units. Information has historically been shared through the use of ListServs, technical exchange meetings, internal wikis, and Microsoft Sharepoint. For communicating and collaborating with our external partners, MITRE employees have traditionally relied on email, telephone, and face-to-face meetings. Artifacts distributed via email get lost or conflicts occur when different copies of materials are edited. Telephone conversations are ephemeral, and face-to-face meetings involve travel and scheduling issues. The company also uses an external Microsoft SharePoint site as a secure document repository, but the strict authentication requirements and the lack of lightweight functionality make it difficult to collaborate using this tool. In addition to these communication and collaboration problems, individuals want to be able to establish relationships with external parties and manage those contacts as people move from organization to organization. The recent popularity of tools like LinkedIn and Facebook make connecting and tracking possible, but they are not trusted environments for conducting business and are fraught with overwhelming amounts of personal and extraneous information. To address these issues, we explored existing products - commercial technology, open source tools, and hosted platforms - to understand what functionality existed and how people were using such tools. We also performed work practice studies with existing cross-organizational teams in order to document current work practices, assess needs that were not being met, identify breakdowns and ineffective use of technology for communication and collaboration, and gather requirements for our prototype.

3 Case Study Through MITRE’s research program, a team developed and launched Handshake [1], a business networking prototype based on Elgg, an open source platform [10]. The Handshake prototype provides a MITRE-owned space for connecting, collaborating, and networking with both internal and external partners in a trusted environment as well as a research platform for exploring the value of social media in the enterprise. Handshake is open to all MITRE employees. Members can connect with each other, create a profile, and establish groups for any purpose. MITRE staff must explicitly invite external participants to join Handshake, but these external partners can, in turn, establish connections with other Handshake members and join groups.

190

L.J. Holtzblatt and L.E. Damianos

Groups may be open to all members of Handshake, limited to MITRE employees, or restricted to a named subset of members. Groups with moderated access require approval for members to join; in open groups, members can self join. Available group tools include a discussion forum, a file repository, wiki-like “pages,” blogs, photo albums, and a message board. Each piece of content within a group inherits the corresponding group permission but may also be restricted further (e.g., MITRE members of the group). A group owner may appoint one or more co-owners to have group administration rights. Owners and co-owners can add or remove members from the group, change the group access controls, edit group metadata (e.g., tags, related URLs, description, title), or delete any inappropriate content. Awareness of activity on Handshake is provided through activity streams and customizable email alerts. Email alerts provide profile information on the creator of the content, the content itself, and quick links into Handshake. At the time of the study, Handshake had been available for just over one year. There were approximately 3300 registered members, including 800 external participants, and 300 established groups. MITRE use of Handshake was voluntary, as was the creation of each of the groups.

4 Approach For this study, we looked at 243 groups that had been created within a one-year period. Most of the groups were still in the growth stage [11] and had not yet reached maturity. We did not look at newly created groups because they were too new to have established regular usage patterns. We also omitted test groups and hidden groups that had no members. Most of the groups examined were moderated and restricted to group members. On average, MITRE employees represented 87% of group membership although 12% of those groups had less than 50% MITRE employees as members with a few of them almost exclusively made up of external partners. Some groups had up to 6 co-owners, but the median was just one per group. The median size of each group was 11, with 977 members belonging to the largest group (open to All MITRE). We created a short survey that was distributed to 222 owners and co-owners of the 243 groups. We asked 6 questions about the purpose/goal/objective of the group, how they would categorize their group, whether there was a designated facilitator, how they rated the success of their group on a 5-point Likert scale, what factors they considered in evaluating the success, and what other tools they used to support this group. We gave the participants 3 days to respond to the survey. At the end of the second day, we sent a reminder to those who had not yet responded. We had 81 participants respond with their input on 128 groups (108 unique groups). For those groups with multiple respondents, we combined their responses and resolved minor conflicts by averaging scores and making judgment calls on group category. Some survey recipients chose not to participant but indicated to us that their groups had not been successful. The categories used in the survey (see Table 1) had been pre-created by classifying each group based on its description. These categories map closely to the corporate

Measuring the Success of On-line Communities in an Enterprise Networking Tool

191

Table 1. Group Categories

Category Community of Practice (COP) Community of Interest (COI) Project

Definition A group of people who share a common domain, field, or profession A group of people who share a common interest or passion, not directly related to their work A team of individuals collaborating on a funded project with a MITRE project number, or a work team for an assigned group project People comprising an area of work that encompasses different projects A task or advisory-level group of people chartered at the corporate, center, or enterprise-level, usually crossorganizational A MITRE organizational unit; e.g., a department or division A group of people who share a common past or present association with an organization such as a university, company, or professional association A group of people attending a specific event (may be a recurring event) (Survey participant as asked to define)

Program Council

Organization Affiliation

Conference / TEM Other

categories used for the internal Sharepoint site although, for this survey, we added Community of Interest and Affiliation, types of communities not found on the Sharepoint site. In addition to survey results, we collected data from the Handshake database and performed some basic calculations (see Table 2). Examining the group activity by month over time, we then classified the activity as consistent (moderated to high), consistent (low), and little to no activity. Table 2. Handshake Data Used in Study

Metrics & Calculations Group membership model Group permissions Group creation date / # days in existence # owners/co-owners # members

Notes e.g., open/moderated e.g., All Handshake, All MITRE, Group only, My Connections

This final number was pulled only at the time of the study. We were unable to ascertain member join date or whether people had joined and then left.

192

L.J. Holtzblatt and L.E. Damianos Table 2. (continued)

% MITRE members # & type of group content # comments on each type of content Median/max thread depth % contributors

% commenters % lurkers Group activity by month

e.g., discussion topics, blog posts, files, pages, albums

Those who contributed original content and may also have made comments Those who commented but did not contribute original content Those who did not participate in any group activity Based on amount of content & comments

5 Initial Results and Discussion Over half the groups established on Handshake were either Projects (both internal and cross-organizational) or Communities of Practice (COP). See Figure 1. Some types of groups were more likely to be rated as successful than other types of groups (Figure 2). One possible explanation for these results is that success is dependent on the clarity of the community’s mission or goal [2]. Projects and Conferences have clear goals and objectives whereas Programs, Organizations, and Affiliations typically do not have stated tasks or objectives. The goals of COPs or COIs may be fuzzier than the goals of Projects but more clear than that of Organizations. Fewer groups had designated facilitators, but having a facilitator did not appear to be a factor of community success (see Figure 3). The overall level of group activity appears to predict the community’s success (Figure 4). Groups with a consistent level of moderate to high activity were more likely to be rated as successful. This finding also holds true across types of content in groups; successful communities were engaged in more discussions, contributed more files, and created more pages than unsuccessful ones (Figure 5). Indeed, success criteria enumerated by survey participants typically included generation of artifacts, amount of shared content, quality of conversations, and ongoing activity. However, the type of activity a community engaged in differed across group category (Figure 6). Comparing the four most frequently occurring categories: Projects, COPs, COIs, and Organizations, we see that: • Projects were more likely to create/edit pages. This is consistent with the observation that project groups work collaboratively around an artifact. • COPs and Projects were equally likely to upload files and much more likely to do so than COIs or Organizations.

Measuring the Success of On-line Communities in an Enterprise Networking Tool

193

• COIs and COPs were equally likely to post discussion topics and much more likely to do so than Projects or Organizations. However, the depth of discussions in COIs is greater than in COPs. From a separate interview study, we learned that lurking (not actively contributing to group content) may actually be an active behavior, and that lurkers may highly value a community because of what they learn. These findings are consistent with the conclusions of other researchers who view lurkers as valued participants in online communities [9]. Because of the negative connotations often associated with the term “lurker”, Preece and Schneider [12] suggest that the term “reader” be adopted to describe on-line participants in a community who read, but seldom contribute content. Farzan et al. [13] note that there are several reasons for valuing the presence of lurkers in an on-line community. Through their participation they may acquire and use new knowledge they read for beneficial purposes. In addition, lurkers may spread their new knowledge beyond the community and, through word-of-mouth- draw other participants into the community.

Fig. 1. Percentage of Handshake Groups by Category

Fig. 2. Success Ratings Based on Group Category. Bars show the percentage of groups within each category that were rated as successful, neutral, or not successful.

194

L.J. Holtzblatt and L.E. Damianos

Unfortunately, we could not adequately measure the activity of lurkers. While we have access to logins and page views, we have no way of knowing if group members were reading and forwarding the Handshake-generated email alerts. Lacking visibility into these activities, we are unable to determine their level of participation in the community. Were they engaged? Were they reading but not contributing? Were they creating conversations around group content outside the group space?

Fig. 3. Success Ratings Based on Having Designated Facilitator. Bars show the percentage of facilitated or non-facilitated groups that were rated as successful, neutral, or non-successful.

Fig. 4. Success Ratings Based on Level of Activity. For each of 3 levels of activity, bars show the percentage of groups that were rated as successful, neutral, or not successful.

Fig. 5. Success Ratings Based on Content Type. Bars show the percentage of contributed content by type for groups rated successful, neutral, or not successful.

Measuring the Success of On-line Communities in an Enterprise Networking Tool

195

Fig. 6. Content Type by Group Category. This graph shows average distributions of content type across group categories.

6 Conclusions The initial results of our study have proved interesting. We have seen that not all measurements of success apply to each kind of community; group category and type of content are distinguishing factors. The size of a community is not necessarily a measure of success. Our data also indicates that having a group facilitator does not always ensure success, and not all groups require a facilitator to achieve success. New Handshake groups are continuing to form, some existing groups are becoming more established, and other groups have faded away. (In the 4 months following the conclusion of this study, 120 new groups were created). As groups continue to change over time, we will observe the activity of Handshake groups over a longer period as part of a follow-on research study. In addition, we would like to explore lurker behavior. We believe that the percentage of contributing members of a community is not in itself a measure of success; lurkers may be contributing to the success of a community although this is more difficult to measure. Acknowledgments. We thank Betty Fisher and Ernie Kim for their help in creating the survey.

References 1. Cuomo, D., Damianos, L.: Extending MITRE’s Reach: Business Networking For & Beyond the Enterprise (2010), http://www.mitre.org/work/tech_papers/2010/10_2435 2. Majchrzak, A., Malhotra, A., Stamps, J., Lipnack, J.: Can Absence Make a Team Grow Stronger? Harvard Business Review (2004) 3. Maybury, M.: Collaborative Analysis for Information Driven Safeguards. In: International Atomic Energy Agency Symposium on International Safeguards: Preparing for Future Verification Challenges, Vienna International Center, Vienna, Austria, vol. 145 (2010). IAEA-CN-184/145 4. Probst, G., Borzillo, S.: Why communities of practice succeed and why they fail. European Management Journal 26, 335–347 (2008)

196

L.J. Holtzblatt and L.E. Damianos

5. Pugh, K.: Sustainable Communities: Top 40 CSFs for Keeping the Faith (2010) 6. Suarez, L.: Community Builders – Building and Sustaining On-Line Communities by Steve Dale (2010), http://www.elsua.net/2010/08/27/communitybuilders-buildingand-sustaining-on-line-communities-by-steve-dale/ 7. Preece, J.: Sociability and Usability: Twenty Years of Chatting Online. Behavior and Information Technology Journal 20(5), 347–356 (2010) 8. Lazar, J., Preece, J.: Online Communities: Usability, Sociability and Users’ Requirements. In: van Oostendorp, H. (ed.) Cognition in the Digital World, pp. 127–151. Lawrence Erlbaum Associates Inc. Publishers, Mahwah (2002) 9. Preece, J., Maloney-Krichmar, D.: Online Communities: Focusing on Sociability and Usability. In: Jacko, J., Sear, A. (eds.) Handbook of Human-Computer Interaction, pp. 596–620 (2003) 10. Elgg, open source social networking platform, http://www.elgg.org/ 11. Iriberri, A., Leroy, G.: A Life-Cycle Perspective on Online Community Success. ACM Computing Surveys (SCUR) 4(2), 1–29 (2009) 12. Preece, J., Schneiderman, B.: The Reader-to-Leader Framework. AIS Transaction on Human- Computer Interaction 1, 13–32 (2009) 13. Farzan, R., DiMicco, J.M., Brownholtz, B.: Mobilizing Lurkers with a Targeted Task. In: th Proceedings of the 4 Int’l AAAI Conference on Weblogs and Social Media, ICWSM 2010 (2010)

Business-to-Business Solutions for the Cosmetic Industry Manami Koblentz and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected]

Abstract. This paper develops e-cosmetic business-to-business solutions based on strategic value chain management analysis. Specifically, it first investigates the e-business adoption in the major activities involved in the cosmetic business based on strategic value chain analysis. Then, it decomposes these activities into a set of implementable business-to-business solutions. Finally, the implementation patterns of these solutions for the top ten companies are analyzed. The results will be helpful in accelerating e-business adoption in the cosmetic business. Keywords: strategic value chain, electronic business.

1

Introduction

Cosmetics companies are getting more and more high tech in their research for effective anti-aging skin care products. Matching the leaps of technology in pharmaceuticals and biotechnology, cosmetics manufactures are getting more and more sophisticated in its ingredients, delivery, and operational system. In order to increase its efficiency and effectiveness, cosmetics companies often implement supporting information technology (IT) infrastructures. Some IT solutions have been commonly used in the cosmetics industry such as Electric Data Interchange (EDI), Enterprise Resource Planning (ERP), Material Resource Planning (MRP), Supply Chain Management (SCM), and Customer Relationship Management (CRM) [1]. Most cosmetics companies are experiencing moderate growth and are looking for the ways to support upcoming increasing volumes; thus, improving its IT infrastructure would enhance customer service and support upon the business expansion [2][5][8]. In order to develop the companies’ IT infrastructure, they need to ensure the employees’ understanding the need to change and its benefits for the companies. Normally, managers and employees are resistance to change for new systems; thus, it is important to ensure that managers and employees have proper training programs prior the implementation. Policy and procedure changes may cause reluctance among employees and lower productivity rates. Managers and leaders must ensure that the corporate culture adheres and supports to the new IT changes [6]. New IT system implementation issues include the lack of e-business skills, lack of understanding and support from employees, deficiencies in back and front-end integration, together with maintaining compatible technologies with current alliances or partners [7]. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 197–202, 2011. © Springer-Verlag Berlin Heidelberg 2011

198

M. Koblentz and J. Wei

The objective of this paper is to develop a set of implementable business-tobusiness solutions for the cosmetics industry has incurred as well as analyze the independence that the industry with IT related business systems. The remaining of this paper is organized as follows: Section 2 develops a strategic value chain based on e-business analysis. Section 3 decomposes these activities involved in the value chain into a set of implementable B2B items. Section 3 presents a benchmark for the top ten companies to analyze the pattern of the B2B implementation. Lastly, the paper concludes with discussions on how to use these implementable items with recommendations.

2

Strategic Value Chain Model

Based on Michael Porter’s value chain [7], this paper presents an e-value chain model for the cosmetic industry based on five primary and four support activities. These activities involve information technological advancements and information systems. They are designed to increase efficiency and effectiveness of the company’s core ebusiness activities together with improvement of the products and services value. 2.1

Primary Activities

Inbound Logistics – Technological advancement for inbound logistics includes an inventory control management systems such as a reorder point (ROP) system. Just In Time (JIT) inventory systems and inventory database systems are also the useful tools to reduce the level of the inventory. When the company’s inventory decreases, its warehouse costs also become lower. By cutting these costs, the company can use financial resources to produce more benefit for its shareholders. Operations – Technology can contribute to make operations more efficient. Computer-aided design (CAD) technological system has shortened the new product development phase from years to months. Electronic data interchange (EDI) is useful to transfer electronic documents or business data from one computer system to another computer system without human intervention; thus, it makes transactions and communications more speedy and accurate. Those value added systems will be necessity in order to bring more revenue for the cosmetics industry in the future. Outbound Logistics – For the outbound logistics, most of the companies’ retail stores use the point of sale (POS) system. Early electronic cash register (ECR), that has very limited function and communications capability, is mostly replaced by POS system that provides businesses with the capability to retain and analyze a wide variety of inventory and transaction data on a continuous basis. Another advanced technology is radio-frequency identification (RFID). RFID is innovated for the purpose of identification and tracking, which are more advanced technological system, yet still expensive to implement. Marketing and Sales – The companies’ website, e-mail, newsletter, press release and information portals are among the top e-business technologies used to generate more value to the company. Most cosmetics products are consumable goods; thus, the company’s website availability is particularly important for the reorders among its

Business-to-Business Solutions for the Cosmetic Industry

199

repeated customers. Many companies distribute their newsletters including the special deals and web coupon codes to promote more online sales. Also, the company’s website has a role to inform new items and provide continuous information on products. Technology is necessary to offset the future losses from competition [7]. Thus, technology becomes necessity to stay in the competition. Service – In order to serve better for its customers, many companies utilize their website to maintain customer satisfaction and provide the cosmetics products information and feedbacks. Many online stores have hassle-free return and exchange policy to increase customer satisfaction. Customer Relationship Management (CRM) system is a widely implemented system for managing a company’s interaction with customers, clients, and sales prospects. This technology is useful to organize, automate, and synchronize the business process for sales activities, marketing, customer service, and technical support. CRM system’s main objectives are to find, attract, analyze, and win new customers and reduce the costs of marketing and client services [3]. Firm Infrastructure -- E-mail system, internet and intranet, and EDI systems have contributed to improve the communication efficiency between the members of the company’s value chain. Recently, Enterprise Resource Planning (ERP) system that is an integrated computer-based application is widely implemented in many companies. ERP system manages internal and external resources, such as tangible assets, financial resources, materials, and human resources to facilitate the flow of information between all business functions [6]. ERP system eliminates redundant works within companies and improves its efficiency and effectiveness. Human Resource Management – Human resource management uses the technology mainly as hiring and training its employees. Companies also use the website for posting new jobs and providing continuous education such as eLearning system. Email and internet technology become very useful tools to shorten the communication time and decrease the costs. Technological advancement is an integral part of the HR process and creates value through implementation of supporting activities. Technology – Technology affects every aspect of the e-value chain. Technology created the value for the customers and companies by decreasing costs, saving time, and increasing a company’s effectiveness. Technology is the key components of the successful companies in many industries in the future. Procurement – Technology advancement also added value for procurement activity. Online ordering system reduced the time and created quicker process. Also, EDI system linked companies to suppliers, manufactures, distributors and retailers for providing effective communications and transactions.

3

Business-to-Business Solutions

Table 1 below describes the B2B applications that can be adapted in cosmetics industry. The B2B mainly focuses on commerce transactions between businesses, such as between suppliers, manufactures, wholesalers, and retailers via Internet.

200

M. Koblentz and J. Wei Table 1. Business to Business Solutions for the Cosmetics Industry

4

E-Business Applications in Top 10 Cosmetics Companies

Table 2 below indicates the top 10 cosmetics companies and its progress level of implementation with the proposed E-business applications. The top 10 companies are ranked based on the annual sales in 2009 in the US. All these companies are publically-traded toiletry and cosmetics companies. In Table 2, a checkmark was given if a company implements a particular B2B solution. The percentages were also shown to analyze the B2B implementation patterns. B2B E-Business applications include E-procurement, E-mail, store-to-store inventory look-up capabilities, order tracking system, electric payment system, multilingual service, extranet interface with partners, product information management, RFID technology, and E-commerce website for wholesale. All companies have implemented at least half of applications out of 10. The companies that have higher rate of implementation are P&G, L’Oreal, Unilever, Estee Lauder, and Johnson & Johnson. Applications that have lower rate of implementation are order tracking and RFID technology. RFID technology is commonly used for more expensive items such as computers rather than cosmetics products. Because of its high cost of implementation, RFID technology still has not have cost advantage in the cosmetics industry.

Business-to-Business Solutions for the Cosmetic Industry

201

Table 2. Business to Business Applications in the Top 10 Cosmetics Companies

PNG A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 Total

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

LOR

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

UNI

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

AVO

BEI

EST

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

6 60%

7 70%

9 90%

SSD

KAO

JNJ

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

6 60%

7 70%

9 90%

HEN

✓ ✓ ✓ ✓ ✓ ✓ 6 60%

Percentage (%) 100% 100% 60% 50% 100% 100% 100% 100% 0% 60% 77%

Note: PNT- Procter & Gamble, LOR - L’Oréal, UNI – Unilever, AVO – Avon, BEI – Beiersdorf, EST - Estee Lauder, SSD – Shiseido, KAO - Kao, JNJ - Johnson & Johnson, HEN – Henkel.

From this analysis, some companies have more advanced technologies implementations compared to the others; however, all companies recognize that the implementation of technology is a crucial aspect in the cosmetics industry. The cosmetics industry’s main products are obviously beauty and cosmetics products; however, in order to perform and serve better for their customers, cosmetics companies also need to focus on improvements of technological advancement and IT infrastructure. Thus, implementation of E-Business applications will become more important in the future.

5

Conclusions

Cosmetics industry, as well as the many other industries, is largely dependent on technology in every aspect of its value chain and customer chain. Since those technologies can add significant values for its daily operations, utilizing technology advancement E-Business application implementation can be a company’s core competency. On the other hand, lack of advanced technology can easily set back the company’s position in the competitive market. Porter’s value chain model categorizes technology as one of the supportive activities; however, all of main activities also use some technology in its operating. Technology is no longer just a supportive activity; it is now one of the company’s bloodline. The wise use of technology can have positive effects on a company’s operations as increasing revenue, cutting cost, providing better training for its management and employees, and increasing customer satisfaction [4]. The poor use of technology can affect those

202

M. Koblentz and J. Wei

areas negatively and hurts the company’s trusts and reputations. Technology is the key of the growth for the companies in cosmetics industry. Future research will be conducted to develop the Business-to-Consumer, Businessto-Government, and Business-to-Internal solutions. Moreover, more companies will be added for checking the e-business implementation patterns, and provide recommendations for the cosmetic industry.

References 1. Au, K.F., Ho, D.C.K.: Electronic commerce and supply chain management: Value-adding service. Integrated Manufacturing Systems 13(4), 247–254 (2002) ; from ABI/INFORM Global (retrived November 28, 2010) 2. Bare Escentuals, I.N.C.: - Form 10-K - February 24 (n.d.). Internet FAQ Archives (2010), http://faqs.org, (retrieved November 28, 2010), http://www.faqs.org/sec-filings/100225/BARE-ESCENTUALSINC_10-K/ 3. CRM strategy checklist: Planning for CRM and customer service success (n.d.). CRM Call Center information, news and tips - SearchCRM.com (2010), http://SearchCRM.techtarget.com/feature/CRM-strategychecklist (retrieved December 1) 4. Sanger, D.E. (n.d.). The New York Times - Breaking News, World News & Multimedia. The New York Times - Breaking News, World News & Multimedia (2010), http://www.nytimes.com/ (retrieved November 10) 5. Sephora Weekly Specials (n.d.), http://Sephora.com/promo/beautybonuses/index (retrieved November 28, 2010) 6. Shields, M.G.: E-business and ERP rapid implementation and project planning. Wiley, New York (2001) 7. The Economist Intelligence Unit, The e-business value chain: winning strategies in seven global industries. New York, NJ (2000) 8. Wei, J., Van Der Ende, L., Lin, B.: Customer-Focused E-Business Model for the Oil Industry. The Journal of Computer Information Systems(2009); from ABI/Inform Global (retrieved on November 29, 2009)

Online Crowdsourcing in the Public Sector: How to Design Open Government Platforms Giordano Koch1, Johann Füller2, and Sabine Brunswicker3 1

Chair of Innovation, Technology & Entrepreneurship at the Friedrichshafen Institute for Family Businesses, Zeppelin University, 88045 Friedrichshafen, Germany [email protected] 2 Department of Strategic Management, Marketing and Tourism, Innsbruck University School of Management, 6020 Innsbruck, Austria, CEO Hyve AG, Munich, Germany [email protected] 3 Competence Center Innovation Management, Fraunhofer Institute for Industrial Engineering, 70569 Stuttgart, Germany [email protected]

Abstract: The trend towards “open innovation” has revitalized firm’s interest in tapping into external innovation sources. Firms purposively open their business models to connect internal and external ideas, and to co-create value with partners and users. Internet-based crowdsourcing and co-creation platforms have changed the way how firms implement open innovation. They allow new participatory problem solving and value-creation processes. However, the current discussion on open innovation has hardly touched upon the public sector. This paper investigates if crowdsourcing platforms can be applied in the governmental context, and under which conditions. Results show that crowdsourcing may generate strong interest among citizens and may serve as source of new high quality input. However, our findings also indicate that design principles derived from open innovation projects in the corporate world may not be directly applied in the governmental context; they need to be adjusted and complemented. Keywords: open innovation, crowdsourcing, virtual co-creation platform, design principles, public management, open government.

1 Introduction So far, research on open innovation, distributed and participatory problem solving has focused primarily on the corporate world. In a corporate context the ultimate objective of open innovation is to create valuable offerings for a firm’s customers, to profit from investment’s into innovation, and to improve an individual firm’s economic performance [1, 2, 3]. While this research indicates that “unknown” outsiders can constitute an important source for innovation and value-creation, there is hardly any research dealing with the public sector and public organizations dealing with production, delivery and allocation of goods for citizens [4]. Following the origins of A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 203–212, 2011. © Springer-Verlag Berlin Heidelberg 2011

204

G. Koch, J. Füller, and S. Brunswicker

the word “public”, “public” refers to matters pertaining to people of a community, a nation, or state. “Private”, in contrast, is set apart from government as a personal matter [5]. In the lights of this difference, we claim that the public sector and governmental problems constitute an important area of research on distributed and open problem solving, what we call “open government”. Recent case examples on problem solving activities in the public sector undermine the need for openness in governmental processes. The “Stuttgart 21” initiative in Germany - a large urban development and construction project - is just one example that highlights the limitations of closed problem solving activities. The controversial discussions on the decisions to replace the existing railway station with a new infrastructure mostly located underground exemplify that state agencies should consider citizens as users of public services, who wish to be actively involved. Indeed, treating citizens as customers should be an important objective of public management reforms [4]. To bridge this gap, the following paper tackles the question whether or not internet-based crowdsourcing platforms can be applied to the public sector and under which conditions.

2 Conceptual Foundations: Online Co-creation and Crowdsourcing in the Corporate Context In the open innovation model firms purposively use inflows and outflows of knowledge to accelerate internal innovation, and to expand markets for external use of innovation, respectively [6, p.1]. When opening up to external influences, firms can involve a range of different external actors ranging from suppliers, research organizations, universities [7], customers and users [8, 9]. In fact, the potential impact to be gained from interaction with customers and consumers (and users who may not be customers) during the innovation and value-creation process has widely been recognized. It highlights that consumers and users should not be perceived purely as “value receivers” but also a “value generators” and “co-creators” [10]. However, only when external sources such as consumers and users are willing to engage in co-creation projects and are enabled to share their creative ideas, honestly state their preferences, and comment on existing concepts, valuable contributions that lead to significantly better results can be expected. A compelling and enjoyable cocreation experience is considered as an important success factor and essential prerequisite for valuable external input to co-creation projects [11, 12, 13, 14, 15]. A carefully designed co-creation platform providing engaging and immersive virtual environments is a critical factor in co-creation projects. Consequently, researchers dealt with the question of how to successfully design co-creation and crowdsourcing platforms that ignite users’ interest in participation, allow for creative collaboration, and help to build lively social networks [11]. For example, Nambisan and colleagues [12, 14, 15] suggested that experience in virtual environments subsumes four components: (1) the pragmatic experience, (2) the sociability experience, (3) the usability experience, and (4) the hedonic experience. Based on these dimensions, Nambisan and Nambisan (2008) suggest a set of practices such as the establishment of rating systems, social interaction tools, participant recognition systems, exclusive forums, process transparency, brand fests, and clean technical flows, companies should consider when designing virtual platforms [15].

Online Crowdsourcing in the Public Sector

205

Füller (2010) introduced a virtual co-creation framework from a participant’s perspective addressing five dimensions: 1) content; 2) process; 3) people; 4) motives; and 5) personal characteristics that should be considered when designing a co-creation platform [16]. Based on this framework he presents practical recommendations to design co-creation and crowdsourcing platforms concerning the offered tasks, intensity and extent of interactions, kind of multimedia-rich environment, interaction among participants, offered incentive as well as preferred interaction partners (Table 1). Table 1. Overview of key design principles of crowdsourcing platforms Design Principle Tasks

Description Provide tasks which differ in kind as well as level of complexity. Enable participants to take on different roles such as designer, evaluator or networker.

Intensity and Extent

Allow consumers to engage more often and on a continuous base. Motivate consumers to leave their fingerprint on the platform.

Tools and Multimedia-rich Environment

Provide supportive and empowering contexts. Provide an immersive but simple-to-explore environment.

Interaction among Participants

Offer platforms which encourage intense interaction among participants. Allow relationships to be established and a community to be built. Social networking functionality enriches the communication between participants. Connection to existing social networks allows the leveraging of existing relationships Give direct and honest feedback Offer additional monetary compensation or prizes related to the performance of the participants. Offer a branded platform which allows direct interaction with the company’s developer’s team.

Incentives

Partner

These co-creation frameworks and design principles for crowdsourcing platforms used in the corporate world – mostly by profit oriented companies - may offer valuable hints also for the design and management of co-creation platforms in the public sector by governmental agencies and public administration. However, they may not be directly applicable. Governments and public administration differ significantly from companies, as they face different challenges and duties. Concepts, which are relevant and get discussed in this context, are the approaches of [17], good public governance [18], civic education, and performance management [19]. Hence, it is not only a discussion on giving citizens more “customer voice” but also a discussion on how to educate the citizens, to enable possibly all of them (reach), and to establish a lasting and bidirectional communication process between the citizens and the political institutions and politicians in order to ensure democratization. As mentioned above, little is known if and how crowdsourcing platforms can be applied in the governmental context and how those should look like. Thus, our study aims to contribute to a better understanding and close this gap of knowledge.

206

G. Koch, J. Füller, and S. Brunswicker

3 Open Government Platforms in Practice: Results from a German Action Research Project 3.1 The ‘Aufbruch Bayern’ Case Study and Its Research Design We have chosen the “Aufbruch Bayern” open government initiative to analyze our research question. The initiative represents a unique open government project due to the scope, the functionalities of the platforms, the number of participants and also the output of the open government initiative (http://www.archiv.aufbruch-bayern.de/ start.php). As this project represents an “outlier” case, we chose a single case study research. As we had access to the platform, we were able to collect rich data and observations. This data provided the basis for an in-depth case analysis. We choose an action and participatory research strategy [20]; it leverages a combined approach – participatory action research [PAR, see 21]. The review of existing research on crowdsourcing shows that there is a lack of understanding on the role of crowdsourcing in the public sector. In addition, a more detailed understanding on the appropriate design and management is required. Participatory research design supports the researcher to develop and test design principles. It creates direct insights and reflections in practice. In line with action research principles, this research went through the following major phases: Planning, acting and reflection of actions. In the planning phase the problem of the open government project was identified. Then, we designed the crowdsourcing initiative based on existing findings on how to design crowdsourcing and co-creation platforms. In the action phase, the research team worked in close collaboration with the representatives of the state agency and directly interacted with the participants of the crowdsourcing initiative. During this phase, observations and experiences of the individual researchers were highly important. After completion of the acting phase, the actions were reflected. The reflections and in-depth analysis of data collected during the action phase support the adaption and refinement of existing design frameworks and principles concerning the specifics of crowdsourcing and co-creation in a governmental context. 3.2 Data Sources and Data Collection In July and August 2010, the Bavarian State Government, announced an online participation platform. With the publication of the “call for participation”, Bavarian citizens were asked to suggest their ideas, concepts and best practice cases related to three main policy areas: Family, education and innovation. The duration of the initiative was eight weeks. We implemented the crowdsourcing with an online community platform, and thus we had access to a large set of data collected via the publicly accessible platform. Data such as logfiles (visits, amount and form of activity per user), user data (usernames, sent and received messages on profiles) and information on contributions (number of received comments, number and average of received evaluations) were exported from the platform system data base (MySQL). Data such as comments, contributions and ideas were thoroughly analyzed and reflected by the research team. Observations and direct interactions with participants provided a further important source of information.

Online Crowdsourcing in the Public Sector

207

3.3 The Design of the Crowdsourcing Platform and Evaluation In our case study we designed and implemented the crowdsourcing platform in a structure process. We relied on existing findings and design principles [15; 16]. We built upon the co-creation framework and design principles introduced by Füller (2010) when designing the crowdsourcing platform [16], and further extended and enriched it. To cater for the specifics of governmental matters and problem solving, and to better analyze our research question, we chose a very comprehensive design approach. The implemented design is presented and discussed in the following: • Purpose and goal of the participation platform: The aim of the open government platform project was clearly specified. The online participation platform “Aufbruch Bayern” was the first project of a triennial campaign aiming for systematically integrating the citizens with the means of stat-of-the art communication tools, such as crowdsourcing and social media. It was set-up as a contest. The goal of the initiative was to bring the Bavarian State Government and the citizens closer together by collaboratively discussing and evaluating ideas, concepts and best practices. The concept of participation platform enabled citizen to (1) deal with political questions and challenges, (2) to develop an understanding of the given restrictions of political decision making, and (3) thus to potentially reduce the disenchantments with politics, as well as (4) to increase the perceived empowerment. Moreover, the platform was designed to address and finally reach a variety of groups, which are typically very hard to reach by the traditional communication channels, such as print media, radio or television. By applying a comprehensive recruiting strategy both, the politically interested citizens as well as adolescents, emigrants, and academics which are typically hard to get access to, were reached. • Functionality and tasks of the participation platform: In order to involve these different groups, which differentiated in their interests as well as their skills and cognitive capacity, the online participation platform offered a variety of different functionalities and tasks. Besides the rather complex task of proposing and contributing a well-developed concept, the participants were asked to upload single ideas or best practices. In order to enable participants to take on different roles, such as the above mentioned idea contributor, evaluator or networker, the platform supported the upload of different supplemental materials like written concepts, photos or even documents. Furthermore, the platform allowed participants differentiating between comments on the idea and messages on the pin wall. The latter feature can be seen as a very useful tool in order to enable the communicators and networker to keep the community vivid. It allows participants interacting independently from one’s own contribution. Members were also invited to vote for already existing contribution with an easy “thumb-up” or “thump-down” mechanism. They were also able to comment to the designs of their fellow participants. Due to the fact that all content could be explored without a registration a large number of silent visitors were able to observe the ongoing discussion (400,000 page visits versus 2,094 registered users). This supported educational objectives and awareness creation of the open government platform.

208

G. Koch, J. Füller, and S. Brunswicker

• Intensity and extent: To address intensity and extent, community members were regularly updated about latest developments on the platform. Community managers implemented the following measures to increase intensity and extent: (1) newsletters to reengage “sleepers” into an active community role; (2) an automatic news feed, when somebody commented or evaluated your idea, (3) features of the “latest participants” and “latest ideas” on the front page, which typically has the highest click-rate. These activities helped to engage visitors immediately. • Tools and multimedia-rich environment: The online platform was equipped with a comprehensive portfolio of different social media tools. A twitter channel, the “share” functionality with different social networks, and a separate Facebook fan page, ensured that user generated content could be spread easily and could trigger a “viral” process. Indeed, we observed that participants were highly engaged and also skilled in the production of “creative materials”. At the same time, they manage to drive the distribution of information due to their high web 2.0 literacy. Multimedia enrichment and interconnectivity supported the social and informal exchange of information among users across different communities and via social media, so called word-of-mouth [22]. • Interaction: From research on platforms used in a corporate context we know that people participate in crowdsourcing platforms simply because they can interact with other platform members who share similar interests [13]. When we analyzed the click-behavior of visitors and members on the platform, it was obvious that platform visitors are more interested in other members than in content. They typically explored the community members first before they looked into the ideas or concepts. To address this interaction aspect, the platform was designed to encourage intense interaction among participants, and to allow relationships and community building. Users could write messages to other members, and an up-todate social media connection allowed, encouraged and enabled social interaction. The fact that overall about 450 ideas were shared in other social networks highlights the interest in interacting with members on the platform as well as with private network friends of registered users. • Incentives: We implemented different incentive mechanisms to motivate citizens to participate. Besides intrinsic motivation of individuals, non-monetary extrinsic motives spurred the participation in the crowdsourcing initiative. The three best contributions per category were awarded exclusive prices and the chance to be part of the next governmental declaration. This selection process was executed by the jury (members of the state government), which considered the community vote in their decision. In order to motivate interested participants who lack the skills for contributing an idea, the activity of participants on the platform such as voting and commenting was also rewarded with non-monetary prizes. It is worth pointing out that profiles of politicians were accessible and also accessed very often. Indeed, numerous questions were addressed to politicians. Apparently, the direct exposure and ability to interact with politicians represents an effective incentive.

Online Crowdsourcing in the Public Sector

209

• Partner: In the private sector research suggests offering a branded platform which allows direct interaction with the company’s developer team or even prominent people, such as the star designer or CEO [16]. We transferred this design principle to the public sector and invited well known politicians to the platform, to discussions on the twitter channel and to the Facebook fan page. Overall, the platform was clearly branded as online platform of the Bavarian state government. As mentioned above, we implemented additional functions, features and management processes to cater for the specifics of governmental problem solving. They complement and expand existing crowdsourcing frameworks and design principles: • Community management: We considered community management as a crucial design factor as open government projects usually aim for a sustainable development of an active community. Thus, we emphasized community management, its visibility and also its transparency when designing the platform. Indeed, the “Aufbruch Bayern” project team implemented a very transparent and open communication and community management strategy with precisely defined rules. The team managed to motivate community members to be involved in the management of the community and decision making. This strategy should help to reduce the risk of conflict and to increase the legitimacy of the community management. Observations during the case study and analysis of the contest data also suggested that successful community management requires self-confidence and eventually the courage to resist complaints from members. • Combining online and offline events: Another fruitful tool to recruit and to drive community growth was the combination of online and offline events. The underlying logic is quite simple. By selecting highly engaged participants and giving them the possibility to meet in “real life” and further develop the ideas and concepts of the platform, especially with the support of politicians, the political institutions gave participants the respect and recognition they deserved. Our observations and analysis of data indicate that such events have a tremendous impact on the activity and motivation of the community. The challenge is to successfully connect the offline event and the platform. The “Aufbruch Bayern” initiative had a special subpage where pictures, comments and ideas of the workshops were gathered. During the offline event the community managers posted live on Facebook and used the twitter channel actively. Afterwards the workshop participants were encouraged to spread their thoughts and experiences into the community. The discussion above highlights that our design of the open government platform embraced design principles known from the corporate world and additional complementary design aspects. The platform design was comprehensive and took into consideration the specifics of “Aufbruch Bayern”, a public initiative.

4 Concluding Discussion: Successful Design and Management of Open Government Platforms Overall, the crowdsourcing project was well received and showed a high participation. Indeed, it generated a strong interest among Bavarian citizens. 2,094

210

G. Koch, J. Füller, and S. Brunswicker

participants registered on the platform in order to contribute 750 ideas, concepts or best practices and generate approximately 1,540 pages of content. The community provided 10,932 community evaluations, and 6,342 contributions to the discussion. This indicates that citizens showed a high interest in contributing to the crowdsourcing initiative. Over the period of eight weeks participants spent 760 working days on the platform which sums up to more than 364,800 minutes of residence time on the platform. From phone calls, emails and postings we know that participants spent much more additional time at home developing and discussing their ideas within the family or own circle of friends. This high commitment of time indicates the high interest of citizens to be actively involved in governmental problem solving activities. It shows that citizens should not be treated “just as tax payers” or receivers of public services but rather as active participants and problem solvers in a governmental context. The fact that more than 400,000 viewers grappled over the period of two month with the interactive dialogue-platform undermines the impact of the crowdsourcing initiative in terms of awareness creation. An analysis of the feedback data collected in the post-project survey supports this argument. On average, participants evaluated the design of the platform positive both in terms of functionality and “innovation”. While descriptive figures on the quantity of user generated content highlight a high activity, awareness towards and interest in the initiative, we also looked into the quality of the “output”. In our case study, our observations suggest that user generated content had a positive and constructive connotation. Although there was some criticism, comments were not destructive or even insulting. In addition, officials of the Bavarian State Government stated that ideas and concepts showed a high quality. They also assessed the innovation potential of these ideas and concepts as high. When implementing the action research project, we built upon Füller’s (2010) design principles for crowdsourcing in the corporate world [16]. Indeed, our reflections indicate that these design principles also matter in public problem solving. However our study also revealed that existing frameworks applied in the private sector do either over- or undervalue some design aspects, or completely lack important ones. Table 2 highlights our adapted design principles suggested for the public sector. Summarizing the identified results, we can state that design frameworks derived from research on open innovation platforms and crowdsourcing in the corporate world can support the development and implementation of crowdsourcing in the context of open government. However, for some design dimensions of Füller’s (2010) framework we identified deviating emphasis and slightly different means to best design the respect dimension [16]. In addition, we added two new dimensions, namely “community management” and “offline-events”, which should additionally be considered in future crowdsourcing projects in the public sector. Overall, our research proposes first ideas for an adapted framework to design and implement open government projects. These design principles need to be examined more rigidly in future empirical research. At the same time, they are of high practical value as they guide the design of future open innovation platform projects.

Online Crowdsourcing in the Public Sector

211

Table 2. Adapted design principles of Open Government crowdsourcing platforms Design Principle Tasks

Intensity and Extent Tools and Multimedia-rich Environment Interaction among Participants

Incentives

Description Since political issues are typically very complex and rather difficult to explore, we observed that tasks should be designed in a visual manner so that they can be addressed with photos or stories. The dimension intensity and extent did not show any specifics in terms of platform design but should be taken into account seriously. The “Aufbruch Bayern” case showed that citizens are very cautious to shelter their private sphere when they participate in open government projects. The virtual environment has to ensure this additional requirement. In a political context, the number of lengthy messages and comments was extremely high; Feedback and comments via “thumb-up” or “thumb-down” were hardly used; they don’t seem to be that attractive in governmental matters. Long an intense discussions were the dominant mode of interaction Citizens are often politically or societal motivated. These motives should be considered in addition to the recommended incentives.

Indeed it was considered to interact with a trustworthy partner. The participation of known politicians was important We added community and event management as further principle to Community the framework as it turned out to be crucial for success. Even it may Management not be seen a platform feature it has to be considered as community managers should have access to to special features in order to guide, steer, and manage the community. Interesting features are mechanisms to connect similar ideas with each other, or to mark ideas and members in order to get back to them. Statistics like incoming and outgoing links provide further interesting information for the management. Combining Offline & We further added the dimension offline & online events as policy making is characterized by in-depth and lengthy discussions on Online Events complex problems and topics. We found that a combination of an online platform with offline events (e.g. workshop with politicians) serves as a (1) motivator for participants; (2) multiplier in the context of participant recruiting; and (3) enhances the quality of the content due to offline workshops Partner

References 1. Chesbrough, H.W.: A better way to innovate. Harvard Business Review 7, 12–13 (2003) 2. Chesbrough, H.W. (ed.): Open innovation. The new imperative for creating and profiting from technology. Harvard Business School Press, Boston (2006) 3. Chesbrough, H.W.: In: Chesbrough, H.W., Vanhaverbeke, W., West, J. (eds.) Open innovation: Researching a new paradigm, pp. 1–12. Oxford Univ. Press, Oxford (2006) 4. Hilgers, D., Ihl, C.: Citizensourcing: Applying the Concept of Open Innovation to the Public Sector. The International Journal of Public Participation 1, 67–88 (2010) 5. Perry, J., Rainey, H.: The Public-Private Distinction in Organization Theory: A Critique and Research Strategy. The Academy of Management Review 2, 182–201 (1988)

212

G. Koch, J. Füller, and S. Brunswicker

6. Chesbrough, H.W.: Open business models. How to thrive in the new innovation landscape. Harvard Business School Press, Boston (2006) 7. Fabrizio, K.R.: Absorptive capacity and the search for innovation. Research Policy 2, 1–13 (2009) 8. Brockhoff, K.: Customers’ perspectives of involvement in new product development. Int. J. Technology Management 5/6 (2003) 9. von Hippel, E.: Sticky information and the locus of problem solving. Implications for innovation. Management Science 4, 429–439 (1994) 10. von Hippel, E.: The sources of innovation. Oxford University Press, New York (1988) 11. Franke, N., Piller, F.: Value creation by toolkits for user innovation and design. The case of the watch market. Product Innovation Management, 401–415 (2004) 12. Nambisan, S., Sawhney, M.: A buyer’s guide to the innovation bazaar. Harvard Business Review 6, 109–118 (2007) 13. Nambisan, S., Baron, R.A.: Interactions in virtual customer environments: Implications for product support and customer relationship management. Journal of Interactive Marketing 2, 42–62 (2007) 14. Nambisan, S., Baron, R.A.: Different Roles, Different Strokes: Organizing Virtual Customer Environments to Promote Two Types of Customer Contributions. Organization Science 2, 554–572 (2010), http://orgsci.journal.informs.org/cgi/content/abstract/21/ 2/554 15. Nambisan, S., Nambisan, P.: How to profit from a better virtual customer environment. MIT Sloan Management Review, 53–61 (Spring (2008) 16. Füller, J.: Refining Virtual Co-Creation from a Consumer Perspective. California Management Review 2, 98–122 (2010) 17. Moore, M.: Creating Public Value: Strategic Management in Government. Harvard Business School Press, Boston (1995) 18. Stowers, G.N.L.: Becoming cyberactive: State and local governments on the world wide web. Government Information Quarterly 1, 113–114 (2000) 19. Holzer, M., Kloby, K.: Public performance measurement. An assessment of the state oftheart and models for citizen participation. International Journal of Productivity and Performance Management 54, 517–532 (2005) 20. Checkland, P., Holwell, S.: In: Kock, N. (ed.) Information Systems Action Research, pp. 3–17. Springer, Boston (2007) 21. Whyte, W.F. (ed.): Participatory Action Research. Sage, Newbury Park (1991) 22. Kozinets, R.V., Wilner, S., Wojnicki, A., de Valk, K.: Networks Of Narrativity: Understanding Word-of-mouth Marketing In Online Communities. Journal of Marketing 2 (2010)

An M-Pill Framework in the Electronic Healthcare Nien-Chieh Lee, Hi Tran, Albert Yin, and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected], [email protected], [email protected]

Abstract This paper aims at developing a mobile pills framework in the electronic healthcare by using mobile information technologies. Specifically, an electronic based framework is developed to show how to mobile information technology and information system can be adopted in mobile pills. Then, a set of usability solution items are developed based on this framework. A prototype was created to show the real implementation of an m-pill system with these important features. The findings from this paper will be helpful to managers when making decisions on m-pills development. Keywords: m-pills, electronic healthcare.

1

Introduction

Some electronic (e-) pharmacy prescriptions have been used in the healthcare system. However, mobile prescription on the nationwide scale still does not widely used. In order to offer a new way to patient, getting prescription refill quickly and conveniently, this paper will develop a new framework that will assist pharmacy in saving patient’s time, helping doctor to track doctor shopping, and preventing misread medication names. There are some proposed concepts on electronic prescription. The e-prescribing system benefits over traditional handwritten prescription. However, the system does not send electronic prescription to the patients’ mobile devices. The e-prescribing system suggests that printing out the prescription and hand it to patients, or doctor can send it directly to the pharmacy. This system has limitation as patients may not want to buy prescription immediately or want to find the pharmacy by comparing prices and selecting the lowest price [2]. Numerous articles have been written on electrical records and mobile health [1] [2] [4] [5] [8]. People are embracing and adopting the implementation of mobile technology because mobile technology brings tremendous benefits over traditional paper record. However, there is a lack of research on mobile pills. The purpose of this paper is to develop an m-pills framework with usability features. Specifically, in Section 2, a framework was developed to show the adoptions of mobile technologies used in pills. A set of features were decomposed from this framework. A questionnaire is developed to illustrate important features. In Section 3, a prototype of an m-pills system is developed to implement these important features. Section 4 presents discussions and conclusions. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 213–218, 2011. © Springer-Verlag Berlin Heidelberg 2011

214

2

N.-C. Lee et al.

M-Pill Frameworrk

M-pill framework is a mob bile prescription system framework that not only preveents doctor shoppers but also saves patients’ time in getting medication. Figuree 1 illustrates this framework. It I provides functions such as (a) Doctor uses PC to loginn to m-pills, (2) m-pills generattes authentication code including saving a copy in the mpills Database and sending the code to patient’s mobile device, (3) Patient uuses mobile device to conduct seearch for medication via m-pills, (4) Patient places an orrder on medication, and (5) Pharrmacy receives the order & transmits the estimated pickk-up time to patient’s mobile dev vice.

Fig. 1. M-Pill Framework

In Fig. 1, the doctors creeate their account with m-pills so that the doctors can use their username and passwo ord to log in to m-pills to generate electronic prescriptiions for their patients. When an a electronic prescription is generated through m-pills, mpills saves a copy of thee prescription generated and then it assigns a unique authentication code for each h electronic prescription and sends that authentication code to the patient’s mobile device. Then, the patient can log g in with his own user name which will be made up off the combination of the patien nt’s last name and last four digits of his social secuurity number. For example, Smiith2345; and password will be the authentication code for the first time the patient logs in. Once the patient is logged in to m-pills, he w will change the password to wh hatever password he wants. The patient will use the saame username (combination of last name and last four digits of social security numbber) and the newly created passsword to log in the next time. He then can view the list of prescriptions he has with the most recent prescription on the top of the list. T The patient then selects the deesired prescription. Then, m-pills input the authenticattion

An M-Pill Framework in the Electronic Healthcare

215

code corresponding to the selected prescription into m-pills and do a search on the particular type of medication prescribed. The patient is also asked to enter the search criteria to narrow the search by indicating his preferences for the pharmacies i.e. cheapest price, nearest pharmacy or pharmacy loyalty. M-pills will display the search result by listing the prices for the medication and the corresponding pharmacies according to the preference indicated by the patient. When patient makes his choice, he will be prompted to pay. The patient can either pay then with his credit card through m-pills or pay later when he goes to pick up the medication at the pharmacy. M-pills will also display an option letting the patient indicate whether the medication will be picked up by himself or by another person such as his spouse. If the patient indicates that another person will be picking up the medication for him, a message will display reminding the patient to tell that person who will be picking up the medication for him to bring ID with him as the pharmacy will need to verify the person picking up the medication. When it’s time for the patient to get refills on the medication, the patient can simply enters the authentication code again onto m-pills. Then, m-pills will display the number of refills the patient has left. When there is no more refills allowed for a patient, m-pills will display a message saying “maximum number of refills reached, see doctor.” M-pills is going to keep the expired prescription for a few years just in case the patient is unaware or forgot that his prescription has expired. To prevent the patient from abusing the refill i.e. trying to get another prescription for the same medication from another doctor, a doctor is going to do a search on the patient’s name in m-pills when he logs into m-pills. If the patient already has an unexpired prescription, m-pills will bring that situation to the attention of the doctor by displaying a warning message saying “Patient XXX already has an existing prescription, do you still wish to prescribe one?” The doctor then can decide whether or not he still wants to issue a prescription to that particular patient. The further decomposition of flows in Figure 1 produced a set of features that are presented in Table 1 below. Table 1. Decompositions of information flows in m-pills

Feature 1 2 3 4 5 6 7 8 9 10

Definition Doctor’s username & password Doctor’s endorsement Patient’s info & medication info Authentication code Formatted prescription info Encrypted authentication code Patient’s username & password Patient’s verified username & password List of patient’s prescriptions Current price info

216

N.-C. Lee et al. Table 1. (continued)

11 12 13 14 15 16 17 18 19 20 21 22

Updated price info Price info of a particular type of medication Selected prescription Search result Order info Estimated pick up time Patient’s phone number Verified estimated pick up time Location of pharmacy Verified Doctor’s username & password Amount due Order confirmation

These developed features in Table 1 were given to several medical doctors who used e-pills before for validation. Some features were modified and added based on the comments from these subjects. The final table was sent back to these subjects for verifications again. There is no further modification.

3

Prototype of M-Pill Business System

Based on the features developed in Table 1, a prototype m-pills system was developed. The m-pills system is a mobile prescription system that not only prevents doctor shoppers but also saves patients’ time in getting medication. The nine attributes of a system are components, interrelated components, boundary, purpose, environment, interfaces, input, output and constraints. The components of our system are doctor’s PC, patient’s mobile device, pharmacies’ databases, m-pills web server and all of the components are interrelated to each other. The boundary of the system is that the system can only provide electronic prescriptions, not medical records. The purpose of the system is to facilitate a means for patients to obtain his prescribed medication more conveniently using his mobile device and also to prevent doctor shopping. The environment for the system is doctor and pharmacy acceptance, government support and funding, and patients’ support. The interfaces for the system are the doctor’s PC, patient’s phone, pharmacies’ databases. The inputs to the system are the prescription the doctor creates through m-pills; the patient’s preference i.e. cheapest price, nearest pharmacy or pharmacy loyalty for the medication search; the pries for the prescribed medications from different pharmacies and the pharmacy location. The outputs of the system are authentication code, the pharmacy location as a result of the patient’s preference, the filled prescription to the pharmacy, payment options for the patient. The constraints on the system are that the patient must have a mobile device with internet capability, and that the doctor and pharmacy must also have access to the internet. In the system analysis phase, the process modeling such data flow diagrams were used to construct users’ required features in Table 1.

An M-Pill Framework in the Electronic Healthcare

217

In the m-pills system design phase, we focused on database design and user interface design. A relational database model was created using Microsoft Access based on the m-pills entity relationship diagram in Figure 1. Designing the user interface is the most important aspect of the m-pills system. There are two limitations of mobile devices: “limited attention span and the device constraint.” [7] They recommend that all interface designers need to focus on these limitations. The scholars favors simple and easy to use interface. Wu, Shu, and Lin [8] also find out that “compatibility, perceived usefulness, and perceived ease of use significantly affect healthcare professional behavioral intent.” As a result, a good interface design would solve all these issues when designing the m-pills prescription. Some key features of the m-pills system are user-friendly, meaningful functionality, and fun-to-use. Details of these features are auto text complete, windows-like display, and help menu. The Visual Basic.net was used to design and implement menu driven interface and Microsoft Access for relational database. The system was built targeting to offer userfriendly interfaces as well.

4

Discussions and Conclusions

The current paper developed a mobile pill system by considering the usability features. It first developed a conceptual framework for m-pill, and then breakdown the m-pill information flows from this conceptual model to determine what features need to be concentrated on. The findings from the current research indicate that developing user-friendly m-pill features is crucial to the success of m-pill system development. There are major findings in the current study. First, in the current research, we are trying to build the gap by trying to utilize the mobile technologies so that physicians can access patient’s medical history anywhere. Physician can access and update patients’ record instantly, reducing the times to enter the same information twice. The system also prevent misread medication name because handwritten prescription sometimes is hard to read. Meanwhile, m-pills system saves patients’ time to get medication [3]. Second, the developed usability features in the current paper help an m-pill development company when consider distinguishing its system from its competitors, and enhance the user-friendliness for mobile pill interfaces’ design. Third, m-pill systems have usability limitations including limited size, display window, processing power, and bandwidth, comparing with e-pill system [6] [7]. The m-pill system development company needs to gain popularity by considering these features. Fourth, m-pills systems have security concern, lack of technical support, and organizational support [4]. The systems lacks of technical support and organizational support because mobile healthcare are fairly new. One limitation is that this prototype does not have evaluation. Future study is to evaluate this prototype this developed m-pills system, and investigate the importance of factors influencing m-pill adoptions by developing a survey so as to quantitatively measure these features.

218

N.-C. Lee et al.

References 1. Coolest Products of 2010. Health Management Technology 31(12), 8–12 (December 2010) 2. E-Prescribing on the Rise. Chain Drug Review 32(6), 77–87 (2010) 3. Lee, Y.E., Benbasat, I.: Interface Design for Mobile Commerce. Communications of the ACM 12(46), 49–52 (2003) 4. Lu, Y., Xiao, Y., Sears, A., Jacko, J.: A review and a framework of handheld computer adoption in healthcare, vol. 74(2), pp. 409–422 (2005) 5. Medvedeff, D.: Early Experiences in e-Prescribing. Health Management Technology 24(12), 12–36 (2003) 6. Tarasewich, P.: Designing Mobile Commerce Applications. Communications of the ACM 46(12), 57–60 (2003) 7. Wei, J., Ozok, A.: Development of a Web-based Mobile Air Travel Ticketing Model with Usability Features. Industrial Management and Data Systems 105(9), 1261–1277 (2005) 8. Wu, J., Wang, S., Lin, L.: Mobile computing acceptance factors in the healthcare industry: A structural equation model. International Journal of Medical Informatics 76, 66–77 (2007)

Online Design Discussion Sites: Emerging Resource for Creative Design Moushumi Sharmin and Brian P. Bailey University of Illinois at Urbana-Champaign Department of Computer Science Urbana, IL, USA {sharmin2,bpbailey}@illinois.edu

Abstract. Online design discussion sites are a popular networking platform drawing thousands of designers from all around the globe. These sites show promise as emerging design resource by enabling designers to learn about design techniques and tools, share ideas and to seek (receive) critique to help refine their ideas, and to learn from other designers’ experience. In this paper, we report results of a study that aims to understand designers’ experience and activities in online design discussion site and how they relate to design. We have analyzed member profiles (N=12164), posts submitted by one-time posters (N=2854), and post-contents (N=1004) for a month and conducted follow-up interviews (N=5) to investigate what roles such sites play in supporting design. Our findings reveal that while designers find such sites extremely promising, ineffective representation of ongoing activities, lack of visibility of contribution, and difficulty in finding needed information hinder (even discourage) participation. We propose actionable implications that can lead into better site design. Keywords: Creative Design, Design Community, Online Design Discussion Site, Design Critique.

1 Introduction Online discussion sites provide a place for informal conversation for people with similar experience [1] as well as offer opportunities for building and maintaining relationships. Online discussion sites serve as effective sources of communication [2], information gathering and entertainment [3]. Online design discussion sites (ODDS) are a small but important subset, attracting many designers especially students and freelancers - offering an informal environment for communication and collaboration. Design researchers consider creative design to be a social process [4, 5] and creative outcomes are attributed to the discussion and collaboration among designers as it promotes divergent thinking [6]. ODDS thus show promise in supporting design creativity by allowing designers to share and receive critique on their ideas from a diverse group of designers, making design sociable. For many freelance and student designers who don’t have access to a design team or colleagues to assist in the design process, ODDS are the only source of critique. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 219–228, 2011. © Springer-Verlag Berlin Heidelberg 2011

220

M. Sharmin and B.P. Bailey

Due to the popularity of the online networking and socializing sites, a large number of researchers have focused on understanding collaborative content-creation [7, 8, 9] and users’ socializing behavior in this setting [10]. In addition to the facilities that any online discussion site may provide, ODDS present an opportunity to share early ideas and receive critique from a diverse group of designers – allowing the designer to consider different perspectives and thus making these sites valuable resource for creative design. We believe research focusing on understanding designers’ experience and activities in ODDS and how these ongoing activities relate to design will lead to better design support sites - enhancing the utilization of ODDS as a resource for creative design. We extend research in this area by taking a first step towards this goal – aiming to offer insight about current practices and to identify areas that need to be addressed to increase the effectiveness of such sites. In this paper, we report results from a study that investigated activities, contents, and user experience in a popular ODDS aiming to understand the role of such sites as a design resource. Our study offered several important insights - (1) requests for critique and information about design techniques account for 70% of the total activity; (2) novice (mainly student) designers utilize these sites to seek critique and to learn about design techniques and tools; (3) expert designers (mainly freelance) utilize these sites to find collaborators and to promote their work; (4) critique and information requests receive notable number of responses within a short period of time (on an average, 7.8 and 5 replies respectively within the same day); (5) contribution is highly skewed with only 6.1% members contributing 83.28% of total content; (6) contribution is a function of members’ need, rank, and experience with the site; and (7) members’ experience varies as a function of involvement with the site and the information provided on their profiles. We also offer actionable implications that can improve the design of ODDS.

2 Related Work Internet facilitates information sharing using various systems including online networks [11]. Online discussion sites are successors of online bulletin boards which created the culture of mass communication [1,12]. Bulletin boards and early discussion sites concentrated on information sharing and problem solving in specific domains such as sports, politics, religion, technical hot lines, etc. With the rapid increase of Internet users, discussion sites became a part of the online culture [13]. Recent years have seen tremendous growth in the popularity of online design discussion communities (ODDS) (e.g., core77, highend3d, about.com). ODDS show potential to support creativity in design by providing a platform for exchanging ideas, refining ideas based on critique from a diverse group of designers through open communication and collaboration. However, ODDS look exactly like any other discussion group and provide little support for identifying the ongoing design activities - hindering the effective utilization of these sites as a design resource. Little is known about the designers’ needs and experience surrounding ODDS. In our research we take a first step towards filling this gap by offering insight about existing practice and identifying areas that deserve attention.

Online Design Discussion Sites: Emerging Resource for Creative Design

221

Researchers examined the pattern of participation in the online setting, especially sites that create encyclopedic material [9], provide entertainment resources [7, 8], and/or offer social networking capabilities [10]. Work on online movie databases examined factors that influence contribution [7]. Considerable amount of work also focused on the contribution pattern in Wikipedia [14, 15, 9]. While both entertainment and content-creation communities try to motivate users to contribute by creating an opportunity for self-satisfaction, ODDS additionally offer an opportunity to enhance the quality of designs by receiving critique on ongoing designs from a diverse group of designers. The difference in ongoing activities and users’ expectation makes it difficult to directly apply lessons from other communities to encourage effective participation in design discussion sites. Our work attempts to extend existing research on online communities by offering deeper understanding about ongoing activities and factors that influence designers’ experience in these sites.

3 Methodology The purpose of our research was to gain insight about activities and motivations for participation in ODDS and whether and how they relate to design. Our study combined both quantitative and qualitative research methods to investigate the activities and learn about user experience of core77, a thriving online design discussion site [16]. For our study, we have chosen core77, but we believe that our findings are generalizable to other ODDS. We have collected and analyzed member profiles (N=12164) along with post history (rank, number of posts, number of replies) to examine participation pattern and to identify factors that influence members’ experience in the site. We have also analyzed all contents posted in a month (N=1004, 177 threads receiving 827 replies) to investigate types of posts and their relation to different design activity. Additionally, we have analyzed all posts (N=2854) submitted by the one-time posters (members who posted just once) to identify underlying factors, if any, that motivated the only post but hindered or discouraged further active participation. To better understand members’ experience, we have conducted semi-structured interviews with five core77 users who have varying level of involvement with the site - ranging from administrator to semi-active users. See table 1 for a sample of the interview questions. Table 1. Sample of the questions asked during the interview Participation What factors influence your participation in the site? Does your participation pattern change with time? What factors influence this change? Motivation What motivates you to contribute in the site? From your perspective, what are the potential benefits of active participation in the site? Challenge From your perspective, what are the main obstacles that hinder active participation?

222

M. Sharmin and B.P. Bailey

4 Study Results 4.1 Post Categories – Conversational, Design Technique and Information, Informational, and Critique Designers participate in ODDS for a variety of reasons, including informal conservation, query about design tools and techniques, seeking critique about ongoing ideas or just sharing information about design events. In core77, there are 21 different topic categories from which we identified 12 most active categories, such as general design discussion, sketching, portfolio, etc. in terms of number of posts and views. Analysis of the posts indicated that posts related to critique and design technique dominate 10 of the 12 most active categories. Overall, requests for critique and question regarding design techniques and tools accounted for over 70% of the posts in the 12 category studied. This clearly indicates the predominant utilization of these sites as a resource for creative design as opposed to a place for informal networking. We have analyzed the post-contents to identify types of posts and whether and how they relate to different types of design activity. Content analysis revealed that posts can be broadly categorized into four types – (i) Conversational: Informal experience sharing - targeted to initiate discussion (e.g., good examples of “design language”); (ii) Design Technique and Information: Query seeking information on tools and techniques (e.g., How do I do good transition surfaces?); (iii) Informational: Posts containing links of design events, jobs, blogs, or articles (e.g., design job websites); and (iv) Critique: Posts seeking or providing critique on a design, often including a link to the project or portfolio or images along with a brief description of the project (e.g., feedback please...maybe a critique here and there). Figure 1 presents a screen capture of one such category indicating that different types of posts get represented in the same manner providing no information about the type or content.

a

b

c Fig. 1. Posts from one category of core77. Posts can be categorized as a) conversational; b) informational; and c) design techniques and tools, but the current interface represents all these similarly, providing no feedback to the members about the post types.

Online Design Discussion Sites: Emerging Resource for Creative Design

223

Table 2. Post categories and activities surrounding posts by one-time posters in core77 Category Design Tech. and Info. Conversational Critique Informational

Average (Reply) 5.06 12.23 7.82 1.58

Stdv (Reply) 10.58 18.86 7.90 2.48

Average (View) 2210.26 5458.22 2844.88 1168.11

Stdv (View) 4584.07 13570.98 2620.85 1009.13

Our interviewees expressed that not all designers are interested in all types of posts and consider it difficult to find posts of interest. Similar finding has also been reported in [17] for online Q and A sites, where some users preferred to look at informational vs. conversational posts. The high cost associated with finding posts that demand attention and can be benefitted by a members’ response limits the utilization of such sites. As a result, to ensure effective participation, it is utmost important to distinguish between different types of posts. While categorization and filtering of posts based of types will allow the members to find posts of interest effectively, categorization solely based on content is difficult and often imperfect as the original and subsequent posts can fall into multiple categories. Providing option for indicating the post type along with content-based categorization may lead to better representation of posts. 4.2 Pattern of Participation: A Function of Need, Rank, and Expertise Posts are prompted to satisfy information need: 75% of one-time posters posted on the same day they have become registered members. This behavior may occur due to the fact that users are “required to register” to post. 60% of the one-time posters were seeking information on specific design tool or technique while 28.6% engaged in conversational topics. Interestingly, only 5.6% of the one-time posters were requesting critique – emphasizing members’ tendency to become familiar with the site before posting a critique request. See table 2 for the community response for posts submitted by one-time posters – reflecting that even new members receive significant number of replies and views from the community. Interestingly, 30% of these one-time posts also replied to other designers’ posts, indicating the presence of passive participation from a significantly large number of users. Ranking as an ineffective mechanism to encourage contribution: Like many existing online communities, Core77 uses a ranking model to categorize its members. The current ranking model utilized is as follows: Administrator, Moderator, Full-self Realization, Step four, Step three, Step two, Step one, and then members without any rank. See table 3 for a rank-based distribution of members and their contribution. Moderators and administrators not only maintaining a clean and welcoming environment but also are the most active users (23% posts are contributed by them who are 0.13% of the total members). Most of the other members take a semi-active role, posting when needed. Positions such as moderators and administrators are usually awarded to members who have high rank (calculated based on number of posts). While ranking is used as a mechanism for encouraging participation, the current ranking is proved ineffective as 93% users in the site have no rank, with 39.5% never posting anything and an additional 23.5% posting just once. High effort required for finding thread of interest is reported as one of the main reasons hindering contribution.

224

M. Sharmin and B.P. Bailey Table 3. Contribution statistics of members in core77

Rank Category

Post (Required)

Post (Avg.)

Administrator Moderator Full selfrealization Step four Step three Step two Step one No rank Spammer Banned

>=600

3275.0 1608.3

200~599 100~199 50~99 20~49 <20 -

943.1 340.6 136.9 68.7 30.2 2.0 0.00 0.19

Member (% of Total) 0.02 0.11 0.19 0.57 0.75 1.40 3.06 89.5 3.19 1.2

Post (% of Total) 7.35 15.6 16.2 17.6 9.3 8.7 8.4 16.7 0.00 0.02

Member (cum %)

Post (cum %)

0.02 0.13 0.32

7.35 22.95 39.15

0.89 1.64 3.04 6.1 95.6 98.8 100

56.75 66.05 74.75 83.15 99.85

Fig. 2. Distribution of rank-wise daily posts

Time of involvement with the site and expertise influences participation: In core77, 93% of all daily posts are replies from members who are involved with the site for a long time and are highly ranked members of the community (see figure 2). While a majority of the new members post on their first day, only 0.005% of the daily post is submitted by the new members. Expertise of the members also plays an important role in the participation. In core77, expert designers rarely request critique on their own design, but offer critique and/or information on design to novice members. While novice designers mostly seek help, expert designers almost always offer help. Interview revealed that expert designers utilize other medium for seeking critique – mostly sending e-mails to friends whom they found and build relationship through the site. Expert designers also consider providing critique as a way of repaying the design community. The quotes from two interviewees nicely summarize this behavior: “If designers are just going into the industry, I feel more apt to share information with them as I don't want them to fall into the same holes where I fall on to…I try to respond to requests for design feedback as much as possible.”

Online Design Discussion Sites: Emerging Resource for Creative Design

225

“I started using core77 as a student and I was blown away with the idea of being able to interact with real designers. I like the fact that as a professional, I am accessible to students and can give back what I've earned over the past decade in a way that is easy to access and fully documented.” 4.3 Motivation for Participation Finding mentors and collaborators: Student and freelance designers, who have limited access to expert designers, utilize ODDS to contact expert designers in their area of design. The site provides opportunity to create relationship with expert designers and find mentors and collaborators in an informal environment. Freelance designers often mention ODDS as the place for meeting other designers, finding mentors and collaborators for future projects. However, while long-time members find mentors or collaborators by judging the interaction history, new comers in a site have no information about how to search for a potential mentor/collaborator. One designer commented: “If you are looking at sites, where you see people again and again, it’s the perception that you grow about the people, what type of materials they post, how is the quality of information, how do they help, etc. to see how he will be able to mentor me personally. Ask questions and see how the designers respond. (But) background and expertise information about designers would be great as then there will be experts to whom you can go to.” Access to design resources: Members utilize the site to receive critique from a diverse group of designers. Additionally, the site features articles and events of latest and greatest happenings in the world of design. Designers view it as a one stop place for their informational, emotional needs, nicely summarized by a designer: “We joined to discuss among us, but it was a useful source for getting feedback, professional people give you advice, you can get news about things that interest you, there are groups that discuss overseas jobs, recruiters also participate in this site. It is a place where you can get all sorts of news and entertainment.” Experience sharing in a less competitive environment: Designers become users of online sites, primarily to interact with designers who share similar experiences but could not be reached by any other medium. ODDS also provide an environment where designers can share ideas without the stress of constant evaluation and peer pressure, which can enhance design creativity [5]. One designer stated: “Online forums create a much more accessible, less formal community than attending local IDSA or AIGA events where the designer is often with people he/she is directly competing with. The forum can be anonymous and less threatening.” 4.4 Influence of Profile Information on Member Activity and Experience While race, gender, and social status can highly impact the interaction of participants in offline communities, in online setting it is likely to have less impact. However, users’ self-portrayal in an online setting influences her networking and communication behavior. Lampe et al. discussed how inclusion of certain information in user’s Facebook profile enhanced their friend-making capacity [18]. In core77, members can

226

M. Sharmin and B.P. Bailey

add (optional) information such as website, occupation, and avatar in their profiles. The average number of replies received is 23.2 vs. 4.3 for users who provided additional info vs. no information. While there is no option for providing information on race, gender or social status, our interviewees expressed that gender and affiliation play a role to gain more attention. One designer commented: “Better school usually gives a certain amount of respect to posters. Women are usually given special attention because they are rare (in ID at least).” In core77-like setting, it is difficult to verify user profile and racial and gender association. In-depth analysis is needed to learn how these information impacts members’ experience with the site. 4.5 Lack of Control over Audience Forces Interaction to Be Not So Informal As a medium, ODDS preserve all user interaction history and it is possible to track down members responsible for posts by any other member and at any given point. This makes participants conscious about their interaction as they can be held responsible for a comment that they made 10 years ago when they were unfamiliar with the setting. This everlasting memory of the site hinders fluid participation as it is impossible to imagine target audience [14] and participants have no control over it. Allowing participants to define their own privacy model (e.g., Facebook allows users to define who can view their profile, posted materials, etc.) may encourage more designers to participate more as they will have control over their posted materials.

5 Design Implications and Discussion 5.1 Create Awareness of Posts of Interest Though there are high level categories that guide users to post their requests under a specific topic, each of these categories contain a number of subcategories and get numerous new posts every day. The overhead of finding threads that require help and would benefit from a reply is extremely high. This leads to inefficient distribution of efforts where some posts get more responses while others get overlooked. One solution can be routing posts to different users based on their area of expertise. This routing can be done automatically by collecting data from a user’s profile and by analyzing previous posts and interaction history. 5.2 Provide Activity-Based Summary of the Categories Like almost all ODDS, core77 interface provides a linear list of discussion categories with an icon indicating new posts. To read a post, users have to go through a series of steps and sometimes several pages before finding anything interesting. Also, to get an idea about the post content, users have to click on the individual posts to see the original and subsequent posts. Additionally, all the posts look similar despite differences in their contents. One possible solution is to provide a visual summary (graphs indi-

Online Design Discussion Sites: Emerging Resource for Creative Design

227

cating new posts and number of replies and views) of the activities of each category in the discussion home page – providing better awareness of ongoing activity. Highlighting keywords from new threads may also reduce the burden on users to search for topics of interest. 5.3 Support Categorization of Posts Based on Content At present all posts are presented under the same label irrespective of the content. However, not all members are interested in all types of posts and categorizing posts such as conversational or critique will benefit members to effectively find posts of interest. Similar finding has been reported by [17] in context of Q and A sites. Providing a default set of tags or keywords such as conversational, critique will allow users to attach these to their posts. Alternatively, content classifiers can be used to suggest tags or keywords to a user for a new post and users can accept or select a tag that best describes the post. 5.4 Provide Awareness of (Quality) Contribution Like many other sites, ranking in core77 is solely based on number of posts - ignoring the quality and impact of the posted material (e.g., an in-depth critique is considered similar to a comment like “great work!”). The ranking system also ignores works such as efforts required to maintain a clean environment (site). Users who contribute quality materials or are engaged in the maintenance often feel demoralized as their contribution is not visible and well-recognized. Making contribution visible will make users aware of their and other’s contribution. Adding an automated activity graph in users’ profile can help in this direction. To reflect the impact and quality of a contribution, users can award points or rank the post and it can be used along with number of posts to determine overall ranking. As contributions will be judged in terms of quality and quantity, users may feel motivated to contribute frequently and it may also improve the quality of contribution. Additionally, providing a summary of top contributors may motivate more users to actively participate as they become aware of the level of effort necessary to achieve higher rank in the community.

6 Conclusion This paper offers insight about activities and designers’ experience and participation pattern in ODDS. The study sheds light on the factors that influence particpartion in these sites – highlighting areas that need improvement. As more and more designers are getting involved in these communities every day, it is important to understand what factors hinder their participation and how we can assist in lowering these barriers. As a new form of creative community there is a lot to learn about the audience of these groups, individual’s perception of privacy and security in this setting and how that can be maintained, and how individuals can best utilize this setting for enhancing design creativity. Ineffective representation of activities and information are considered as one of the major factors that hinder participation. Another major issue observed is the lack of active participation by the majority of the members. To ensure the success of such communities, it is essential that a significant number of members

228

M. Sharmin and B.P. Bailey

actively participate and eventually reduce the burden on the few committed ones who carry the community. In-depth research is needed to understand how to close the gap between passive and active participation and encourage more people to effectively participate. Another area of future research is the role of status and power in this setting and how that affects overall participation.

References 1. Wellman, B., Gulia, M.: The network basis of social support: A network is more than the sum of its ties (1999) 2. Joinson, A.N.: Looking at, looking up or keeping up with people? motives and use of facebook. In: Proc. CHI, pp. 1027–1036 (2008) 3. Nardi, B.A.: Beyond Bandwidth: Dimensions of Connection in Interpersonal Communication. CSCW 14(2), 91–130 (2005) 4. Cross, N., Cross, A.: Observations of Teamwork and Social Processes in Design. Design Studies 16(2), 143–170 (1995) 5. Warr, A., O’Neill, E.: Understanding design as a social creative process. In: Proc. Creativity &Cognition, pp. 118–127 (2005) 6. Roy, R.: Case Studies of Creativity in Innovative Product Development. Design Studies 14(4), 423–443 (1993) 7. Beenen, G., et al.: Using Social Psychology to Motivate Contributions to Online Communities. In: Proc. CSCW, pp. 212–221 (2004) 8. Lange, P.G.: Publicly Private and Privately Public: Social Networking on YouTube. Computer-Mediated Communication 13(1), 361–380 (2007) 9. Viegas, F.B., Wattenberg, M., K., D.: Studying cooperation and conflict between authors with history flow visualizations. In: Proc. CHI, pp. 575–582 (2004) 10. Boyd, D.: Social Network Sites: Public, Private, or What? Knowledge Tree, 13 (2007) 11. Mislove, A., Marcon, M., Gummadi, K.P., Druschel, P., Bhattacharjee, B.: A Measurement and analysis of online social networks. In: Proc. ICM, pp. 29–42 (2007) 12. Whittaker, S., Terveen, L., Hill, W., Cherny, L.: The Dynamics of Mass Interaction. In: Proc. CSCW (1998) 13. Kollock, P., Smith, M.: Managing the Virtual Commons: Cooperation and Conflict in Computer Communities. Computer-Mediated Communication: Linguistic, Social, and Cross-Cultural Perspectives, 109-128 (1996) 14. Cosley, D., et al.: SuggestBot: using intelligent task routing to help people find work in wikipedia. In: Proc. IUI, pp. 32–41 (2007) 15. Kriplean, T., Beschastnikh, I., McDonald, D.W.: Articulations of WikiWork: Uncovering Valued Work in Wikipedia through Barnstars. In: Proc. CSCW (2008) 16. Core77 Discussion Board, http://boards.core77.com/ 17. Harper, M., Moy, D., Konstan, J.: Facts or friends?: distinguishing informational and conversational questions in social Q&A sites. In: Proc. CHI, pp. 759–768 (2009) 18. Lampe, C., Ellison, N., Steinfield, C.: A familiar face(book): profile elements as signals in an online social network. In: CHI 2007, pp. 435–444 (2007)

Social Computing for Home Energy Efficiency: Technological and Stakeholder Ecosystems Marc Torrent-Moreno1, Rafael Giménez1, Regina Enrich1, José Javier García1, and María Pérez2 1

Barcelona Digital Technology Centre, Roc Boronat 117, 08018 Barcelona, Spain 2 Gas Natural Fenosa, Avenida de San Luís 77, 28033 Madrid, Spain {mtorrent,rgimenez,renrich,jjgarcia}@bdigital.org, [email protected]

Abstract. Many initiatives exist with the goal of providing the required tools to improve energy efficiency of households. Encouraged by public administrations, supported by the private industry, and demanded by an environmentally-aware society, several systems are gaining attention, mainly, energy consumption sensors, smart meters, and energy management platforms. The latter pretend to increase their impact and engagement with emerging social networks focused on energy efficiency matters and applying knowledge engineering disciplines. This paper describes the current home energy efficiency ecosystem, outlines some missing pieces and advances required to achieve its potential outcome, and surveys the existing social media platforms.

1 Introduction Carbon emissions are one of the major concerns in our society. We can witness different private, public and coordinated initiatives aimed at improving different energy-related aspects around the globe. A primary example is the ‘climate and energy package’ agreed upon by the European Parliament and Council in December 2008 in order to implement the 20-20-20 targets by 2020, which is encouraging the use of renewable energy sources, enhanced materials and improved energy management systems. In this paper, we explore existing initiatives and systems where social media is envisioned to play a key role in assisting private users in reducing energy consumption in their households. Energy management systems will enable monitoring and recommend more sustainable practices and, according to existing studies, up to 35 million equipped households can be expected worldwide by 2014 [1]. Of course, these forecasts depend on several factors, most relevant being the timing of smart meter rollouts and early adopter engagement. Here, we will focus on the latter and describe the technological enablers that will assist in increasing consumer commitment, which is a major challenge. Indeed, while the social crowd claims to have become greener and publicly demands a more environmental friendly planet, there is a lack of individual adoption of conservation practices and a lack of knowledge with respect to behaviour or carbon footprint. The principles and procedures of social media applied to the energy efficiency scenario can provide the engagement capabilities required to effectively integrate A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 229–238, 2011. © Springer-Verlag Berlin Heidelberg 2011

230

M. Torrent-Moreno et al.

consumers into the service provider strategies. In addition, the performance and reach of these platforms can be enhanced by refining services, contents and interaction according to the user profiling data captured and processed from smart metering devices. The tools provided to customers should supply more information than just how much energy they are consuming, but also how are they doing compared to other customers with the same profile as theirs, whether they are improving and how they can do better by applying personalized recommendations according to their profile, history and preferences. As a result, both user adoption and obtained value from these novel “smart” social media are expected to become major drivers for energy efficiency programs. Further, the deployment of these platforms would benefit all actors in the play. Customers would save money by reducing their energy consumption and by profiting from time-based pricing. Access to information would lead to increased competitiveness between utilities, which in turn would offer improvements in the services offered. Moreover, utilities would see how their energy efficiency and loyalty programs are widely spread through comments from experiences of customers, providing the feedback necessary to evaluate their degree of success. Knowing how customers behave and what their expectations are could help utilities to design market strategies to flatten the load demand curve by shifting peak demand consumption to off-peak demand periods, thereby avoiding investments in new infrastructures to cover sporadic peaks in demand. In this paper we first introduce the home energy efficiency ecosystem and identify the goals of the most relevant stakeholders (Section 2). Then, we describe the technological enablers that are required for the short-term success of energy management solutions and those which are envisioned to play a key role in the aim of developing sound solutions capable of bringing energy savings to consumers and achieving longterm user engagement (Section 3). Finally, we survey the existing cloud-based solutions developed to improve energy efficiency at homes, devoting special emphasis on implemented social computing features (Section 4).

2 Stakeholders In this section we present the main stakeholders involved in the home energy efficiency ecosystem. First, however, we outline the main building blocks of the envisioned system as defined by the market for clarification and further reference. As we can observe in Fig.1, we can identify three main building blocks: the utility is the energy provider (in this paper we focus on electricity, although it could also refer to gas, water, or a combination of any) which distributes electricity and commercializes the service (although companies distributing and commercializing could be different, it is not relevant within the scope of this paper). The utility owns the IT infrastructure required for billing and CRM purposes, i.e., includes all consumers’ relevant data and profile, as well as the communication and energy distribution channels up to the smart meter. AMI (Advanced Metering Infrastructure) is the chosen term to name the smart meter and the communication channel up to the utility which enables remote readings. Further, new generation smart meters, located at the consumers’ buildings, can include an additional communication interface that provides real time readings to the HAN (Home Area Network).

Social Computing for Home Energy Efficiency

231

Fig. 1. Envisioned building blocks in the energy efficiency ecosystem

The energy HAN refers to the sensor network installed at home capable of monitoring relevant parameters of energy consumption (including devices’ energy consumption as well as temperature, presence, etc.) and includes the energy management devices acting as energy hub. The energy HAN should be also capable of obtaining real time readings from the smart meter and contains a gateway to the internet (e.g., through the consumer’s ADSL or 3G router) in order to connect to cloud-based services. Cloud-based services are offered by an energy management platform implemented in the cloud, which extend those implemented in the home equipment. Obtaining realtime and detailed readings from user’s energy consumption, these platforms enable key features such as data analysis and visualization, social computing for direct user interaction, and personalized utility services, among others. Derived from the building blocks introduced above, we can, at first sight, differentiate the following stakeholders in the home energy efficiency field, each of them pursuing their own different goals: Utility. Having the AMI being deployed in many countries (and to be soon in many others) to enable remote readings, utilities see the next step in implementing a direct, bidirectional and constant communication channel with their clients. Their interest in implementing this channel together with encouraging social media platforms is twofold. On the one hand, they can develop a new customer relations model based on better knowledge of the customer, including personalized services and loyalty programs. On the other hand, it will allow the building of more accurate demand estimations, which in turn allow for optimization of wholesale energy purchase. Early alliances with HAN manufacturers and cloud-based platforms will guide the direction of future business models and potential approaches to consumers. In a second phase, utilities pretend to develop better energy management strategies, including demand response programs and pricing policies. Further, these strategies will aim for the incorporation of renewable energy sources as well as the integration of electric vehicles. HAN Provider. Current first-generation energy HAN systems are commonly composed of a clamp sensor, which monitors the aggregated energy consumption at home, and an optional energy management display. Both can exchange data wirelessly and send it to a cloud-based platform through the home internet connection. A dedicated home display offers the advantage, when compared to interacting with a cloud-based service through a computer, smartphone or TV, of being at the optimal place to quickly check the current energy consumption state or tariffs before switching on another appliance and triggering an alarm. Although a simple sensor can

232

M. Torrent-Moreno et al.

already differentiate some consumption patterns in an aggregated curve, e.g., identifying home appliances being used, a better consumption pattern will be acquired and consequently a higher saving potential will be achieved with a larger amount of sensors at the house, i.e., monitoring not only appliances independently but also relevant parameters such as temperature or humidity. Note that these independent sensors can also be actuators and therefore the user (or eventually the energy management unit or even the utility) could remotely control their usage depending on the current power consumption levels and tariffs. HAN providers’ sole objective is the sale/installation of its products, either directly to clients or in partnership with a cloud-based service provider and/or a utility. Cloud Services Provider. We have witnessed the launch of a few cloud-based services in the last couple of years, which are reviewed in Section 4. These can obtain real readings from home sensors or through web forms filled by the user. Developed by energy HAN manufacturers or third parties, their goal is to provide an appealing user interface with engaging features that achieve real energy savings. The most advanced aim is to analyze the user’s consumption patterns and identify behavioural improvements through AI (Artificial Intelligence) techniques. Further, social computing features help engage and assist users with tips from other users, from experts or even automatically from the system. The business model behind these platforms varies, depending on whether they are standalone systems, if they are part of a complete solution including an energy HAN, or if they are deployed in partnership with a utility. Indeed, while as a social network advertising can be a direct source of revenue, the high interest of utilities in acquiring and managing the data (including user conversations) encourages both parties to discuss joint initiatives. Public Administration. In accordance with sustainable planet goals, governments are establishing more concrete actions and policies which are relevant for this field. Although some, such as the US and several European countries, have begun enforcing the adoption of smart meters, their position with respect to additional policies is still unclear. It would definitely encourage the adoption of energy HAN systems and boost the engagement of users if they would take further actions such as enforcing communication interfaces between meters and energy HAN, imposing time of use pricing, subsidizing equipment, and providing incentives for consumption reduction. Consumer. When referring to their energy consumption, users have two main motivations: saving money in their monthly bill and reducing carbon emissions. However, since these savings will not always be very significant or will not be achieved without compromising on daily comfort, easy access and utilization as well as appropriate engagement methods are essential requirements. Currently, early adopters fit the profile of tech-savvy, ecologically-aware consumers and are interested in features such as real-time monitoring and remote access to the energy systems. However, when aiming at the mass market, solutions will have to provide more appealing interfaces, be extremely easy to use, and probably be integrated in a single product. Note that present-day consumers are in control, i.e., they can contract any utility, HAN provider and service that they wish, independently of each other. However, one dealer integrating the complete solution would certainly encourage its adoption by the mass market. According to [1], 81% of these would pay at least $150 for energy management equipment if they could achieve up to 30% savings, and more than 52% are willing to spend $5/month if they can save up to 10%.

Social Computing for Home Energy Efficiency

233

Additionally to those described, the interest of other entities is witnessed which see energy management as a potential killer application for home control systems. These entities, which could help boost energy HAN adoption, are mainly providers of home networking, domotic and security systems as well as telecom operators.

3 Technological Enablers Although the future looks promising and the benefits are clear, we are still some steps away from achieving the maximum potential outcome of a complete solution. Indeed, several ICT building blocks still require advances in terms of standardization or technological enhancements. In the following section we describe the state of the art and current trends of the main ones: namely, smart metering and energy HAN systems, knowledge engineering capabilities, human-computer interaction technologies, and social engagement strategies. 3.1 Communication Interfaces between Smart Meters and the Energy HAN The smart meter is a key component of new AMI deployments. As such, communication aspects from the smart meter toward utilities backend systems are clearly defined and standardized by institutions such as NIST in USA, or by Mandate 441 of the European Commission. Also, although different communication technologies are implemented in existing energy HAN products, such as Z-Wave or HomePlug, ZigBee seems to be the current trend for these networks. With respect to the communication between both systems, i.e., the smart meter and the energy HAN, although a large percentage of current solutions implement ZigBee, it still remains as an unresolved challenge how best to cover multiple scenarios, from small houses to large apartment buildings. Along this line, standardization efforts like those fostered by EU commission mandate 441 and the CENELEC TC 205 working group will further strengthen the “smart metering – HAN” as an enabler for innovative energy efficiency applications and services. 3.2 Knowledge Engineering The transformation of raw data into knowledge supporting the operation of expert systems plays a central role to realize the full potential of the described platforms. Although there’s no a unique vision on the technical approaches to fulfill the lifting from data to knowledge, the concept of knowledge engineering is regularly used to embrace all the involved disciplines. In the following section we have depicted those which are more relevant in the scope of this paper, ranging from traditional data mining techniques to those closer to artificial intelligence such as user profiling and recommendation, with a special focus on the semantic technologies as the state-of-the-art framework for knowledge acquisition. Ontological Modeling. Within computer science, ontologies provide an exhaustive and explicit conceptualization of a given domain that fosters the transfer and processing of information between systems. In the specific domain of the described platforms, the rich domain conceptualization enabled by an ontology offers a powerful framework for the development of semantic recommender systems by increasing the

234

M. Torrent-Moreno et al.

understanding of the user and social dynamics [2]. From the basis of the existing ontologies for user modeling (SUMO, GUMO [3], UbisOntology), social graphs (XFN and FOAF) and social media content (SIOC Core Ontology), modeling the target domain requires an extension of those to correlate specific topics related to energy efficiency such as climate, building characteristics, household appliances and household composition. User Profiling. User profiling techniques provide the basic background for profitable and sound performance of recommender systems. Through user profiling techniques systems can map the interests, preferences and wants of users and match the generated recommendations on a systematic and model-based basis. As previously stated, semantic modeling based on ontologies provides the most powerful framework today to maintain and exploit user profiles. Techniques for user profiling are grouped under two basic models: implicit and explicit. In the depicted scenario user profiling based on customer attributes is also connected to the characterization of the consumption behavior based on consumption data mining, and the identification of proper refinement mechanisms for created models arises as one of the main challenges. Data Mining. As previously stated, the discipline of data mining is highly relevant for systems based on the processing of massive amounts of data such as those described here. In the context of energy platforms data mining is useful to reinforce user models according to the different consumption patterns found in the data. The most common techniques used in data mining are decision trees, nearest neighbor classification, neural networks, rule induction, and K-means clustering [4]. Some open research challenges deal with finding out which works best for profile representation, how to categorize variables, how to normalize them or which is the distance function which defines if two users are similar concerning energy efficiency. Recommenders and Semantic Reasoners. In the described system, recommender engines provide the computational capabilities to select and present information that is interesting to the user. Recommender engines compare stored user profiles to some key attributes in the information to predict the “rating” that a user would assign to an item he doesn’t know. The performance and accuracy of recommender engines is crucial for user engagement in energy management platforms. The integration of context-awareness into recommendations, the enhancement of system scope by adding learning capabilities supporting changes in consumer’s behavior, and the ability to collect data from disseminated and heterogeneous data sources are some of the main open challenges being addressed by research in the area today. Social Media Tracking. The application of information retrieval techniques to social computing provides a great potential for knowledge acquisition from user-generated content, but these techniques must be properly adapted to the specific features of the social data: scalability, real-time, content duplication and lack of structure. As above mentioned, the central corpus of knowledge acquisition methods for social media can be expressed as applied adaptations of some of the traditional information retrieval techniques: data harvesting or web scraping, filtering, opinion and sentiment mining, activity tracking, and pattern detection and learning on similarity metrics [5]. 3.3 Human-Computer Interaction The relevance of Human-Computer Interaction (HCI) as a success factor for social computing platforms is increasing in accordance with the growing complexity of

Social Computing for Home Energy Efficiency

235

contents and activities these systems deliver. Providing an optimal user experience as a tool to elicit of the level of participation is the final goal of HCI in social computing. At the same time, recommender systems such as the ones depicted in this paper also pose some specific challenges to the discipline of HCI. Together with good algorithms, recommender systems must provide usable and intuitive interfaces for presenting suggestions and capturing opinions [6]. The visual representation of reputations and trust levels is also a challenging aspect addressed by research today [7]. In addition to that, the heterogeneity of the devices shown in the energy efficiency scenario also introduces some issues to HCI. Devices with constrained computational capabilities and reduced interface capabilities are highly useful in multiplying the information displays which the system offers to the user, but technologies must be adopted to fit those specific device requirements. 3.4 Social Engagement Although it should not be specifically considered as a technological research area, the strategies for user engagement are a subset of the interaction design with a highly relevant impact on social computing. Modern approaches on user engagement agree on the need for a consistent design of the reputation and reward policies, the development of a shared culture among the platform users and the deployment of a consistent behavior from the platform administration side. Techniques and mechanics inherited from game play are performing today as those displaying best results to raise user engagement. In a phenomenon known as gamification, this approach makes technological platforms more engaging by encouraging desired behaviors. Well-known techniques such as achievement badges, leader boards, or the delivery of virtual currency are in the state of the art of the user engagement through gamification.

4 Current Solutions There are many ways in which the combination of resource consumption data and social networks can be leveraged to influence the behavior of occupants and managers. The use of social media solutions as a way to reach end users and provide valueadded services towards energy efficiency (whether in homes, offices, or any other scenario) is relatively new. The main objective is to empower, engage, educate, and motivate users toward smarter resource usage, both individually and within a community as a whole. In general, these social media-based energy efficiency solutions are based on appealing graphical user interfaces that allow users to introduce energy consumption and specific user profile data, keep track of their energy consumption, or receive valuable feedback based on this information. Integration with other social networks such as Twitter or Facebook is a common mechanism used to foster interaction between users, make data even more visible and accessible, and finally promote the use of the tool within a larger community. At the time of writing, several solutions are either available on the market or nearing launch. The main ones and their distinctive features are as follows:

236

M. Torrent-Moreno et al.

Welectricity is a free stand-alone web service with a community of users from 72 countries, which is based on a dashboard that let users compare their energy consumption with that of others with similar profiles and keep track of their energy efficiency goals as well as other features. Building Dashboard Network is a cloud-based platform released in 2010 by Lucid Design Group, which claims to be the first social network for buildings. It is focused on the use of evolvable tools that promote individual behavioral changes towards community-wide energy and water use and conservation goals. It is currently deployed in commercial, institutional and residential buildings, mainly located in the USA and Canada. Microsoft Hohm is a free online application based on the use of advanced analytics licensed from Lawrence Berkeley Labs and the US Department of Energy, developed to provide personalized energy-saving recommendations to users based on their profiles and consumption data which they provide (or which is obtained through integration with third-party hardware solutions). PeoplePower Energy Services Platform is a software platform based on a broad range of data mining, analytics, and reporting tools, designed to work in conjunction with an embedded hardware platform intended to provide energy monitoring and control capabilities to virtually any electronic device. Google Power Meter is a cloud-based tool designed to be easy to integrate and to use data provided either by utilities, smart meters, or third-party energy monitoring devices. The platform is able to learn the energy consumption habits of a user, as well as to derive community-level patterns used to provide personalized useful tips, engagement, and motivation tools among users. Enerbook is a web-based social network created by the Portuguese company ISA as a complement to ISA’s iMeter in-house solution for energy efficiency monitoring and control. The portal provides iMeter users a set of evolving social networking tools to enable interaction and sharing of experiences. Current solutions comparison. In Table 1 we present a summary and comparison of the main characteristics of the social media solutions analyzed. We focus our comparison in four key areas: energy data source (to provide an idea of how the solution obtains the data used), user-focused features (either personalized tools for a specific user or tools provided to enable social interactions between users), deployment responsibility (whether the service is integrated in the utility’s back-end, or is a standalone service), and the current state of the solution in the market. Although it is too early to identify which will be the most adopted solution, it is clear from the results obtained from first deployments of these solutions that social community tools and the integration with other successful social networks will play a key role in current and future energy efficiency solutions. The use of user profiling techniques, a reliable source of data, and powerful data processing mechanisms will allow these solutions to offer users better functionalities and services. However, there is still a large margin for improvement, particularly in the current knowledge engineering mechanisms that are the heart of these energy efficiency social tools.

Social Computing for Home Energy Efficiency

237

Table 1. Feature comparison of existing cloud-based energy management services including social computing features

5 Conclusions and Outlook Social media platforms are envisioned as key components for the successful adoption and long term engagement of consumers with household energy management systems. Such systems can benefit many actors in the play, from customers to utilities. On the one hand, private users would be assisted by service providers or other peers in the goal of reducing energy consumption, would obtain enhanced and personalized services, and would earn social acknowledgement. On the other hand, utilities would have access to data that would allow them to build loyalty programs as well as implement improved demand response and pricing programs. Currently, several cloud-based solutions exist that aim at monitoring user energy consumption and incorporate social media features. However, these could be further improved with knowledge engineering algorithms which would assist analysis and recommending engines that could, in turn, improve user acceptance and engagement. Furthermore, not only product-based improvements, but also business models and the way consumers will be approached (e.g., which stakeholder will sell and install the energy management system) need to be well defined. The complete service must be easy to contract and use in order to avoid user disappointments which would make it difficult to achieve large penetration rates. Early alliances among stakeholders will define future business models encompassing relationships between utilities, energy HAN manufacturers, and cloud-based service providers. Currently, utilities are very active in stimulating the market and positioning themselves, even though technology may not be ready for the mass market. Obviously, the role of the public administration will be key in the development of this sector, with the ability to encourage it by enforcing communication interfaces between smart meters and HAN, subsidizing energy management equipment, or providing incentives for energy reduction. Further, a second generation of energy management systems is envisioned which will offer automation features to further optimize consumption rates and could be

238

M. Torrent-Moreno et al.

capable of identifying both malfunctioning devices (which are not efficient any longer) or suggest new and more efficient ones. Moreover, other aspects such as the introduction of renewable energy sources and the integration of the electrical vehicle into the house grid may drive the smart energy management system to become an indispensable device in every household. Acknowledgments. The authors would like to thank Greg Gladman for his insightful comments which significantly contributed to the final state of this paper.

References 1. Hatler, M., Gurganious, D., Chi, C.: Smart Energy Homes A Market Dynamics Report. In: Onworld Study, Onworld (2010) 2. Sheth, A., Nagarajan, M.: Semantics-Empowered Social Computing. IEEE Internet Computing 13(1), 76–80 (2009) 3. Heckmann, D., Schwartz, T., Brandherm, B., Schmitz, M., von Wilamowitz-Moellendorff, M.: GUMO – The General User Model Ontology. In: Ardissono, L., Brna, P., Mitrović, A. (eds.) UM 2005. LNCS (LNAI), vol. 3538, pp. 428–432. Springer, Heidelberg (2005) 4. Berthold, M.R., Hand, D.J.: Intelligent Data Analysis. Springer, Heidelberg (2003) 5. Appan, P., Sundaram, H., Tseng, B.: Summarization and Visualization of Communication Patterns in a Large-Scale Social Network. In: Ng, W.-K., Kitsuregawa, M., Li, J., Chang, K. (eds.) PAKDD 2006. LNCS (LNAI), vol. 3918, pp. 371–379. Springer, Heidelberg (2006) 6. Knijenburg, P., Bollen, D., Schmidt-Thieme, L.: Workshop on User-Centric Evaluation of Recommender Systems and Their Interfaces. In: RecSys 2010. ACM, New York (2010) 7. Egger, F.: Interface: How to Maximize Perceived Trustworthiness? In: Proceedings of the International Conference on Affective Human Factors Design, pp. 317–324. London Academic Press, London (2001)

Conversational Lives: Visualizing Interpersonal Online Social Interactions Heng Chen and Aisling Kelliher School of Arts, Media and Engineering Arizona State University {chenheng,aisling.kelliher}@asu.edu

Abstract. The dynamic and distributed nature of interactions on popular social network platforms creates challenges in personally understanding the collective meaning and impact of this disparate activity over time. We propose the development of the Social Reflector application, a dynamic visualization interface for exploring individual social network activity over time. We discuss the design, implementation and evaluation of three visual strategies for representing key facets of social communication interactions - conversation potential, activity rhythms, and interpersonal communications. Findings and results from a preliminary user study with Facebook users are presented and promising future research directions introduced. Keywords: Social Network Visualization, Reflective Interfaces.

1 Introduction Much of our daily reflections, communications and commentary appear today integrated as part of our social media network [7,18]. Social network sites such as Twitter, Facebook and YouTube vary in the consumer interests and communication practices supported, but their key technological features are relatively consistent. As defined in Boyd et al., they allow individuals to 1) construct a public or semi-public profile within a bounded system; 2) share information and ideas to a list of other users; 3) communicate among their list of connections [2]. Our everyday interactions on these sites can be considered as forming a significant part of our digital life-stream, complete with fragmentary statements, media sharing and expressive behaviors. Personally understanding the collective meaning and impact of this disparate activity over time can be challenging, given the dynamic, momentary and forward moving impetus of most social network sites [14]. Social behavior can be understood as the operational effect of two distinct information processing systems: the impulsive system and the reflective system [17]. These dual systems operate in parallel, but the reflective system requires considerably more cognitive processing. Within the impulsive system, knowledge structures arise that “bind together frequently co-occurring features”, whereas in the reflective system, generated knowledge is “accompanied by a noetic state of awareness, which consists of knowledge that something is or is not the case” [17]. Our online social interactions occur across a multi-dimensional space comprising people, artifacts, actions and time A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 241–250, 2011. © Springer-Verlag Berlin Heidelberg 2011

242

H. Chen and A. Kelliher

[11]. The speed and diversity of these interactions lend themselves well to the operations of our impulsive system, but forming more reasoned opinions and understandings about the broader impact of our online social activities requires the integrated consideration of our reflective operations. In this paper, we propose the design, implementation and evaluation of the Social Reflector visualization application. This tool provides three dynamic social network visualizations aimed at supporting reflection and the development of insight about online interactions on social network websites. Our online activities comprise both direct communications and the consumption of messages [3]. In our research, we are interested in investigating perceptions regarding how personal statements and communications are consumed and responded to through reply mechanisms such as comments or ‘likes’. Our visualizations provide users with a dynamic interface for exploring their online data from a distance, in a form that emphasizes the quality and nature of their interactions over time.

2 Prior Work Our work expands and learns from prior research in reflective practice systems, social communication representation and social network visualization. 2.1 Reflective Practice Systems Reflection is described by educational theorist John Dewey as the examination of the basis for a belief [5]. Research within the field of cognitive science suggests that engaging in reflective activities is a fundamental component of the thinking process, vital for decision making and self-learning. Donald Schön studied reflection within the realm of professional practice [16] and defined two types of reflection as occurring during problem solving activities: reflection-in-action and reflection-on-action. Iterating between in-the-moment and after-the-fact reflective thinking supports individuals in becoming more explicit, accountable and revisionary in their everyday lives. Maeno et al. [13] developed a mediation reflective system displaying dialog utterances between mediator and disputants. Results from preliminary studies demonstrated the value of their visualization as a training mechanism for mediation students to reflect on their dialogue in moving from an impasse point to ultimate agreement. 2.2 Social Communication Representation The Sociable Media Group at MIT Media Lab have developed a number of applications for analyzing and representing social communication patterns [6]. Projects such as Coterie, Authorlines and the Loom Project, visualize detected patterns in online conversations, thus allowing participants to better understand the activities and behaviors of their virtual communities [20]. More specifically at the level of the individual, applications such as CrystalChat [19] visualize IM chat history from a personal perspective and provide opportunities for reflection and storytelling. Email exchanges have similarly been studied from an ego-centric perspective. For example, Themail [21] visualized the interaction history between participants in an individual email archive, while Thread Arcs [10] visualized overall conversational threads. Additional

Conversational Lives: Visualizing Interpersonal Online Social Interactions

243

research has focused on analyzing and representing the hierarchical, correlational and temporal patterns [15] in email repositories, while other email visualizations encourage reflection for improving social decision-making [14]. 2.3 Social Network Visualizations Visualizing social networks has a long and rich history, from hand-drawn pictures of social patterns to computationally derived images of massive online networks [8]. In recent years, numerous end-user visualization applications have been developed to investigate the structure and behavior of popular social network platforms including LiveJournal1, Friendster/MySpace [9] and Facebook2. Researchers within Facebook have also developed several applications such as Project Palantir 3 that visualizes Facebook activities happening in real-time and the “The Road to 200 Million” infographic4 which displays the world-wide growth of the Facebook network.

3 Design and Implementation There are an estimated 30 billion pieces of content (web links, news stories, blog posts, notes, photo albums, etc.) shared each month on Facebook, making it a complex multi-dimensional platform that can be observed at many scales. Our Social Reflector application focuses on the ebb and flow of information on a user’s wall, specifically the status updates (posts), comments and likes left by profile visitors. We propose three visual strategies for representing key facets of these social communication interactions - conversation potential, activity rhythms and interpersonal communications. Within the Social Reflector tool, the WordCloud, Clock and Circles interfaces express these patterns respectively.

Fig. 1. Screen shot of the WordCloud interface depicting an individual’s Facebook wall posts, likes and comments as a weighted 2D list

1

http://www.touchgraph.com/facebook

2 http://vansande.org/facebook/visualiser;

http://thomas-fletcher.com/friendwheel 3 http://blog.facebook.com/blog.php?post=41339392130 4 http://www.nytimes.com/imagepages/2009/03/29/business/29face.

graf01.ready.html

244

H. Chen and A. Kelliher

A common property of social network sites is the ability to respond to statements or mediated messages by posting comments or indications of support (e.g. digg, like). The conversational potential of a statement or artifact can be understood by the volume, diversity and value of the responses it generates. The WordCloud interface (Fig.1) presents an individuals’ Facebook status updates5 as a weighted 2D list. In this interface, the font size of each status post indicates the number of associated comments (the more comments the larger the font) and the font saturation depicts the number of associated ‘likes’ (the more likes, the more saturated). Understanding when (i.e. during the workday, only at weekends), how (i.e. browsing news feeds, searching for friends) and why (i.e. communicate with family, personal diary) people use platforms such as Facebook can offer insights into contemporary interpersonal communication dynamics. The Clock interface (Fig. 2) depicts an individuals’ monthly Facebook status posts and received comments in a radial layout [1]. Here, each circle represents a day in the selected month, and posts and comments are represented as icons placed on the circle lines according to their time of posting. The icons are color-coded blue (posts) and orange (comments) to give participants an at-aglance overview of the ratio between their outward expression and their inward feedback reception. The circles are presented using a tree ring metaphor, where the innermost circle represents the last day of the month, while the outermost circle represents the first day. A “reverse” function (reverse the order of the circles) was also implemented to allow closer inspection of densely populated inner circles.

Fig. 2. Screen shots of the Clock interface depicting Facebook data from three users. (a) Data depicts significant inward traffic on a user’s birth-date; (b) Data indicates regular morning posting by a user; (c) Data indicates little interactive communication from a user’s network.

As our online social networks expand and contract, we develop and discard interpersonal relationships by maintaining or ignoring communication interactions. The Circles interface (Fig. 3) allows users to explore the inward and outward flow of communications within their Facebook network. The rings in the visualization depict status updates (inner), comments (middle) and friends (outer) as selectable, interrelated bars. Users can click on a bar in any of the three rings (e.g. a post bar) and any related bars (e.g. comment and contributor bars on that post) will be highlighted, together with general statistical information (e.g. overall % comment contribution of selected friends). 5

http://apps.facebook.com/my-year-in-status

Conversational Lives: Visualizing Interpersonal Online Social Interactions

245

Fig. 3. Screen shots of the Circles interface depicting data from three user categories as detailed in Table 1: (a) Low number of friends and low activity volume; (b) Low-medium number of friends and medium activity volume; (c) High number of friends and high activity volume.

3.1 Implementation The Social Reflector is a cross-platform desktop application developed with Adobe AIR and the ActionScript SDK for the Facebook Platform, using the Facebook GRAPH API to make calls to get and set Facebook data. Users are prompted to authorize the application to allow access to information in their News Feed. The Social Reflector application does not collect or store any News Feed data and individual’s are logged out of their accounts upon application exit.

4 Evaluation We evaluated the design and reflective utility of our application in a multi-activity user study with a diverse group of Facebook users. In the next section, we describe the study participants, methodology and results and findings. 4.1 Participants We recruited 20 Facebook users (12 male, 8 female) for our study. The participants ranged in age from 23 to 59, with an average age of 30. The majority of participants were recruited within a large public university in the United States, with representatives from the student, staff and faculty bodies. The participants derived from a wide variety of disciplinary backgrounds, including Design, Music, Computer Science, Engineering and Media Arts. Additional study participants represented engineers working in industry. At the beginning of our study, participants completed a questionnaire survey examining both their general social communication activities and their specific use of the Facebook platform. Overall, participants ranked phone communication as the primary method by which they communicated with friends and family, followed in order by email, social network services, instant message, Skype, microblog, blog and postal service. In addition to using Facebook, 75% of participants identified as Twitter users, while YouTube and Flickr were used by 55% of the study population respectively. 25% of participants used MySpace, while 25% also stated that they used RenRen, a popular Chinese social networking platform (7 of the study participants were Chinese nationals).

246

H. Chen and A. Kelliher

The average user on Facebook has 130 friends, with 50% of active users logging onto Facebook on any given day6. The participants in our study represented a relatively active group of Facebook users, with 70% indicating they logged on at least once a day. 3 participants indicated they logged on several times a week, 1 checked Facebook several times a month, while 1 other indicated rarely using the service. Our study group also presented a higher than average size friend network, with participants on average having 319 Facebook friends. However, 30% of participants did have fewer than 130 friends, with the smallest network representing 57 connections, in contrast to the largest at 955 friends. In terms of regular activity, 35% of participants stated that they added content to their Facebook wall (e.g. status update) at least several times a week, with 40% indicating they did so several times per month. Participants estimated that their posts on average attracted 3 comments, with 12 being the average maximum number of comments ever received. 4.2 Methodology Participants in our study completed a survey, a two-part comparative activity and a short interview, with an overall study duration length of 40 minutes. The first module of the study, as described in 4.1, comprised a short questionnaire survey, examining participant demographics, their general social communication activities and their specific use of the Facebook platform. The second component comprised a comparative activity session, where participants were asked to indicate their beliefs and understandings about their usage patterns and interactions on Facebook. Participants completed the following four tasks twice, first while viewing their data on Facebook and second while viewing their data using the Social Reflector interface: 1.

2.

3. 4.

Communication Network: we asked participants to complete an ego-centric sociogram diagram depicting the friends/alters of their social network in a series of four concentric circles, placing closer friends nearer the center (Fig. 4a). Sociograms have proved helpful not only in reliably recording network data, but also function as a strong cognitive aid helping participants reflect and think about their social ties and associated activities [4]. Events: we asked participants to name some of the most important events in their life in the past year (e.g. travel, birthdays, achievements). Participants were also requested to indicate if they posted about those events on Facebook, and if so, whether or not they received comments of communications in response. Activity: we asked participants to indicate where and when they regularly use the Facebook platform (e.g. home/work/in transit and time of the day). Activity Distribution: we asked participants to characterize their Facebook activity along a distribution from purely maintaining an online profile (i.e. only adding to your own wall) to high levels of interaction and communication with friends (i.e. extensive commenting and activity on others’ walls). Participants were asked to indicate their answer by marking their location on an activity distribution bar (Fig. 4b).

6 http://www.facebook.com/press/info.php?statistics

Conversational Lives: Visualizing Interpersonal Online Social Interactions

247

Fig. 4. Completed elements of comparative activity forms from two users: (a) Completed sociogram depicting the closeness of members in a user’s social network; (b) Marked activity distribution bar depicting a user’s self-identified level of interaction with members of their social network.

The final study component comprised an approximately 15-minute audio-recorded semi-structured interview, where participants were asked a series of questions about their use of the tool, their perceived value of the approach and suggestions for design refinements.

5 Results and Discussion Analysis of the reported and observed study data indicate the potential of different facets of our approach for certain types of social network users and communicators. While participants were able to successfully use and navigate our tool, some users indicated confusion with particular elements of the design and functionality of the interfaces, indicating the need for additional refinement and improvement. 5.1 General Findings We categorized the study participants along two primary dimensions (number of friends and volume of activity) in order to analyze the usefulness of our approach across diverse types of users. Combining k means and average clustering, the participants were grouped into 4 friend size collections: low, low-medium, medium-high and high taking into consideration both the number of friends of the average Facebook user and that of our study participants. Secondly, we examined screengrabs of each participant’s Facebook activity captured using the Circles visualization, categorizing user activity based on the sparseness or density of the circular representation. Table 1 below indicates the distribution of the network size and communication activity of the study participants. The majority of participants amended or updated their initial responses for each task in the comparative activity, ranging from little or no change to substantial response reconsideration following their use of our visualization tool. The Circle visualization proved most effective in helping participants identify and add forgotten network members or to redefine relationships between members by deleting, moving or connecting them on their depicted sociogram. One participant noted during the

248

H. Chen and A. Kelliher Table 1. Taxonomy of Facebook usage diversity of our study participants

Low # of Friends

Low-Medium # of Friends

Low Activity Volume Medium Activity Volume High Activity Volume

6 (30%)

3 (15%) 1 (5%) 1 (5%)

MediumHigh # of Friends

High # of Friends

2 (10%) 4 (20%)

1 (5%) 2 (10%)

interview that they “spent a lot of time looking at the names of people who wrote things” and at the “breakdown of percentages to see how much people are commenting – that was interesting”. Several participants indicated surprise at the amount (or lack) of communication from some members in their network, with one participant commenting that “the surprise was that there are some people who commented ... like this person contributed 5 percent of my comments, but it’s not somebody I consider very close.” One participant in particular outlined a theory about the possible disconnect between their perception of a communication relationship, versus that depicted in the visualization. “I knew this person was commenting but this is the interesting part: I think the relationship has to be connected with my own activities changing over time. So this person commented in the early part of the year and not the later part of the year. So now I don’t remember him - it’s not because he is not important but he’s not part of my life now. I think somehow there needs to be a connection between these changes in relationship and activities and time...it’s surprising. It’s not consistent in my understanding of how the relationship has changed over time.” Overall, the majority of participants indicated a preference for the WordCloud visualization, with several commenting on its usefulness in identifying posts that attracted interesting conversation (i.e. comments) versus those that their network just seemed to like. For example, one participant noted that “I really like the text cloud a lot because it shows a useful function to me which is just first of all, I don’t have any other way at all to look at all the posts I published, and then I can also see what people really like, or what was less interesting.” Of the three interfaces, the WordCloud visualization appealed and made sense to the greatest number of participants, primarily because of its legibility regardless of the level of communication activity or network size. Over half of the participants placed themselves in a different location on the activity distribution diagram after exploring their data using the Social Reflector visualization. While many of these changes were relatively minor (less then 2 unit change), several were notable in that the participant placed themselves over 4 units away from where they originally indicated. Several users noted their change of opinion as being based on an incorrect assumption such as “I think that I have even less status updates then I thought” or “I didn’t realize that I had a post with 5 comments this year. I guess I’m just not aware.”

6 Conclusions and Future Work In this paper we presented the design, development and preliminary evaluation of the Social Reflector application, a reflective interface that visualizes the conversation

Conversational Lives: Visualizing Interpersonal Online Social Interactions

249

potential, activity rhythms and interpersonal communications of an individual’s social network. Results from a preliminary user study indicate the utility and potential of our approach in instigating reflection and promoting awareness of the underlying patterns of our online communications. Of particular value is the ability to reveal an overall interpretation of aggregated activity by presenting disparate data in one interface (ie WordCloud) and an opportunity to precisely explore interactions with specific network members as a percentage of overall activity (ie Circles). The study also highlighted several limitations in our design, particularly in representing the network structure and activities of highly connected, highly active individuals. We intend to address this issue in future iterations of our work through the use of search and zoom features that will enable varying levels of scaled exploration from whole network to specific members. Creating tighter integration between the three interfaces and the underlying data (e.g. revealing all post or comment content) would also allow users to more thoroughly explore particular areas or patterns of interest. Acknowledgements. We would like to thank the participants in our study for their time and thoughtful commentary and our colleagues in the School of Arts, Media and Engineering for their helpful suggestions and feedback.

References 1. Bergstrom, T., Karahalios, K.: Conversation Clock: Visualizing audio patterns in colocated groups. In: 40th Annual Hawaii International Conference on System Sciences, HICSS 2007, pp. 78–86. IEEE, Los Alamitos (2007) 2. Boyd, D.M., Ellison, N.B.: Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication 13(1), 210–230 (2008) 3. Burke, M., Marlow, C., Lento, T.: Social network activity and social well-being. In: Proceedings of the 28th International Conference on Human factors in Computing Systems, pp. 1909–1912. ACM, New York (2010) 4. Carrasco, J.A., Hogan, B., Wellman, B., Miller, E.J.: Collecting social network data to study social activity-travel behavior: an egocentric approach. Environment and Planning B: Planning and Design 35(6), 961–980 (2008) 5. Dewey, J.: How we think: A restatement of the relation of reflective thinking to the educative process. DC Heath Boston (1933) 6. Donath, J.: A semantic approach to visualizing online conversations. Communications of the ACM 45(4), 45–49 (2002) 7. Ellison, N.B., Steinfield, C., Lampe, C.: The benefits of Facebook “friends:” Social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication 12(4), 1143–1168 (2007) 8. Freeman, L.C.: Visualizing Social Networks. Journal of Social Structure 1 (2000) 9. Heer, J., Boyd, D.: Vizster: Visualizing online social networks. In: Proc. IEEE Symposium on Information Visualization (InfoVis), pp. 32–39 (2005) 10. Kerr, B.: Thread arcs: An email thread visualization. IEEE Computer Society, Los Alamitos (2003) 11. Lin, Y.-R., Sundaram, H., Kelliher, A.: JAM: Joint Action Matrix Factorization for Summarizing a Temporal Heterogeneous Social Network. In: Proceedings of the 3rd International AAAI Conference on Weblogs and Social Media, ICWSM 2009 (2009)

250

H. Chen and A. Kelliher

12. Lin, Y.-R., Sundaram, H., Kelliher, A.: Multi-Relational Characterization of Social Network Communities. In: Furht, B. (ed.) Handbook of Social Network Technologies and Applications. Springer, Heidelberg (2010) 13. Maeno, Y., Nitta, K., Ohsawa, Y.: Reflective visualization of dispute resolution. In: IEEE International Conference on Systems, Man and Cybernetics, SMC 2009, pp. 1698–1703. IEEE, Los Alamitos (2009) 14. Morrison, D., Gooch, B.: ConnectDots: Visualizing social network interaction for improved social decision making. In: Schuler, D. (ed.) HCII 2007 and OCSC 2007. LNCS, vol. 4564, pp. 134–140. Springer, Heidelberg (2007) 15. Perer, A., Smith, M.A.: Contrasting portraits of email practices: visual approaches to reflection and analysis. In: Proceedings of the Working Conference on Advanced Visual Interfaces, pp. 389–395. ACM, New York (2006) 16. Schön, D.A.: The reflective practitioner: How professionals think in action. Basic Books (2006) 17. Subrahmanyam, K., Reich, S.M., Waechter, N., Espinoza, G.: Online and offline social networks: Use of social networking sites by emerging adults. Journal of Applied Developmental Psychology 29(6), 420–433 (2008) 18. Strack, F., Deutsch, R.: Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review 8, 220–247 (2004) 19. Tat, A., Carpendale, S.: CrystalChat: Visualizing personal chat history. IEEE Computer Society, Los Alamitos (2006) 20. Viegas, F.B., Smith, M.: Newsgroup Crowds and AuthorLines: visualizing the activity of individuals in conversational cyberspaces. In: Proceedings of the 37th Annual Hawaii International Conference on System Sciences, p. 10 (2004) 21. Viegas, F.B., Golder, S., Donath, J.: Visualizing email content: portraying relationships from conversational histories. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 979–988. ACM, New York (2006)

CICRO: An Interactive Visual Interface for Crowd Communication Online Masao Ohira, Hitoshi Masaki, and Ken-ichi Matsumoto Graduate School of Information Science, Nara Institute of Science and Technology, 8916-5, Takayama, Ikoma, Nara, Japan {masao,hitoshi-m,matumto}@is.naist.jp

Abstract. As a means for online communication has become sophisticated and diverse (e.g. twitter, SNSs, YouTube, and etc.), a large indefinite number of users actively communicate each other in online communities. Thorough such the communications, they sometimes form the “crowd mind” which is temporally shared values and beliefs among mass users and can affect our society either positively or negatively. In this paper we present a concept of the online crowd and introduce CICRO, an interactive visual interface for the crowd communication. As a result of our experiment, we found that CICRO encouraged 1.7 times more active discussions among users than that of BBS since it provides users with an easy way to understand the existence of a variety of others’ opinions and/or impressions in the discussions. Keywords: Crowd communication, Visual Interface, Online community.

1 Introduction As a means for online communication has become sophisticated and diverse (e.g. twitter, SNSs, YouTube, and etc.), a large indefinite number of users actively communicate each other in online communities. Thorough such the communications, they sometimes form the “crowd mind” which is temporally shared values beliefs among mass users and can work for our society either positively or negatively [1]. Although a small number of authorities (e.g., political party, government, religious organization and so on) and mass media have strongly influenced the formation of the crowd mind among people so far, people are currently able to be proactive in forming it by using internet and information technologies. The crowd mind among people is not static but rather dynamic one. It can be changed over time depending on outside circumstances. A current interface for online communication, however, does not carefully consider the characteristics of such the user group and its communication. Most of online communication tools still provide users with traditional, text-based interfaces which cannot sufficiently capture or help the dynamical formation of the crowd mind. This situation might accelerate the “daily me” [2] phenomena where people prefer to get the information only they are interested in. It might not a democratic way of using the information technology [3]. Based on the findings from the literature of social psychology, in this paper we define such the online communication as “crowd communication” and try to show that A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 251–260, 2011. © Springer-Verlag Berlin Heidelberg 2011

252

M. Ohira, H. Masaki, and K.-i. Matsumoto

the theory of the crowd mind in the physical (real) world can be applied to that formed online. We also introduce an interactive visual interface for the crowd communication, so called CICRO.

2 Crowd Communication Online 2.1 The Notion “crowd” The notion “crowds” were invented through discussions among French social psychologists in order to explain the characteristics of large scale, collective behaviors of people, which had been observed saliently after the Industrial Revolution. Le Bon [1] described that the crowd is a group of people which is psychologically united by loosing sense of self from any cause. Tarde [4] used the notion “public” to describe the crowd clearer. The notion of public means a psychological collective formed among people who are physically distributed but intermediated by media such as news papers. In this sense, the crowd is a temporal group which is formed of massed people in a limited physical space and which acts together toward a certain shared goals. Summarizing these discussions, the crowd can be defined as follows: • • • •

The crowds consists of a number of spatially-massed people, is a temporal group, shares a uniform environment, and is organized by common interests and situations defined by the environment.

The “crowd” in this study is also a temporal group of people which consists of massive online users on a virtual space. Different from the psychological unit (Tarde’s “public”) which is formed by the influences of one-way media such as news papers and TV, an online group shares common interests and situations on a uniform (virtual) environment by using interactive (bidirectional) online media such as chat and BBS. Their behaviors are constantly changing since the online group is affected by the summation of the interactions among the users. In this sense, the notion “crowd” in this study can be interpreted as the same notion of “crowd” in the social psychology, excepting the difference between a physical environment and virtual environment. 2.2 Characteristics of Crowds In the literature of social psychology, the categorization of “crowd” and its formation mechanism has been studied [5, 6]. Table 1 shows the phenomenological classification of collective behaviors of crowds [5]. Roughly dividing crowds into two kinds of crowds (mob: active crowd and audience: positive crowd), the classification seems to succeed to create a consensus on understandings of crowds. In contrast, we created the classification of online crowds as shown in Table 2. As Table 2 indicates, the classification of collective behaviors in a physical environment can be applicable to a virtual environment (online) since we can see that each type of the crowd exists even in a virtual environment.

CICRO: An Interactive Visual Interface for Crowd Communication Online

253

Table 1. Phenomenological classification of collective behaviors of crowds [5] type offensive escape mob material expressive accidental audience intentional

example crowd involving violent behaviors in disturbances crowd escaping from disaster sites crowd rushing to panic purchases and deals to make a profit crowd enjoying to express feelings and impressions at festivals crowd consisting of third parties such as curious crowds crowd watching sports or listening to musical performance

Table 2. Collective behaviors of online crowds (massive user groups) type offensive escape mob material expressive accidental audience intentional

2.3

example user group causing framing on BBSs or blogs user group worrying about the welfare of their family on BBSs user group rushing to online shopping sites to buy incorrectly-priced items user group posting their feelings and opinions to BBSs or blogs user group seeing accidental affairs such as flaming on BBS or blogs user group cheering for sports teams on micro blogs while watching TV

Communication Media for Online Crowds

Through the comparison of physical and virtual crowds, we were motivated to build a new interface for the crowd communication online because a current interface for online communication does not carefully consider the characteristics of online crowds and most of the online communication tools still provide users with traditional, textbased interfaces which cannot sufficiently capture and help the dynamical formation of the crowd mind. As a first step toward creating communication media for online crowds, in this study we develop a visual interface for the crowd communication to meet the following requirements. • It is easy to grape summations of opinions and impressions among users. • It is easy to understand the scale and structure of conflicts between crowds. • It provides users with a means to chime in the discussions among users (i.e. a means to quickly replicate the discussions in the past) for users joining the communication later. In this study, we do not intend to cover all the types of crowds as shown in Table 2, but we would like to support the communication for the “expressive” crowd.

3 CICRO: A Visual Interface for the Crowd Communication This section describes a visual interface for the online crowd communication we call CICRO (Communication Interface for CRows Online). Figure 1 shows CICRO’s interface that visualizes the crowd communication in real time so that users can easily understand the scale of the crowd communication and relationships between opinions of the crowd. CICRO also has a function to replay crowd communications in the past

254

M. Ohira, H. Masaki, and K.-i. Matsumoto

Fig. 1. A screenshot of CICRO. In this situation, users discuss about President Obama while watching a YouTube video. The groups of supporters and opponents are differently colored. Demos are available from http://crowdcom.sakura.ne.jp/hcii2011-demo/.

so that new users can quickly understand the past discussions among the crowd and join the current crowd communication. CICRO is built with Microsoft Silverlight 3 and runs on web browsers installed the Silverlight plug-in. 3.1 Implementation Figure 2 shows the architecture and data flow of CICRO. CICRO consists of three components: CICRO Remark Controller for preprocessing input data (remarks) from users, CICRO Sever for storing the preprocessed data into a XML database, and CICRO Visualization Controller for interactively displaying the data to users. Sending remarks. As shown in Figure 1, users send their remarks to CICRO by inputting the remarks with a keyboard and clicking any position to put them on the visualization area. CICRO Remark Controller takes remark data from users and coordinates data which is defined by the position users clicked. At the same time, it randomly determines colors of remarks on the visualization area according to the coordinates if users click a position where other remarks do not exist. It uses the color as same as existing remarks if users click on the existing remarks. After this, it sends the data to CICRO Server using HTTP. Storing data. After CICRO Server receives the data from CICRO Remark Controller, it convert the data into the XML format using a custom Python script and then stores the XML data into a XML database. Visualizing the crowd communication. CICRO Visualization Controller acquires remarks, coordinates, and colors data in real time from the XML database and stores it into the temporal data store which is always updated to have the most recent data up to 50 remarks. The canvas controller of CICRO Visualization Controller visualizes colored, circular remarks on the visualization area.

CICRO: An Interactive Visual Interface for Crowd Communication Online

255

Fig. 2. Architecture and data flow of CICRO

3.2 Interactions with CICRO Representation of a crowd. In the visualization of CICRO, “crowds” are represented as overlapped remarks with the same color, which mean that users put own remarks onto existing other users’ remarks to express agreements and/or interests with other users’ remarks (i.e., opinions and impressions). If a user disagrees with other users’ remarks or other crowds, she can represent a counter opinion as a different colored remark by putting it on the blank space. Anonymity. In the current implementation, CICRO represents a scale of a crowd not as the number of users but as the number of remarks. If a single user remarks again and again, one crowd can be formed only by the user. In contrast, if CICRO represents it as the number of users by coloring each user’s remarks, each user will be identifiable from other users and it might result in hesitating to remark own opinions and impressions. This is a reason why we implement the way to visualize users’ remarks anonymously. However, in the future we need to study this anonymity and its influence on the crowd communication. Crowd communication in real time. In order to represent the crowd communication in real time, CICRO makes the size of an existing remark on the visualization area smaller little by little, each time a newer remark is created. In addition, CICRO only visualizes remarks up to 50 at the same time to avoid too complicated discussions among users. In the current implementation, 51th older remark is removed from the visualization area to display a new remark. Replaying the crowd communication. For users who chime in the crowd communication, CICRO has a function to quickly watch the communication in the past. Users can randomly skip to any scene with the seek bar in the left bottom of the interface.

256

M. Ohira, H. Masaki, and K.-i. Matsumoto

4 Comparative Experiment This section describes our comparative experiments to evaluate the usefulness of the CICRO’s visual interface in the online crowd communication. 4.1 Overview To evaluate the usefulness of the usefulness of the CICRO’s visual interface in the online crowd communication, we conducted a comparative experiment where 34 subjects in total used CICRO and a common bulletin board system (BBS) respectively. While they watched a Japanese comedy show lively streamed on the internet, they used the two kinds of the communication interface to talk each other about the comedy show. After the session, we asked the subjects to answer a questioner on the usefulness of the two interfaces. We also counted the number of subjects’ remarks in the session and analyzed (1) whether CICRO had an effect on the increase of the remarks or not, and (2) whether CICRO had an effect on the creation of the crowd mind among the subjects or not. 4.2 Procedure Prior to the experiment, we recruited Japanese 34 undergraduate and graduate students (20-24 years old) as subjects from four different universities in Japan. We strongly prohibited them from telling anyone our experiment in order to ensure the anonymity of the subject during the experiment. We then divided them into two groups of randomly-selected 17 subjects. The subjects were asked to use his own computer to participate in the experiment where he could ensure the space without anyone else. The subjects were also asked to watch a Japanese comedy show lively streamed on the internet and to use both CICRO and BBS to freely discuss the show with other subjects who did not know each other. Figure 3 shows the procedure of the experiments. The groups of the subjects participated in two sessions: one session for CICRO and the other session for BBS. In consideration of the order effect, the order of the interface use was counterchanged. Each session spent 45 minutes in total. In order to deeply look at the effects of the interfaces on online communication, we controlled that only the half subjects (8 or 9) of each group could join a session for first 10 minutes in the session. In this setting, we were interested in understanding which interface eased to chime in the discussions.

Fig. 3. Sessions in the experiment

CICRO: An Interactive Visual Interface for Crowd Communication Online

257

During the sessions, we recorded all the subjects’ remarks on the two interfaces and captured subjects’ screens. After the sessions, we asked the subjects to answer questioners on the usefulness of the two interfaces. We also interviewed them based on their answers. 4.3 Results Number of remarks. Figure 4 shows the number of each subject’s remarks in the sessions. The grayed bar is the number of remarks in using CICRO, and the white bar is the number of remarks in using BBS. We can see that the most of the subjects’ remarks in using CICRO was larger than that in using BBS. In fact, the number of subject’s remarks with CICRO was 1.7 times larger than BBS as shown in Figure 5 and Table 3.

Fig. 4. Number of each subject’s remarks in the sessions

Fig. 5. Box plot of subjects’ remarks in total

258

M. Ohira, H. Masaki, and K.-i. Matsumoto

Table 4 shows the result of the questionnaire#1: “does the interface help you understand a variety of opinions and impressions in the sessions?” From this question, we tried to understand which interface is easier for the subjects to grasp others’ opinions and impressions during the crowd communication. The result indicates that CICRO is superior to BBS for that purpose. The Wilcoxon signed-rank test also shows there was a significant difference between BBS and CICRO. Some of the subjects also answered our interview as follows: • “I think one of the advantages of CICRO is that some topics can be discussed at the same time.” • “For colored remarks, I could intuitively (visually) understand responses from others.” Table 5 shows the result of the questionnaire#2: “does the interface help you understand the scale of opinions and impressions in the sessions?” From this question, we tried to understand which interface is easier for the subjects to understand the existence of crowds and its scale. The result indicates that CICRO is also superior to BBS for that purpose. The Wilcoxon signed-rank test also shows there was a significant difference between BBS and CICRO. Some of the subjects answered our interview as follows: • “It was fun because I could see that users was interested in the same topic and sharing the interests among each other.” • “It was easy to understand the extent to which similar opinions existed.”

5 Discussions The results of the comparative experiment showed that the visual interface of CICRO is superior to a traditional, text-based interface for the online communication since CICRO could motivate users to post their remarks (i.e., 1.7 times larger than BBS) and our subjects basically had positive impressions on CICRO. We believe we can successfully design the interface as we initially intended to help the communication for the “expressive” crowds. One of the reasons why CICRO was better than BBS is that colored remarks on the interface triggered to express counter opinions and impressions from other users, resulting in active discussions among users. Another reason is that in the current implementation of CICRO, the size of older remarks gets smaller each time a new remark comes in the visualization area. Due to this limitation, if there are multiple groups (crowds) of users as shown in Figure 1, each group seems to be oriented to make remarks to keep their position more dominant or protect it in their discussions. We could observe such the interactions among our subjects at many times in case of using CICRO. Although we could have positive outcomes from the experiment, we consider the results of the experiment as preliminary ones. Over several thousands of internet users sometimes communicate to discuss one topic at the same time. The results of our experiments were based on the analysis of the communications among only 17 subjects. This is obviously the limitation of our study in this paper. We need to scale up the user study in the future. To do so, we also need to scale up the functionality of CICRO as it can process and visualize a number of remarks at the same time.

CICRO: An Interactive Visual Interface for Crowd Communication Online

259

We are also aware that CICRO needs to automatically collect, group, merge, and polarize users’ opinions. The current implementation of CICORO requires users to position their remarks manually. In order to overcome this inconvenience, we would like to incorporate natural language processing functions into CICRO in the future. In addition, we would like to develop several kinds of clients for the crowd communication, by extending existing communication tools such as twitter and YouTube. Furthermore, we would like to develop interfaces for other kinds of crowds described in Section 2, though in this paper we only present an interface for the communication of the “expressive” crowds.

6 Related Work There have been many studies on the visualizations of online communications and communities. For instance, a variety of systems have been proposed to visualize conversations in a newsgroup. Newsgroup Crowds [7] and Netscan [8] provides a means to visualize the distribution of topics and users activities in a newsgroup. PeopleGarden [9] is a visual interface which employs a garden metaphor to represent enthusiasm for topics in a newsgroup. Loom [10] also visualizes threads of newsgroup conversations in a circular pattern to grasp active topics at the time. These systems and interfaces are mainly designed to help users understand the current situation in a newsgroup and/or to motivate new users to join newsgroup discussions. Therefore, the conversations are visualized using the data in the past. CICRO places value on real-time visualization of the crowd communication to help “temporal” users better understand the scale and structure of the crowd communication. In this sense, Chat Circle [11], a chat system using the two dimensional remark positioning technique, seems to provide similar interactions to users as CICRO does. However, in Chat Circle, individual users are identifiable. It visualizes remarks so that users can easily find partners to talk with. In contrast, CICRO visualized remakes anonymously to encourage free discussions among users.

7 Conclusions and Future Work In this paper we presented the concept of the crowd and its communication online. We introduced CICRO, an interactive visual interface for the crowd communication and showed the result of the comparative evaluation. We found that CICRO encouraged 1.7 times more active discussions among users than that of BBS since it provides users with an easy way to understand the existence of a variety of others’ opinions and/or impressions in the discussions. CICRO is currently designed to simply visualize the existence of different groups of online users with crowd minds. In the near future, we will built functions of the natural language processing into CICRO, in order to automatically collect, group, merge, and polarize users’ opinions. We also would like to develop several kinds of clients for the crowd communication, by extending existing communication tools such as twitter and YouTube. Acknowledgements. We would like to thank all of the students for participating in our experiment as subjects.

260

M. Ohira, H. Masaki, and K.-i. Matsumoto

References 1. 2. 3. 4. 5. 6. 7.

8.

9.

10.

11.

Le Bon, G.: Psychology of Crowds (1895) Negroponte, N.: Being Digital, Knopf (1995) Sunstein, C.: Republic.com, Princeton University Press (2001) Tarde, G.: L’opinion et la foule (1901) Brown, R., Lindzey, G. (eds.): Mass Phenomena, Handbook of Social Psychology, vol. 2, pp. 833–877. Addison-Wesley, Reading (1954) Turner, R.H., Killian, L.M.: Collective Behavior. Prentice-Hall, Englewood Cliffs (1957) Viegas, F.B., Smith, M.: Newsgroup Crowds and AuthorLines: Visualizing the Activity of Individuals in Conversational Cyberspaces. In: Proceedings of the 37th Annual Hawaii International Conference on System Sciences (HICSS 2004), p. 40109b (2004) Smith, M.A., Fiore, A.T.: Visualization components for persistent conversations. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI 2001), pp. 136–143 (2001) Xiong, R., Donath, J.: PeopleGarden: creating data portraits for users. In: Proceedings of the 12th Annual ACM Symposium on User Interface Software and Technology (UIST 1999), pp. 37–44 (1999) Boyd, D., Lee, H.Y., Ramage, D., Donath, J.: Developing Legible Visualizations for Online Social Spaces. In: Proceedings of the 35th Annual Hawaii International Conference on System Sciences (HICSS 2002), vol. 4, p. 115 (2002) Donath, J., Karahalios, K., Viegas, F.: Visualizing Conversation. In: Proceedings of the 32nd Annual Hawaii International Conference on System Sciences (HICSS 1999), vol. 2, p. 2023 (1999)

Features to Support Persistent Chat Conversation Mohsen Rahimifar and Siti Salwah Salim Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, 50603 Kuala Lumpur, Malaysia [email protected], [email protected]

Abstract. Chat conversation persistency is one of interesting aspects of chat tools which make them even more useful as a communication medium. However, persistency alone is not useful unless there are features that support persistent conversations in order to increase the usability of a persistent environment. Moreover, these features simplify the processes of conversing and information retrieval for chat participants. This paper uncovers how different persistent chat tools have tried to overcome such difficulties by employing design factors and strategies. Design factors are concerned with the particular functionality of a feature and they are generally concepts that the chat tool should support. Design strategies concentrate more on the rules and restrictions applied to the users and the features available in the tools. A set of design factors and strategies for persistent chat conversation is presented and discussed in relation to the features that support persistent chat conversation. Keywords: Persistent chat conversation, features, design factors, design strategies.

1 Introduction Persistency simply means the availability of previous content for later use [1]. Chat conversation persistency is one of interesting aspects of chat tools which make them even more useful as a communication medium. However, persistency alone is not useful unless there are features which support it. These features and different approaches to their integration into chat tools are the basis of this research. Generally, persistent chat tools have to cover two different phases. The first is the dialogue phase or the time when the chat conversation takes place and the chat archive is being created. The second is the information retrieval phase where users or readers go back in the chat archive to extract their desired pieces of information. Mahowald and Levitt [2] indicated that (i) chat conversations are commonly short and limited to a single topic (ii) the tone of dialogue is informal and less appropriate for the workplace, and (iii) multitasking is another characteristic of communicating through chat, which means that users tend to perform several tasks while chatting. On the other hand, during the long period of time that chat has been around its interface has had very minimal evolution due to its simplicity and informality of use [3]. However, this simple interface is in most cases not efficient enough to cope with the complexities of A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 261–266, 2011. © Springer-Verlag Berlin Heidelberg 2011

262

M. Rahimifar and S.S. Salim

information retrieval from the chat archive. These facts affect the formation of a conversation archive which tends to be the resource for the information retrieval phase. The second phase may even happen during the conversation when participants may refer back to the chat archive in order to keep track of the conversation or quote previous utterances. However, there are a number of challenges that result from the informal and textual form of chat conversations which make the information retrieval a difficult task. The following sections review how different persistent chat tools have tried to overcome such difficulties by employing design factors and strategies.

2 Related Work There is a variety of different chat tools which support conversation archive persistency. Based on the purpose of each specific chat tool, different feature sets are implemented to fulfill that purpose. Five chat tools have been selected to be investigated in detail. The basis for this selection is the availability of features which support persistent conversation. The following section studies Babble, Conversation Space, Threaded Chat, Chat Spaces and BackTalk as the selection of persistent chat tools. Babble: Accountability and awareness are the two major aims of this tool [4]. This tool presents four different areas. A section in which the chat log is saved on a cumulative textual basis and each new utterance is labeled with the name of its producer and a time tag. The second section shows the list of topics or “Places” that exist in the system that can be a symbol of topics in a discussion. The other two sections are the list of participants and the visualization of the user’s activity status during last few minutes. Environment awareness is promoted based on the topic and user activity visualizations. Moreover, the concept of places helps the organization of the chat log into different subject matters. However, the chat log remains flat text as far as the chat log of each thread is concerned. There is also room for non-persistent private conversations. Conversation Space: This tool supports multithreading along with basic textual structure for the chat log (called Remarks) [5]. The results indicated that persistence enabled users to collaborate efficiently on shared objects and material [1]. The tool was generally developed to act as an asynchronous communication medium; however, later on it turned out to be restricting if it remained asynchronous. Threaded Chat: Smith et al. [6] emphasized the organization of conversations. Since one of the major problems in the regular chat tools is distinguishing different topics of the current conversation that are captured together in a single chat log, Threaded Chat separates subjects from the very beginning so that the users can easily focus on any topics they like. The tool is implemented in a way that participants see the chat log like a tree and they are able to create new branches, each of which stands for a new topic. Users are also able to navigate to a desired topic and expand it by appending their responses to end of its branch. Chat Spaces [7]: This uses tree structure to bring threading into action; however, unlike with Smith et al. [6], environment awareness is improved by employing a temporal presentation of updates in the tree. This tool also has the ability to visualize the total time that a participant spent on a conversation.

Features to Support Persistent Chat Conversation

263

BackTalk: Although the main focus in BackTalk [1] is shifted towards persistency, there is a logical bond between the concepts of persistency and those introduced and analyzed in structured conversation. Generally, this tool aims to make persistency through a set of features that let the users manipulate the chat log in a richer way than just flat textual contents. These features include filtering, threading, tagging and visualization.

3 Analysis of Available Features Although different tools have selected several features and used them in combination to support information retrieval, it is not easy to state which features have to be selected and whether there is an optimized way of combining them, since some features affect others. It is also crucial to know how to implement a certain feature so that users can use it more effectively and intuitively. To discuss the feature sets that the five previously-mentioned tools have supported, the approaches are divided into two major categories: design factors, which are more concerned with the particular functionality of a feature, and design strategies, which concentrate more on the rules and restrictions applied to the users and the features available in the tools. Design factors can be implemented in the tool regardless of how users will use them. This provides the possibility of having diversity in their implementation. Design factors are generally concepts that the chat tool should support, for example, filtering chat conversation is a design factor. However, the way the filtering is implemented in a certain tool and its functionality details are decided based on the tool design. On the other hand, design strategies have more abstract logic behind them. They virtually define the way users should use the chat tool and control their interaction with the tool and other chat participants. Design strategies also cover the general rules that the chat tool should support. An example of design strategies is turn taking. While chat participants are conversing, two different interaction models can be followed. The first is to allow users to converse freely and submit their statements to each other whenever they want, and another approach is to control the turn using the chat tool. The first approach provides more freedom for users but it can lead to unwanted misunderstandings due to the statements’ sequencing. The second approach, which is commonly named strict turn taking, can be limiting for users to some extent but makes their interactions clearer and easier to follow. Common design factors observed in revised tools are discussed in this section. Firstly, there is filtering, which is the ability to skip through unimportant or irrelevant content without reading it. Filtering can be applied based on data or metadata. There are also downsides to filtering, such as improper filtering of useful information due to an inaccurate filtering process. Secondly, there is threading, which is an approach that lets participants create branches of discussion out of the main dialogue in order to concentrate on specific subject matter in sub branches. The advantages of the threaded conversation structure are: (i) ease of finding and focusing on a certain topic; (ii) because of proper referencing features fewer repairs are needed to cover errors caused as a result of fast responses; and (iii) providing the ability to manage the conversation through setting customizable levels of accessibility for each topic. On the other hand, according to the evaluation results [6], users face some level of difficulty in locating

264

M. Rahimifar and S.S. Salim

new incoming messages as they arrive in different branches, which significantly affects the environment awareness of users. Thirdly, there is content visualization, which is the capability to visualize user interactions into proper visual models to convey certain information to the users. Visualization, if accurately used, significantly increases environment awareness. Fourthly, there is semi-synchronous support, which makes it possible for participants to have the freedom of contributing to the conversations regardless of other participants’ presence or absence. Conversing in a persistent conversation can be more useful if users can participate whenever they wish, and the tool can support the necessary requirements for such an interaction model, for example proper environment awareness that helps chat participants to keep track of the conversation even if it happens while they are absent. Finally, there is shared artifacts support, which provides the ability to share external artifacts and to have discussions over them. This feature is important in referencing throughout the conversation. Table 1shows a comparison of the existing systems which have been discussed. It is crucial to mention that each tool can have a different approach to utilizing and implementing these design factors in features. Based on this comparison, filtering and shared artifacts support are the least supported amongst these persistent chat tools. Absence of filtering may cause some difficulties for participants as they try to go back in conversation archives to seek and extract information. Since in a persistent chat environment chat archives are expanding very rapidly, there have to be proper features that support effective content filtering to accelerate the process of information retrieval. Shared artifacts is a way to make the grounding process easier and to reduce unnecessary conversation redundancies by eliminating referencing statements as there is a clear approach to sharing resources amongst participants. On the other hand, threading is the most supported design factor; however, each tool has its unique approach of designing it and each design has its pros and cons. Table 1. Chat Tools Comparison Based on Design Factors

Tool Name

Persistenc e

Filterin g

Threadin g

Content Visualizatio n

Semisynchronou s Support

Babble

Yes

No

No

Partial

Yes

Shared Artifact s S No

Chat Spaces

Yes

No

Yes

No

Yes

No

Conversatio n Space

Partial

No

Yes

No

No

No

Threaded Chat

Partial

No

Yes

No

No

No

Backtalk

Yes

Partial

Yes

Partial

Yes

Yes

Features to Support Persistent Chat Conversation

265

Table 2. Chat Tools Comparison Based on Techniques Strict turn taking

Implicit feedback

Environment awareness

Chat Log reconstruction

Referencing

Babble

No

No

No

No

No

Chat Spaces

No

No

Yes

No

Partial

Conversations Space

No

No

Yes

No

Yes

Threaded Chat

Partial

No

No

Partial

No

Backtalk

No

No

Yes

Partial

No

Tool Name

The common design strategies employed in revised tools are discussed in this section. Firstly, there is strict turn taking, which is when the chat tool controls the occurrence of utterances by limiting users to submitting statements in their own turn (which rotates amongst participants). An advantage of turn taking is that the sequential chat log is easier to read and respond to; however, as a downside it can be very limiting in certain cases. Secondly, there is implicit feedback, which is the ability for the participants to anonymously evaluate any specific content. Thirdly, there is environment awareness, which is letting the users know what happened, when it happened and with whom it took place. Fourthly, there is chat log reconstruction, which is the capability of the chat tool to regenerate the chat log based on specific criteria which are given by the users. Finally, there is referencing, which is the ability to link together and associate pieces of content. Table 2 represents a comparison of the previously-mentioned chat tools regarding a number of popular design strategies. Supporting design strategies are slightly different from design factors. Based on the overall design of the chat tool and its feature set, certain design strategies can be selected or applicable. Hence, supporting more design strategies will not necessarily lead to a better and more effective design. Design strategies should be selected with caution as a wrong selection may lead to decreased usability and user performance.

266

M. Rahimifar and S.S. Salim

4 Design Factors and Strategies for Persistent Chat Conversation Each chat tool which supports conversation persistency should provide features that support persistency. Based on the investigation of the five previously-mentioned chat tools, each tool supports a customized set of features for this purpose. Moreover, these features simplify the process of conversing and information retrieval for chat participants. Based on these two facts, it can be said that any chat tool which aims to support persistent chat conversations should consider the employment of a predefined, normalized and standard feature set, which simplifies the process of chat tool design according to previous practices of feature set combinations. Although the design process is generally subjective to each chat tool, similar to other best design practices, certain feature consistencies and inconsistencies can be converted into a guideline solution to prevent design pitfalls in the area of chat conversation persistency. For example, a feature such as threading can reduce the usability of a chat tool if it does not have a proper environment awareness strategy. An example of features that can support each other is the combination of referencing and filtering in order to provide more accurate information. Implicit feedback and content visualization is another useful combination to simplify the process of decision making.

5 Conclusion Based on previous practices of chat tools, including persistent chat tools, there is a crucial need to provide a generalized feature set that supports persistent conversations in order to increase the usability of such persistent environments. However, the design of each feature and their combination can significantly affect the usability of the chat tool, so having a variety of feature combinations and evaluating their compatibility can be important in the formation of a unified feature set design and combination guide.

References 1. Fono, D., Baecker, R.M.: Structuring and Supporting Persistent Chat Conversations. In: Proc. CSCW, pp. 455–458 (November 2006) 2. Mahowald, R., Levitt, M.: Finding a place: Corporate instant messaging market forecast & analysis, IDC Report (2000) 3. Muller, M.J., Raven, M.E., Kogan, S., Millen, D.R., Carey, K.: Introducing chat into business organizations: Toward an instant messaging maturity model. In: Proceedings of the 2003 International ACM SIGGROUP Conference on Supporting Group Work, pp. 50–57. ACM Press, Sanibel Island (2003) 4. Erickson, T., Smith, D., Kellogg, W., Laff, M., Richards, J., Bradner, E.: Socially translucent systems: social proxies, persistent conversation, and the design of Babble. In: Proc. Computer- Human Interaction (CHI), Pittsburgh, PA, pp. 72–79 (1999) 5. Popolov, D., Callaghan, M., Luker, P.: Conversation space: Visualizing multi-threaded conversation. In: Proceedings of AVI 2000, Palermo, Italy (2000) 6. Smith, M.A., Cadiz, J., Burkhalter, B.: Conversation Trees and Threaded Chats. In: The ACM 2000 Conference on Computer-Supported Cooperative Work (CSCW 2000), Philadelphia, PA, pp. 97–105 (2000) 7. Geyer, W., Witt, A., Wilcox, E., Muller, M., Kerr, B., Brownholtz, B., Millen, D.: Chat spaces. In: Proceedings of DIS 2004, pp. 333–336 (2004)

Design and Evaluation of an Online Social Support Application for Family Caregivers Matthieu Tixier and Myriam Lewkowicz Université de Technologie de Troyes - ICD/ Tech-CICO - UMR CNRS 6279 12, rue Marie Curie, BP 2060, 10010 Troyes, France {Matthieu.Tixier,Myriam.Lewkowicz}@utt.fr

Abstract. A lot of people who live with a loss of autonomy related to ageing or disease benefit from the daily support of some of their relatives in order to perform their day-to-day duties (waking up in the morning, shopping, cooking). Family caregivers are strongly committed to the support relationship with their ailing relative. The assistance they provide takes up a large part of their time and energy. Offering them places for social support appears as a relevant solution. Besides, more and more people are turning towards the Internet to find support when they experience difficult situations. Our research deals with the complex issue of providing online social support to family caregivers through an interdisciplinary research project which aims at getting a better understanding of peer-to-peer social support practices. The healthcare network RéGéMA is our first partner in the requirements analysis, test and deployment of our application, Aloa-aidants.fr. Keywords: Family Caregivers, Social Support, Healthcare Network, Support Group, Interdisciplinary Design.

1 Introduction A lot of people who live with a loss of autonomy related to ageing or disease benefit from the daily support of some of their relatives in order to perform their day-to-day duties (for instance, waking up the morning, shopping, cooking, getting dressed up). The care these family caregivers are providing to their ailing relative is then essential. The French society just starts to become aware of the burden of the family caregivers and solutions are needed to help them. Allowing these caregivers to share their experience and to find a sympathetic hearing to their problems, in other words offering them places for social support, appears as a relevant solution in this context. Social support constitutes a relevant answer to the moral distress expressed by family caregivers [1]. Barnes and Duck [2] define social support as an exchange of verbal or non-verbal messages, which convey emotion or information, in order to reduce the stress or the uncertainty of a person. This support is traditionally provided by family and friends. It is performed through the exchange of information, advices, emotions or more tangible resources which lend to recognize the value of the person in need of support [9]. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 267–276, 2011. © Springer-Verlag Berlin Heidelberg 2011

268

M. Tixier and M. Lewkowicz

However, as people today are often living and working far from their family, more and more people are searching for other ways to find support when they experience difficult situations. They are for instance turning towards the Internet to reach and talk with people who are or were going through the same troubles and find a sympathetic hearing to share their problems. Several research work have highlighted these mediated peer-to-peer social support practices and described existing online communities, as empathic communities [3] [5], Virtual Healthcare Communities [4] and patients’ online communities as “communities of unintended interest” [11]. Developing such online social support space intended to family caregivers was the aim of our design project in the framework of our partner, the RéGéMA healthcare network. Our research deals with this complex issue of providing online social support to family caregivers through an action research project which aims at getting a better understanding of peer-to-peer social support practices. In the framework of this project [6] [7], we have worked in cooperation with researchers from Psychology, Sociology and Conversational Analysis in order to inform the design of an online application dedicated to social support. The healthcare network RéGéMA is our first partner in the requirements analysis, test and deployment of the application. This network provides support to patients suffering from memory disorders, like Alzheimer’s disease, and their family caregivers in the Aube region (North-East of France). This paper presents the approach that we have followed in order to design Aloaaidants.fr, an application dedicated to online social support for family caregivers. In what follows, we first describe in more detail the situation of family caregivers and our partner RéGéMA. Opportunities offered by online social support in this context are highlighted. Secondly, we consider the studies carried out within our interdisciplinary project with the aim to guide designers in developing online services usable by family caregivers and relevant for social support practices. Several results that have been of interest for the design of Aloa-aidants.fr are highlighted in this section. Then, we present the technical translation approach we have elaborated in order to design the functionalities of our application on the basis of the results of the studies formerly introduced. We furthermore illustrate our approach by presenting several functionalities intended to develop a social support activity among family caregivers through our application. A first version of Aloa-aidants.fr is now online and we present the results of the first evaluations that we have conducted with RéGéMA family caregivers as well as with another association interested by the use of our system. Finally, we conclude and address some future research perspectives in order to develop and sustain an active social support community of family caregivers on Aloa-aidants.fr.

2 The Needs of the Family Caregivers of Alzheimers’ Patients Serious neurodegenerative diseases such as Alzheimer’s disease (AD) greatly reduce the patients’ autonomy as their cognitive abilities gradually decline. The patients become unable to deal with their own day-to-day requirements as the symptoms worsen, and their relatives have to assist them increasingly with even the simplest operations, such as shopping, bathing, and getting dressed. Patients’ relatives find it hard to cope with playing the role of caregiver for which they have not been prepared. As shown by several surveys on French caregivers’ situation [1] [8], beyond the

Design and Evaluation of an Online Social Support Application for Family Caregivers

269

financial cost of the disease (assistance with housework, adapting the home, etc.), it makes heavy demands on the supporting spouse or relative in terms of time and attention. The ineluctable cognitive decline of AD patients means that they need an increasing amount of support as their disease progresses. Caregivers tend to be strongly committed to the support relationship with their ailing relative, whether they are the patient’s spouse or child. The assistance they provide takes up a large part of their time and energy and leaves the caregivers little opportunity of escaping and taking care of themselves. Due to all these worries, caregivers express the need for help, apart from the need for financial assistance, which is felt to be the main response provided by social institutions. They clearly lack a means of expressing their distress and finding a hearing with which discuss their problems in order to obtain information and comfort. The most relevant persons in order to provide this hearing could be other caregivers who are facing similar problems. The only means of this kind available are the support groups run by associations. Our partner RéGéMA is a healthcare network which focuses on memory disorders such as AD. The main aim of the network is to coordinate the work of the many professionals (neurologists, general practitioners, social workers, psychologists, etc.) involved in the care and support of the these patients. The RéGéMA healthcare network runs monthly support meetings for spouses and children who are caregivers in order to provide them with a space where they can talk about their experience and discuss their problems. A web-based system appears as a relevant response in order to supplement existing social support devices like face to face support groups. Such system would provide family caregivers with a space available at all times where they can share social support with peers. Online social support can be a way to escape from the burden of caregiving without losing sight of the ailing relative since living the patient alone at home is often difficult. In addition, receiving social support on a daily basis enables people to talk about their problems immediately rather than letting worries accumulate for a long time before they find a hearing. The anonymous communications mediated by the Internet with people living elsewhere might encourage people who are reluctant or unable to participate in face-to-face support groups [12]. Online services of this kind would usefully supplement the services already proposed by medical healthcare networks, as it would improve caregivers’ access to information and help them cope with their distressing situation.

3 Family Caregivers Social Support Practices Designing relevant and usable online social support services for family caregivers raise two challenges. First, we need to get a better understanding of current social support practices (computer mediated and face-to-face). And at the same time, we have to understand the specific needs of the family caregivers we are working with in term of social support. We present here the studies that have been carried out on social support practices and caregivers’ needs. For both, we highlight several findings that have informed the design of our application. Each study is a complete research on its own and we only focus on some of their results that have been of interest for our purpose.

270

M. Tixier and M. Lewkowicz

3.1 Understanding Family Caregivers Social Support Practices In order to get a better understanding of the current social support practices, two studies have been carried out. We have conducted a one year and a half participant observation in the support groups ran by RéGéMA. This study gave us an understanding of the situation of RéGéMA family caregivers, the problems they daily encounter and their current social support practices. It also helped us to stay in touch with end-users representatives all along the design process. The researchers in conversational analysis and psychology involved in this project have studied current computer-mediated social support exchanges taking place on discussion forums on the Internet [13] [14]. They have highlighted the existence of interaction patterns in messages and described a communication contract and sociopsychological factors which underline online social support practices. These studies have highlighted that self disclosure is important for receiving social support. The analysis of messages posted on social support discussion forums show that users unfold personal and intimate details about their situation and the problem they encounter. Theories on computer mediated communication explain the related psycho-social factors through the need to cope with lack of context [15] and the disinhibitory effect of anonymity [16] [17]. Besides, family caregivers are used to unfold a lot of details, sometime very intimate, about their personal situation when they explain the daily problems they encounter on the occasion of the support groups. Information about their children, the help they receive, even the way their home is arranged may become important details in the stories they want to share with the support groups members. Another element reported as a key aspect of online social support practices is the importance of reciprocity. The script elaborated on the basis of the online social support exchanges analyzed by our colleagues’ researchers in conversational analysis has highlighted this element [7] [13]. At the level of the messages sequence, the common occurrence of an evaluative message at the end of social support exchanges underlines the importance of reciprocity. These evaluative messages through which the support recipient generally thanks the other users for their supportive messages are almost always observed. Their absence may be negatively understood by the members of the discussion thread. Another marker of the expectation of reciprocity is observed at the level of speech acts. The support recipient is often anticipating the support she may receive by thanking other users from her first message. These two elements provide evidence in favor of the presence of a norm of reciprocity underlying the studied online social support discussion forums. 3.2 Understanding Our Users In order to understand the family caregivers’ situation and needs, our colleague sociologist has carried out fifteen interviews with family caregivers. This study helped us to gather needs expressed by interviewees and to highlight more latent needs through a better knowledge of their face to face and online social support practices [18] [19]. Eight of the interviewed family caregivers have used online discussions forums in order to help them to cope with their role of caregiver. Some of the interviewees have been recruited on the occasion of RéGéMA support groups.

Design and Evaluation of an Online Social Support Application for Family Caregivers

271

A need for information has been generally reported by the interviewees. The healthcare network and the support groups have been mentioned has an important source of information for them. There they can find resources to navigate within the complex medical and administrative environment associated with the disease of their relative, trustworthy information about diseases are provided and also help to plan assistance with housework and in-home care services. The support groups’ members share knowledge through the story of their caregiver daily experience and the way they tackle problematic situations (e.g. make bathing or dressing their ailing relative). They explain the different solutions they have tried and what finally works. Experienced caregivers and the support group coordinator suggest solutions to common problems. However, since every situation is unique, a solution for one case may not work for another. The support groups’ members explain they appreciate the support they found in the group and the “useful tips” they picked out from each other’s experience. The analysis of the use of online discussion forums by family caregivers [19] underlines that information seeking is their starting point in online social support practices. They pick out and gather information through the stories and questions posted by peers before, if ever, they open their own discussion thread. This is in line with the observation that support in terms of information is central in online social support [14], pure emotional support request are seldom observed and often mixed with information requests [13].

4 Design Approach Integrating the findings brought by these studies coming from different theoretical and methodological perspectives led us to develop an interdisciplinary design framework based on the idea of technical translation [18] [20]. The analysis and results of the studies formerly presented do not constitute a selfevident corpus from which designers can easily derive requirements. In order to make sense of the multi-level features highlighted by social sciences studies, we use the notion of technical translation as a tool to build design requirements from social sciences accounts. Our approach is inspired by the work of Bruno Latour and Michel Callon on the Actor Network Theory (ANT) [21], which provides a framework to describe the dynamic of sociotechnical systems, as are design projects. Through the notion of translation, ANT underlines the dynamic of the multiple voices and interests gathered around an innovation project and the way designed artifacts, as any other actors, take part to project, especially by reifying discourse and acting as representative of multiple intentions. Our purpose is not to describe these dynamics on the framework of the present project. As designer and stakeholder, we do not feel at ease to rigorously conduct such analysis although we recognize the existence of these phenomena. We are more interested we the last element we have introduced about ANT: the ways artifacts can reify discourses about the activity. To address this perspective, we propose the moderate notion of technical translation through which designers can account for the way functionalities of artifacts translate discourses about the activity and practices proposed by the different stakeholders of a design project. The aim of our translation approach is more to provide a tool for designers rather than a formal methodology to translate

272

M. Tixier and M. Lewkowicz

social sciences and activity inputs directly into system functionalities. The core idea behind our approach is to make open to discussion the detail and purpose of the designed functionalities with non technical stakeholders. The technical translation approach proposes to begin by selecting elements from the analysis of the activity, the activity features (AF). These AF are formulated by designers in order to reflect their interpretation and understanding of the studies conducted. For instance, we have selected the following AF from the findings formerly presented (section 3): • Self disclosure is important for receiving social support online. • Reciprocity is a favorable factor for social support. • Information seeking is the starting point of family caregivers in online social support practices. • Informational support plays a central role in social support. • Caregivers explain they appreciate the “useful tips” they pick out on the occasion of support group sessions. Designers define functionalities and build mockups according to the selected activity features. Several AF can inspire one or more functionalities. Designers explain the use they envision of their functionalities through scenario. These materials constitute resources to discuss designers understanding of the activity and the relevance of the proposed functionalities. Through mockups and scenarios, the detailed implementation (user interface and interaction logic) of functionalities is opened to discussion with other project stakeholders [22] [23]. At the end, the technical translations made along the design process constitute a design rationale for the project and a resource for further iterations.

5 Online Social Support on Aloa-aidants.fr We illustrate our design approach by presenting the functionalities - on the basis of the AF formerly introduced – which are offered by Aloa-aidants.fr (fig. 1) in order to support family caregivers. In order to support self disclosure on our application and provide more contextual information to ease mediated communication, every registered user is invited to present herself through a personal pages (1). They can explain their situation and give news to members with the blog attached to their personal page. Information of interest for family caregivers are highlighted as the city where users live - due to the local context of our application - and the helps users are benefiting (e.g. Autonomy Personal Allowance and support group on the figure). Helps tags can be clicked to find members according to this information. A space dedicated to experience sharing is also available (4) and allow users to share personal stories about the way they cope with the burden of caregiving. Users post their story under an existing topic or create a new one, thus experience of different members on a same subject are gathered. Members can comment and post supportive messages on each others’ stories.

Design and Evaluation of an Online Social Support Application for Family Caregivers

273

Fig. 1. The front page of Aloa-aidants.fr (left) and a screenshot of a member’s personal page (right)

Reciprocity cannot be directly translated into the functionalities of a system. Instead we provide opportunities for users to act with a concern for reciprocity. Users are systematically alerted by email when someone replied to their contributions on Aloa-aidants.fr. Users’ contributions that have received few or no messages are emphasized in the front page and at the top of their respective application space. Besides, registered users are invited to greet newcomers. The Document Sharing space (2) and the Questions and Answers (Q&A, 3) space are intended to offer informational support to users. A number of documents and websites useful for family caregivers have been uploaded by administrators. Registered users can supplement this space with the document and websites they find useful. Since information seeking is the starting point of family caregivers in online social support practices, all the content but members' personal pages can be viewed by visitors and consulted with the help of our application search engine. In order to highlight content of practical interest for family caregivers like the advices and knowledge exchanged on the occasion of the support groups, every contribution can be marked as a “useful tip” by the mean of a thumb up icon. Thus a collectively build knowledge source is constituted through the “useful tips” picked out by users on the application content. The “useful tips” are gathered under a dedicated section of the Aloa community space and they are always presented in the context of the question, document or user story where they have been mentioned.

6 Evaluation We have made first evaluations of our application according to our twofold aim of designing an application usable by family caregivers and relevant for social support practices. First, an evaluation of the usability of our application has been conducted with RéGéMA family caregivers in order to unfold problems and potential improvements in our design. Second, an evaluation has been conducted with the aim of assessing the relevance of our application for online social support.

274

M. Tixier and M. Lewkowicz

Usability evaluation has consisted of three evaluation sessions that have been carried out with RéGéMA family caregivers. We asked them to complete several tasks on the application and to express their opinion on the credibility and relevance of the use scenario elaborated during the design process. The application is seen as usable by participants according to their comments and their responses to the System Usability Scale (SUS) [24]. Several minor usability problems have been encountered, like problems with the modification of a past contribution or the lack of clarity of the form for sharing documents and websites. We are working to solve these problems for the later version. The remarks made by participants on the use scenarios and the relevance of our application functionalities have been of more interest. Initially, Aloa offers a new member to present her/himself as a family caregiver or a supporter (of the family caregivers cause). One user pointed out that she no more belongs to one or the other group since her ailing relative died several months ago. She suggests we should enable users to present also themselves as former caregiver since a lot of people are concerned with this case. Six months after having deployed Aloa, the experience sharing space encounters little participation. One user reports the fact that she needs to carefully prepare the story she wants to publish on the application. The idea of experience sharing is perceived as a deep and emotional activity. This fact seems to limit the participation on this space. Besides, it shows that the communication situation suggested through the user interface is not neutral. The family caregivers have also reported on the few time they have to spare on the Internet. We observe we have underestimated this aspect by relying a lot on users’ participation to generate the content of the application. Preparing more content already available online might bootstrap participation. The second evaluation was aimed at assessing the relevance of our application for social support practices. It has been carried out with UNISSON, a French family caregivers association aimed at providing support to family caregivers on the Internet. UNISSON was interested to study the opportunity to benefit from an application like Aloa for their purposes. We have provided them with a white-label version of Aloa where they have made extensive tests simulating the way the association brings support to family caregivers on social support discussion forums. At the end of their evaluation, 77 contributions have been posted and 69 user accounts registered. Feedbacks from UNISSON have been positive and they have confirmed their interest in the use of our application for their purpose of supporting family caregivers on the Internet.

7 Conclusion and Future Work In this paper we have exposed the approach we have followed to develop Aloaaidants.fr, an application dedicated to online social support for family caregivers. The findings of social sciences researchers have been really valuable to ground our design theoretically and empirically. The first version of our online social support application is now online and available for RéGéMA family caregivers since six months. Several messages have been posted by our first users but our aim of supporting an active online community is not currently achieved. As reported by evaluation sessions with family caregivers, more content has to be prepared to motivate users' participation. For instance, we are

Design and Evaluation of an Online Social Support Application for Family Caregivers

275

working to gather typical questions asked by new family caregivers at the onset of the disease of their relative. We are also invited to pursue our research on the means to accompany and develop a dynamic online social support community among RéGéMA family caregivers. Acknowledgements. This research was conducted with the support of Conseil Général de l’Aube in the framework of a strategic research program funded by Troyes University of Technology.

References 1. IFOP: Etude nationale ”Connaître les aidants et leurs attentes” (2008) 2. Barnes, M.K., Duck, S.: Everyday communicative contexts for social support. Communication of social support: Messages, interactions, relationships and community, 175–194 (1994) 3. Pfeil, U., Zaphiris, P.: Patterns of empathy in online communication. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 919–928. ACM, New York (2007) 4. Dannecker, A., Lechner, U.: Online and Offline Integration in Virtual Communities of Patients - an Empirical Analysis. In: Proceedings of the Third Communities and Technologies Conference, pp. 151–170. Springer, Heidelberg (2007) 5. Preece, J.: Empathic communities: reaching out across the Web. Interactions 5, 32–43 (1998) 6. Lewkowicz, M., Marcoccia, M., Atifi, H., Bénel, A., Gaglio, G., Gauducheau, N., Tixier, M.: Online Social Support: Benefits of an Interdisciplinary Approach for Studying and Designing Cooperative Computer-Mediated Solutions. In: Proceedings of the 8th Conference on the Design of Cooperative Systems, pp. 144–155 (2008) 7. Tixier, M., Lewkowicz, M., Marcoccia, M., Atifi, H., Bénel, A., Gaglio, G., Gauducheau, N.: Practices Analysis and Digital Platform Design: An Interdisciplinary Study of Social Support. In: Proceedings of the 9th Conference on the Design of Cooperative Systems, pp. 309–330. Springer, Heidelberg (2010) 8. PIXEL: Etude PIXEL - L’entourage familial des patients atteints de la maladie d’Alzheimer. Novartis (2000) 9. Cohen, S., Gottlieb, B.H., Underwood, L.G.: Social Relationships and Health. In: Social Support Measurement and Interventions: A Guide for Health and Social Scientists, pp. 3–25. Oxford University Press, New York (2000) 10. Neal, L., Oakley, K., Lindgaard, G., Kaufman, D., Leimeister, J.M., Selker, T.: Online health communities. In: CHI 2007 Extended Abstracts on Human Factors in Computing Systems, pp. 2129–2132. ACM, New York (2007) 11. Josefsson, U.: Patients’ online communities experiences of emergent Swedish self-help on the internet. In: Proceedings of the First Communities and Technologies Conference, pp. 369–389. Springer, Heidelberg (2003) 12. Salem, D.A., Bogat, G.A., Reid, C.: Mutual help goes on-line. Journal of Community Psychology 25, 189–207 (1997) 13. Gauducheau, N., Marcoccia, M.: Le soutien social dans les forums de discussion Internet: réalisations interactionnelles et contrats de communication. Psychologie sociale de la communication. De Boeck, Bruxelles

276

M. Tixier and M. Lewkowicz

14. Atifi, H., Gaglio, G.: L’entraide numérique en mots: le cas du forum « aide » des « Marocains d’ailleurs ». In: Actes du congrès de l’AISLF 2009. pp. 336–344 (2009) 15. Klaw, E., Dearmin-Huebsch, P., Humphreys, K.: Communication patterns in an on-line mutual help group for problem drinkers. Journal of Community Psychology 28, 535–546 (2000) 16. Caplan, S.E., Turner, J.S.: Bringing theory to research on computer-mediated comforting communication. Computers in Human Behavior 23, 985–998 (2007) 17. Coulson, N.S.: Receiving social support online: an analysis of a computer-mediated support group for individuals living with irritable bowel syndrome. CyberPsychology & Behavior 8, 580–584 (2005) 18. Tixier, M., Gaglio, G., Lewkowicz, M.: Translating social support practices into online services for family caregivers. In: Proceedings of the ACM 2009 International Conference on Supporting Group Work, pp. 71–80. ACM, New York (2009) 19. Gaglio, G.: Consommation d’informations sur Internet et modulation de la relation aux médecins - Le cas d’aidantes de malades atteints d’une pathologie lourde. Sociologies Pratiques, 63–74 (2010) 20. Tixier, M.: Aloa: un outil de soutien social en ligne pour les aidants familiaux - Proposition d’une démarche d’analyse et de conception interdisciplinaire. Ph. D. Thesis (2010) 21. Callon, M.: Some elements of a sociology of translation: domestication of the scallops and the fishermen of St Brieuc Bay. Power, action and belief: a new sociology of knowledge? pp. 196–223. Routledge, London (1986) 22. Carroll, J.: Five Reasons for Scenario-Based Design. Interacting with computers 13, 43–69 (2000) 23. Erickson, T.: Notes on Design Practice: Stories and Prototypes as Catalysts for Communication. In: Scenario-Based Design: Envisioning Work and Technology in System Development, pp. 37–58. Wiley & Sons, New York (1995) 24. Brooke, J.: SUS-A quick and dirty usability scale. Usability Evaluation in Industry, 189– 194 (1996)

An Investigation into the Social Network between Three Generations in a Household: Bridging the Interrogational Gaps between the Senior and the Youth Tsai-Hsuan Tsai, Yi-Lun Ho, and Kevin C. Tseng Department of Industrial Design, Chang Gung University, Taoyuan, Taiwan [email protected]

Abstract. The traditional pattern of living arrangements makes the family a strong source of financial and emotional support for its members, and it connects grandparents and grandchildren in a unique bond. Unfortunately, it is reported that the number of older people living alone is rising in most countries. Although a lot of effort has been put into slowing down the social and emotional isolation experienced by the elderly, little has been done to investigate the social interaction between generations. In order to bridge intergenerational gaps between three generations of a family, this research work begins with an interview-based study to give insight into individual perspectives among those aged parents living alone and to identify their specific needs in generational connections and interactions.

1 Introduction Everywhere in developed countries, the population is aging rapidly. Co-longevity has greatly increased the duration of family ties. The parent-child relationship may last 67 decades; the grandparent-grandchild bond 3-4 decades[1]. In Chinese tradition, filial piety has always been the ethical core of the family. It also maintains the stability of China’s social structure. Chinese people regard filial piety as a virtue and obligation, and the most basic obligation of filial piety for traditional Chinese people is living with parents. The traditional pattern of living arrangements makes the family a strong source of financial and emotional support for family members and connects grandparents and grandchildren in a unique bond. Hagestad (2006) signified that grandparents significantly influence their grandchildren in many different ways. For instance, grandparents can play a role as a family historian, mentor, nurturer and surrogate parent. In addition, researchers have suggested that grandparents act as the reserve army of the family. When the family suffers from crisis, such as illness, financial problems or divorce, grandparents emerge from behind and support problem resolutions, maintaining close relationships between family members [1]. Furthermore, the relationship between the elderly and adolescents does not only compensate for agerelated deficits (e.g., cognitive performance, cognitive-affective complexity)[2], but also facilitates physical assistance inherent in co-residence [3]. Due to financial issues and caring reasons, three-generation households make family relationships more intimate. By mutual assistance, these ties bridge the relationship and interaction of the A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 277–286, 2011. © Springer-Verlag Berlin Heidelberg 2011

278

T.-H. Tsai, Y.-L. Ho, and K.C. Tseng

three generations. As a result, cultural values make three-generation households a representative symbol of the Chinese culture in most Asian countries. Unfortunately it is reported that the number of older people living alone is rising in most countries. In some European countries, more than 40 percent of women aged 65 and older live alone (NIA, 2007). Even in societies with strong traditions of older parents living with children, such as in Japan, Taiwan and China, traditional living arrangements are becoming less common due to social changes and economic growth [4-6]. In Taiwan, social changes have been accompanied by rapid changes in family structure and relationships, including the spread of schooling, the employment of young people outside the family, increasing separation of the residences of parents and children before and after marriage, and the growing independence of young people [4]. Likewise, the migration of working-age adults from rural to urban China has altered traditional patterns of living arrangements and intergenerational support among elderly persons who remain in rural regions [6]. For the above reasons, the generational gaps between younger generations and their seniors have become clearer and more noticeable, and the correction of the intergenerational gap is therefore a crucial issue. Social networks go through periods of change relating to life 'transitions', both in terms of the people in the network and the importance of those people. For older adults in particular, social network size and membership may be reduced by ‘transitions’ such as divorce, spousal illness/death, and geographic displacement – either to be closer to family or to move to sheltered housing or a nursing home. Retirement also has strong effects[7]. Work is likely to have taken up a large part of older adults’ time and to have played the largest part in the development of their social networks, beyond the inclusion of family members. Once retired, older adults may find it difficult to maintain or to increase membership of their networks. Age-related functional declines, such as reduced mobility, may make pose challenges to meeting people [8]. Recently social gerontologists have been concerned with the emotional well-being of older persons, and the extent to which these effects are mediated by subjective feelings of social integration [9]. Many studies have even focused on healthpromotion interventions that reduce social isolation and loneliness among older people, which include encouraging the elderly to participate in group activities, such as educational input [10-13], providing social support [14-16] and one-to-one home visits, assessments, or provision of information and services [17]. Participation in these social activities can help the elderly to establish social networks and to improve their physical and mental health. In addition to joining general social activities, the main impact on the individual social network involves the existence or quantity of social ties or relationships, which may in turn be distinguished as type (e.g., marital, kin/nonkin) and frequency of contact. Furthermore, the quality of social relationships is another factor that affects social network.[18]. Although a lot of effort has been put into slowing down social and emotional isolation for the elderly, little has been done to explore the social relationship between generations. Traditionally, Asian countries were dominated by the Confucian ethic of filial piety, and they paid attention to social interactions within their families. The traditional living arrangement and formation of families are mutually beneficial for the elderly and adolescents. However, the prevalence of nuclear families makes the interaction between intergenerational family members more difficult. In order to

An Investigation into the Social Network between Three Generations

279

explore the intergenerational relationship and lessen intergenerational gaps in threegeneration families, this study begins with an interview-based study to give insight into the individual's own perspectives among those aged parents living alone and to identify their specific needs in generational connections and interactions. This is followed by the second interview-based study aimed to explore the perspectives of younger people of three generations of a family.

2 Ageing and Generational Relations from Perspectives of Older Adults Living in a Senior Community This study took a qualitative approach. It was conducted among older parents aged 60 years and over, living alone in a senior community. There were 52 volunteers, about one-eighth of the senior community, with 17 men and 35 women. Written informed consent to participate in the interviews was collected through basic information forms, and the interviewer obtained the participants’ verbal agreement and arranged a time and place for the interview. All interviewers were carried out in a senior community where the participants live. The participants were interviewed about stage of life, social lives and activities, and intergenerational relationship in order to explore their social networks and the intergenerational connections among those older parents living alone. The duration of the interviews was between one hour and two hours. The interviewers were audio- and video- recorded and later transcribed verbatim. Content analysis was employed in this study. It describes a family of analytic approaches ranging from impressionistic, intuitive, interpretive analyses to systematic, strict textual analyses. The interviews were guided by the following themes: stage of life, social lives and activities, and intergenerational relationship. In terms of life stage, changes in living arrangement and life transitions were discussed among those older people living in a retired senior community in this study. In addition, the study found that the senior citizens referred to psychological feelings. Therefore, the psychological feeling was another category of this theme. In describing and analyzing social lives and activities, activities that older adults engage in and experience are considered as part of their social lives in this study. In addition, the study analyzed the quality of social structure between the elderly and their friends in the senior community. As past studies have rarely mentioned generational issues, the theme of intergenerational relationship is the most important to this study. The objective of this theme is to give insight into the lives of aged parents of three-generation families, who live alone, to explore the depth of social connections between the young and the old generations and also to investigate the perspectives and social needs of intergenerational relationships that older people desire. Two steps of analysis, from open to focused coding, are conducted in this study. The transcribed interviews were read several times. An empirically-oriented approach, with the themes mentioned in the above section and set as a general point of reference, was employed as the first step of analysis. The researchers individually marked transcribed verbatim/text relevant to the aim of this study, brought condensed "meaning units" for each themes together, and compared and discussed the meaning units. In the second step of the analysis, the meaning units were kept on a more

280

T.-H. Tsai, Y.-L. Ho, and K.C. Tseng

abstract level, guided by the themes. Discussions with a co-reviewer took place in each step of the analysis. The analysis was interwoven with other steps of the research process, as is characteristic of most qualitative research. Stage of Life. The 52 elders all live alone in a senior community. Most participants used to dedicate themselves primarily to their families and children. As they grew older, the adult children encouraged their elderly parent/parents to move to a senior home for safety issues, medical consulting, social support and relations, and quality of life in retirement. Since moving to the senior community, most of the residents were engaged in many and varied social events, leisure activities, and learning programs; in addition, they have opportunities to meet new friends and to create social groups. In addition, the residents kept in touch with their friends whom they had met before moving to the senior community. Some of the participants indicated their retired lives were more pleasurable than before, and some of them expressed that living in a senior home is very monotonous and lonely. Social Lives and Activities. About 60% of the participants engage in activities such as chatting (e.g., sharing experiences or expressing feelings) and playing chess or mahjong, and they participate in games or group activities that volunteers organize. Many participants regularly attend community programs that the village holds (e.g., classes for calligraphy, computer use, painting, ball dancing, tennis or language) or festival-related activities and lectures (e.g., Dragon Boat Festival, birthdays, Mother's Day). The participants think that these diverse activities not only offer a great " making friends" opportunity, but also fill their spare time that creates a feeling of companionship and enriches elderly lives. For some elderly people, although the senior community is a place where they can meet people, share experiences, and learn different things in societies or classes, having such social interactions make them feel lonely or even strange. Intergenerational Relationship. It is noted that a quarter of the participants whose children live abroad visit only on holidays or special occasions. Even though these elderly participants are accustomed to the lifestyle at a senior home and enjoy meeting many friends, they desire and expect their weekly family visits. The children and/or grandchildren’s visits bring a sense of pride to the aged parents. Those older people not only show a high level of participation in family activities, but also value quality of life and generational relationships, measured by the quantity of family visits. Following Eastern tradition and cultures, family relationship plays an important role. However, for those aged parents living alone in a senior community, they express their worries regarding life course transitions, the generational difference, and grandparent-grandchild relationship; family company and generational interactions are presented as a valuable source of emotional support to the 52 participants who live alone in the senior home. Besides, in addition to the conventional interpersonal communication via telephone and in person, many participants attended computer training courses in order to contact their children and grandchildren via the Internet and emails.

An Investigation into the Social Network between Three Generations Table 1. ‘Stage of Life’ from the 52 participants living alone in a senior community Theme

Stage of Life

Category

Social network heterogeneity

Sub-category Meaning unit

Work status

Retirement

Before the age of fifty, I had been working. (Chang8) Ah, a lot of different… ah, the former to be busy, busy to do business᧝(Chen13) I used to volunteer (Yao Ho16) Before the age of 50, every day, work, work, make the civil service. (Hsing17) Busy before the age of 50 ... Ah. Care of business.(Hsu21) Yes, this is ... busy doing things ... the company .... (Chu23) Caring for the grocery store (Chen28) Trained before the age of 50 and go out to serve ... (Li36) Before the age of 50, you have pressure at work and children (Lin34) So you think the biggest difference is busy at work…..(Wang39) There are busy with work before, and now it is very interesting here (Yu Hsia40) Ah for the life of that time is very busy ... tired ... is very hard but also very happy because I'm happy at work ( (Yuan41) Busy both at work and family (Chang44) Before the age of 55, the pressure of work is very large, then retirement is entirely out of the pressure (Chien43) Before the age of 50, I was busy at work….. (Lin45) Before 50 years old, busy with work to make money (Lu50) I looked after the family at home (Yang Chin5) Not only to do housework, but also with children (Chang8) Yes, look after child (Liu15) Right! Busy, then to my grandchildren. (Wu19) In raising children, later with grandson. (Chu20) Doing housework and look after children….. (Hsu21) When I was young, I was busy with my child and my husband……(Hu27) There was a lot of house work before but now they are grown up……(Li36) Before 50 years old, busy with work and children…..(Lin34) ….educate my children……(Yuan41) I was a little different, I do public servants’ work until my husband suffered from an illness …later I looked after him until he died…. (Li42) Both busy for work and families (Chang44) cultivated children ... (Lin45) life is of course not the same; before retirement, worked every day, doing housework, watching child (Lin47)

Category

Many friends. This is my home, I like this. (Yang 5) There are more friends here than at home(Wu19) To go out with friends…… ……chess with friends…… (Chu20) There are more friends here and makes me comfortable…..( Chang24) Go out with friends here(Yu Hsia40) I make a lot of friends here... (Lin45) There are more friends here to accompany with me….(Yang Tsai22) The old friends can visit me here... (Chang24) I meet many old friends here (Chu35) I keep in touch with old friends….(Lu50)

Psychosocial impacts

Sub-category

Positive

Negative

Meaning unit

Life is easier than before(Fang6) in this place is very casual life….(Liang4) The life would be happier here᧝(Yang Chin5) We have had a very good life here, and there is good for retirement (Chang8) So now more like the holiday and enjoy life (Chu20) I felt very comfortable staying here ... just like holiday. (Hsu21) Now is more freedom, more happiness(Chang24) I tell people how happy I came here(Hu27) …..retirement can be simply (Li36) Now I feel interest here…(Yu Hsia40) Life here is so happy...(Lin45)

Life is not a problem ... but here it is so boring ....(Chu23)

281

282

T.-H. Tsai, Y.-L. Ho, and K.C. Tseng

Table 2. ‘Social Lives and activities’ from the 52 participants living alone in a senior community Theme

Social Lives and Activities

Category Sub-category

Meaning unit

Practically Community program I like to dance, social dance…. (Chen 13) There are many lessons here, I now have the computer classes, piano lessons, social dances, a healthy lecture…..(Wu14) Oh .... I sometimes sing the old songs, sing and learn with a teacher…. (Chu20) Oh .... I now have the computer classes (Chen 37) Ah I have to study painting sketch here ... (Yu Hsia40) Ah, painting, calligraphy, piano .... art, ... ... so it’s hard to find me (Li42) Ah.. I forgot today is Monday. I have to calligraphy Ah, (Hsiung46) There are two, such as storytelling, acupressure, meridian massage (Lu50) I only attend computer classes and English class (Lai51)

Festival-related or special activities

Group activities

Do you have to participate in the Dragon Boat Festival Grandpa? Yes(Wang2) I have participate in carnival activities, (Yang1) Mother's Day I seem to attend ah(Wang2) I've been to Father's Day, and Mother's Day (Yang Chin5) Dragon Boat Festival grapefruit peel (Chen13) I participated in a graduation party… (Chen13) I participated in a graduation party and took a photo (Wu14) I’m so happy to volunteer one day a week…(Chien43)

Category

I spend all day with them, playing mahjong. (Yang1) …playing billiards and video game…. (Wu14) Go out shopping with friends Oh.... I will not playing Mahjong, sometimes playing chess with them... (20Chu) Yes, I’m singing…..(Hsu21) I did not ... I ... I just... I'd play table tennis with my wife…I play chess, playing cards on Saturday… (23Chu) Mahjong only (Chang24) Not smoking, drinking, playing mahjong, I’m singing…(Feng25) I'm going to sing ,playing table tennis and climbing the mountain ... (Hu27) I climb the mountain ... (Liu30) I ... I just... playing Mahjong only (Li36) Go out shopping with friends (Yu Hsia40) ……I'm going to sing at night…(Li42) chatting, and chess (Li42) I'm going to sing with friends (Hsiung46)

Relational content

Sub-category

Social regulation or control

Relational demand and conflict

Meaning unit

Everyone like a family… (Wu19) I have a lot of friends here…(20Chu) ……...all the friends here are very nice to me (Hu27) all the friends here are very nice to me, I’m so happy (Chen38) Everyone like a family…they were very good (Li42)

Here is only two people ... with my wife… is so lonely here Ah ... Ah (23Chu) Have rarely neighbour (Liu30) I feel so strange here..(Chen33) I have no way to contact with them…(Chu35) I feel that they regarded me as kindergarten kids I feel very naive (Chu35)

An Investigation into the Social Network between Three Generations

283

Table 3. ‘Intergenerational Relationship’ from the 52 participants living alone in a senior community Theme

Intergenerational Relationship

Category Sub-category

Meaning unit

Reality Meaningful contacts

Category Sub-category

Participate in family activities

How many times does your son visit you? sometimes (Wu19) My child always comes here (Chu20) I use the telephone…I can’t use cell phone...(Chen28) When my son call me, I will go back home……(Chen31) Does your child come back to see you? Yesterday my eldest son came to see me, took me out, He would come almost every week (Chen38) It’s easy to visit me, and take me out to play…. (Yu Hsia40) By e-mail, because my children will write e-mail, so I learned to use…. Ah Ah like looking on Facebook to see a lot of photos (Lin47)

Psychosocial impacts Satisfied My children are very good to me (Chen33) My children are filial…(Chen38)

Meaning unit

They like to go abroad and take me with them… (Yao Ho16) We always go out to have family dinner…(Chu23) My son and grandchildren take me out and go abroad….(Yang Tsai22) Yesterday my eldest son came to see me, took me out, He would come almost every week (Chen38) It’s easy to visit me, and take me out to play…. (Yu Hsia40)

Desired But both advantages and disadvantages, that is quiet good, but also lonely (Yao Ho16) When they need us, take me with them, but now they do not need us and send us back (Chu35) just arrived in the first year things are too hard, life is so bitter, feel homeless (Chen38) few care (chang44)

3 Generational Relations and Social Activities from the Perspectives of Younger Generations To make a comparison with the previous interview-based survey, the study included interviews with 19 middle-generation individuals. Each of them is a parent of a three-generation family and answered questions regarding everyday interactions between the three generations, social demands such as social patterns, family types, the relationship with their seniors, and social behaviours. Family categories and social behaviour include their lifestyles and social needs; family support refers to the intergenerational relationships between three generations of a family; specific needs comprise social behaviours and interpersonal communication. Family Categories and Social Behaviours. Nineteen middle-aged people, aged 3050 years, participated in this study. The family categories were classified to three generations living in a household (8 participants), nuclear families (10 participants), and only one participant who lives with a spouse. Socialization is the social process in which people are engaged in life experiences and learn cultural patterns, including social norms and behaviours. In this day and age, the use of computing technology and the Internet is an essential part of all types of work and has inevitably impacted socialization within the family, the young generation in particular. Based on the findings, socialization through face-to-face conversations, phone, emails, instant messaging (IM), and social network sites (SNS) has been adopted by the participants and

284

T.-H. Tsai, Y.-L. Ho, and K.C. Tseng

presented as customary and preferred social patterns. Communication technologies, such as Facebook, Plurk, MSN, or Skype, have made socialization easier for family and friends to stay connected, and they were considered today’s popular social tools by the participants. Family Support and the Intergenerational Relationship. Families of three generations state that family members interact on a daily basis while nuclear families do so on a weekly basis at least. A participant, for example, indicates that three generations, while living apart, would go for walks three days a week due to concern about the mother’s loneliness at home. Therefore, traditional families all value family interactions either on a daily or weekly basis to enhance their relationships. The elderly reveal that chatting in person or over the phone is the most common way to communicate, and the content of such conversations is mainly about feelings, or new information and interesting daily events. Many traditional festivals or holidays in Taiwan also bring families together. One of the participants mentioned that family members gather once a week to maintain connections; another participant stated that three generations, living close to each other, dine together every day. Many participants share the same interests with their elderly families, such as walking or exercising. One of the participants, for instance, describes that three generations would gather for a walk every two weeks because they do not live together and yet are very close to each other. The participants are satisfied with and wish to sustain their current interactions with the elderly, while over half of the participants wish to have more interactive opportunities. The middle generation shows a great concern for the connection between family members and values good relationships with the elderly. Technology and Intergenerational Socialization. Unlike the elderly in the village, every participant has an experience on the Internet, whether browsing the Web, searching for information, playing games, downloading files, listening to music or the radio, using Facebook, MSN or Skype, writing emails, etc. Getting online has become essential to and habitual in everyday life and is one of their daily activities. The Internet is an important interactive way for the middle generation to communicate. The Internet allows for information sharing, messaging and connecting with different people and is another method of interaction between friends. The participants wish to teach the elderly to use the Internet to have a new way to communicate and to enhance family cohesion. One participant who often travels for business misses his family in Taiwan and states that besides a cell phone, emailing is another way to interact and to care for the elderly.

4 Conclusion In the past, in Taiwan, the majority of aged parents have lived in three-generation households, or with grandchildren in skipped-generation households. However, social changes have been accompanied by rapid changes in the family structure, and the usual living arrangement now includes only the nuclear family. It is interesting that the elderly and middle generation both mention personal wishes. Elders wish that their children spent more time with them and that their grandchildren would visit more often. The middle generation wishes for health, happiness and longevity for the elderly. Based on

An Investigation into the Social Network between Three Generations

285

the interview findings, the elderly who live alone in a senior home expressed fear of loneliness and the desire for companionship. Thus, the elderly actively participate in activities or societies to develop more interpersonal relationships, have diverse lifestyles and sustain their happiness. For young people, social interactions with friends and colleagues enrich their lives and allow for the exchange of new information. Influenced by the traditional culture of filial piety, the elders expect the companionship of children, as the middle generation wishes to spend more time with them. As the Internet and similar technology have developed, network interaction has become widespread. SNSs are deeply integrated in young users' everyday lives. According to interview results, the elderly are gradually using the Internet instead of using pen and paper or telephone to contact their children and grandchildren. With the prevalence of the Internet and applications of Web 2.0-related technologies, interaction between three-generation families is no longer limited to in-person communication. Technology breaks such communication limits and brings more opportunities for contacts and sharing. This also enhances the cohesion between the elderly and their families. Acknowledgements. Special thanks go to the all the participants taking part in the pilot studies, and especially to the residents and staff at Chang Gung Health and Culture Village in Taiwan. This research was funded by the Taiwan National Science Council (NSC 99-2410-H-182-032-).

References 1. Hagestad, G.O.: Transfers between grandparents and grandchildren: The importance of taking a three-generation perspective. Zeitschrift für Familienforschung, 315–332 (2006) 2. Kesslera, E.-M., Staudinger, U.M.: Intergenerational Potential: Effects of Social Interaction Between Older Adults and Adolescents. Psychology and Aging 22(4), 690–704 (2007) 3. De Vos, S., Lee, Y.-J.: Change in Extended Family Living among Elderly People in South Korea, 1970-1980. Economic Development and Cultural Change, 41(2), 377–393 (1993) 4. Arland, T., Chang, M.-C., Sun, T.-H.: Social and Economic Change, Intergenerational Relationships, and Family Formation in Taiwan. Demography 21(4), 475–499 (1984) 5. Domingo, L.J., Asis, M.M.B.: Living arrangements and the flow of support between generations in the Philippines. Journal of Cross-Cultural Gerontology 10(1), 21–51 (1995) 6. Silverstein, M., Cong, Z., Li, S.: Intergenerational Transfers and Living Arrangements of Older People in Rural China: Consequences for Psychological Well-Being. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 61(5), S256–S266 (2006) 7. Francis, D.: Friends from the workplace. In: Hess, B.B., Markson, E.W. (eds.) Growing Old in America, pp. 465–480. Transaction Publishers (1991) 8. Morris, M.E.: Social networks as health feedback displays. IEEE Internet Comput. 9, 29– 37 (2005) 9. Lee, G.R., Ishii-Kuntz, M.: Social Interaction, Loneliness, and Emotional Well-Being among the Elderly. Research on Aging 9(4), 459–482 (1987) 10. Stevens, N.: Combating loneliness: a friendship enrichment programme for older women. Ageing & Society 21, 183–202 (2001)

286

T.-H. Tsai, Y.-L. Ho, and K.C. Tseng

11. Anderson, L.: Intervention against loneliness in a group of elderly women: An impact evaluation. Social Science & Medicine 20(4), 355–364 (1985) 12. Hopman-Rock, M., Westhoff, M.H.: Development and evaluation of ‘Aging well and healthily ’: a health education and exercise program for community living older adults. Journal of Aging and Physical Activity 10, 363–380 (2002) 13. McAuley, E., et al.: Social relations, physical activity, and well-being in older adults. Preventive Medicine 31, 608–617 (2000) 14. Arnetz, B.B., Theorell, T.: Psychological, sociological and health behaviour aspects of a long term activation programme for institutionalized elderly people. Social Science & Medicine 17(8), 449–456 (1983) 15. Caserta, M.S., Lund, D.A.: Beyond bereavement support group meetings: Exploring outside social contacts among the members. Death Studies 20(6), 537–556 (1996) 16. Rosen, C.E., Rosen, S.: Evaluating an intervention program for the elderly. Community Mental Health Journal 18(1), 21–33 (1982) 17. McEwan, R.T., et al.: Screening elderly people in primary care: a randomized controlled trial. British Journal of General Practice 40, 94–97 (1990) 18. House, J.S., Umberson, D., Landis, K.R.: Structures and Processes of Social Support. Annual Review of Sociology 14, 293–318 (1988)

Preschoolers as Video Gamers Alexander Voiskounsky Psychology Department, Moscow State University after M.V. Lomonosov, Mokhovaya str., 11/5, 125009 Moscow, Russia [email protected]

Abstract. The results of a qualitative study of preschool-aged (five and six year old) gamers are presented. Videogames salespeople, parents, caregivers and teachers were surveyed; participating children (N=91) were observed; their behavioral patterns were registered. Adults express either concern or neutral attitudes towards playing video games; few of them are aware of positive and negative aspects, while the others pay little or no attention to the gaming practice. The observation shows that children are in varying degrees dependent on the adults’ support. Preschoolers play educational and entertaining video games: they are interested in both achievements and cognition, react in an animistic manner towards computers and characters in the games. Their problems in comprehending the rules and instructions often result in failures. Gender specifics and age differences are also discussed. Keywords: video games, preschool-aged players, qualitative study, psychology.

1 Introduction Playing video games is a new practice in the development of preschoolers in Russia. Available video games include those for entertainment and/or educational purposes and are either adapted versions of games produced outside of Russia, or games developed by Russian producers. These games are age-specific. The majority of Russian preschool teachers, parents and caregivers have very limited experience in video gaming and can provide limited or no support to the children under their care. Children of five to six years old who play these games are the only age group that is numerous enough and at the same time wholly dependent on their elders (parents, caregivers and teachers). Although younger children are not any less dependent on their elders, gamers who are younger than five years old are less numerous. Older (school age) gamers are quite numerous, but at the same time are able to gain competence with little or no support from adults. Moreover, by virtue of their increased interaction with their peers and older children, they are much less dependent on their elders. This study of gamers was done with Russian children, ages five and six. Because Russian children do not go to school until they are seven, the members of the investigated group can be considered preschoolers. That is to say, not many of them are able to read. As a result, video game rules and instructions need to be conveyed in A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 287–296, 2011. © Springer-Verlag Berlin Heidelberg 2011

288

A. Voiskounsky

an auditory form. To attract the target group, video games need to be rich in color and graphic design. In fact, popular video games produced for preschoolers have to be quintessentially audiovisual. This paper presents an explorative study held in 2008, with conclusions made in 2009. To get data for (longitudinal) comparison, it would be useful to carry on an additional study using the same methodology.

2 Preschool Gaming as a Research Problem Previous studies related to video-gaming habits focus primarily on school-aged children, older teenagers being the largest group observed. While there are some studies related to the use of information technologies (gadgets, computers, game consoles, etc.) by preschool and elementary school aged children, these studies have been done outside of Russia, e.g., a key study outside of Russia was done by Turkle [2]. This study was a groundbreaking study in the late 1970s and early 1980s, i.e. the time when the first generation of preschool-aged children first began to access computers and electronic gadgets in kindergarten and at home. After this pioneering work, nobody carried it out again, although almost all other countries (other than the USA) and many communities were to introduce the access to new technologies to the first generations of children brought up within these countries and communities. This means that the study described in the landmark book “The Second Self” [2] is not only unique, but also unrivaled due to the lack of an adequate counterpart world-wide. Following the above mentioned book, many studies have been done within psychology, education and computer science. These studies refer to the actual processes of the use of computers by young children, including parameters of video gaming, online gaming and the use of consoles. Attempts to organize an irregular and exemplary access of Russian preschoolers to computer technologies started in 1980s, when several enthusiasts established an Association called “Computer and Childhood”. The major activities of this Association were concentrated on the development of educational software, and on organization of classes with this educational software on conveniently located mainframe computers (kindergartens did not have computers at that time), so that children in kindergartens could get an access to them. Later the software was reformatted to be used on desktop computers and laptops. As for the software developed by this Association, it is quite rarely used nowadays due to the fact that PCs became widely available, and new sorts of commercial games became available as well. The newest generations of games for preschool-aged children have a strong entertaining effect and well-developed graphic design; they are rich in colors and often correspond to popular fairy-tales (both related to Russian folklore and to worldfamous fairytales) and/or popular animated movies. The commercial video games can be easily purchased and exchanged, and indeed children and/or their parents/caregivers often exchange games. The games are produced by a variety of software companies – the producers either adapt video games produced abroad or

Preschoolers as Video Gamers

289

produce original products. The variety of newer and older video games on the Russian market is rather wide; diverse types of games are easily available. The variety of video games for preschool-aged gamers changes rapidly: the population of gamers is increasing, and psychological patterns of the gaming processes vary as well. Although the “Computer and Childhood” Association as well as some independent researchers make efforts to study the ways the preschool-aged children play video games [1], the results of these studies have not been updated. Thus, there is no adequate up-to-date information about the most current trends in the use of video games by those children who do not yet go to elementary school. The results of a qualitative pioneer study aimed at investigating current data on the theme are described in this paper. The purpose of the study was to reveal what types of video games are popular within the target group, i.e. five and six-year-old preschoolers; to learn the attitudes – presumably, positive, neutral or negative – of adults (parents, caregivers and teachers) toward the processes of gaming; the likeability of situations in which adults give support to playing children and/or play with them, and to describe some typical playing patterns characteristic for preschool-aged video gamers. The patterns that were being investigated include typical emotional, behavioral and motivational reactions of the target group children.

3 Methods The current study is qualitative and is approached thorough descriptions of the major parameters reported by parents, caregivers and teachers and fixed by the researcher and his assistants as relevant to the processes of preschool-aged children’s playing video games. The parameters are described in this section. The study was carried out in four geographical regions in Russia: (1) in Moscow which has an outstanding position in the country, taking into account the number of citizens who have a high level of education and are economically well-to-do; (2) in a small town in the Moscow region; (3) in Chelyabinsk, which is a large industrial city in the Urals, on the border between the European and Asian territories; (4) in Saransk, which is a less populated and less prosperous industrial town in the centre of the European part of Russia. As many as 91 preschool children participated in the study (31 from Moscow, 34 from Moscow region, 13 from Chelyabinsk and 11 from Saransk), including 43 boys and 48 girls. Each participant played at least two different games: one habitual and one previously unknown to him or her. The observation and registration sessions were performed both at home and at kindergartens; in the latter case, as many as 3-6 children were playing simultaneously; participant used his or her computer, but only one of them was the participant whose behavior was registered during the game session. In 14 gaming sessions, parents or caregivers were sitting next to the children to help them gain experience with a new game; research assistants kept a register of each child’s behavior and took notes with respect to the modes of child-adult interaction and collaboration.

290

A. Voiskounsky

The participants’ computer literacy varied. Nevertheless, all of them were capable of the simplest forms of managing computers and operating computer or console games. The participants’ experience in playing video games was no less than six months; some children had up to two years of experience in video gaming. The number of parents, caregivers and teachers who were surveyed was 52, from all of the abovementioned geographic regions. The number of the retailers and salespeople who replied to the questions is not specified; generalized reports were provided by local research assistants. The methodology included a survey – oral (local retailers and video game salespeople were asked to estimate which games were in highest demand) and paper & pencil (parents, caregivers and teachers filled in a specially developed survey form), as well as an observation and registration of behavioral patterns of five and six year olds who played video games; besides, some of the observed children were presented with questions. The observation and survey were performed by four local research assistants, one from each of the geographical regions. The observation was done using a special registration sheet which was developed and tested in a pre-study session. The pre-study was performed in Moscow exclusively and included consultations with two parents and two teachers, as well as observation and registration of gaming behavior of four five-year-old and four six-year-old children (five girls and three boys). Retailers and salespeople of video games were asked to orally respond as to which preschoolers’ games were most often bought in each of the abovementioned geographic regions. The registration sheet included such parameters as activity/passivity of the participant during the playing session, his or her interest towards the process of playing and its result (i.e., towards a success), attitudes towards losing or winning, readiness and/or eagerness to continue a playing session beyond the time allocated for gaming, peculiarities in the comprehension of the rules of a previously unknown game, the ability to invent correct or partly incorrect rules while playing, fantasy in inventing such rules, the level of absorption and immersion into the gaming process, readiness to ask for support and/or advice from an adult and to use this support/advice, parameters of interactions between a child and an adult, attempts to initiate interaction with a video game environment, types of emotional reactions of the children during the gaming sessions. To learn the intensity of behavioral reactions, the registration sheet was divided into sections corresponding to ten-minute intervals; the duration of sessions varied, but never exceeded forty minutes.

4 Results and Discussion Nine popular games were selected; all of these games were available at every geographical location under investigation. The selection was based on the results of surveys and interactions with game retailers/salespeople and parents/caregivers/ teachers. Two of the games are educational, two are based on internationally wellknown fairy-tales, two games are produced for entertainment, and three of them are based off of well-known animation movies. Thus, the local research assistants were able to register the study participants who played any two of the nine selected games:

Preschoolers as Video Gamers

291

first a familiar game, which the child had played more than once, and second, a new, unfamiliar game. Playing the familiar game was necessary, so that the child would get accustomed to the presence of a research assistant; the rest of the results refer to the reports on playing the second, i.e. previously unknown game. When asked, all the children replied they were fond of the games which were introduced to them, even in cases when the children had very little success while playing. First, we will discuss the results of questioning the adults (parents, caregivers and teachers) who were supposed to report their views in written form and who were partly surveyed orally. The adults’ (parents, caregivers and teachers) registered reports read as follows: the advantages of preschoolers playing video games include gaining competence in the use of computers, being more prepared to acquire cognitive habits and use them efficiently even before going to school, training rapid and effective reactions to screen events, bettering processes of psychological attention and building up logical ways of thinking. Many parents and caregivers (over half of the surveyed) reported that the permission to use the computers is a means of managing and controlling their children (an appraisal for ‘good’ behavior and a mild punishment in case of disobedience). The teachers reject this way of control, insisting that a computer is an educational medium and has nothing to do with disciplinary demands. Moreover, teachers claim that the use of computers should not be a means of manipulation for either children who might acquire habits characteristic of experienced manipulators, or for adults. Neither parents nor caregivers talked about – and supposedly even had such an idea – of manipulation. The negative effects reported by adults are as follows: the parents and caregivers reported concern; firstly, that the visual psycho-physiological functions of their children might be damaged, and secondly, that children might get addicted to Internet gaming. Thirdly, they reported that the after-effects of gaming sessions on their children included nervousness, uneasiness and disobedience, and in the long run – loss of many of their other interests, such as painting or reading books (those children who had previously acquired reading capabilities), or playing with siblings, older brothers and sisters, with friends and neighbors of their age. The parents’ and caregivers’ reports show that over half of them are little or not aware of the content and educational/entertaining characteristics of the games they buy and allow their children to play. The parents and caregivers tend to say that one of the major results of video gaming is that children’s train to amuse themselves independently of the adults’ control and special efforts. About one fourth of the parents and caregivers and all the teachers report they provide support to the preschool-aged gamers by following the process of playing when a new game is being introduced. Otherwise, they believe, the children may have difficulties in understanding instructions of how to play an unknown game, express nervousness and other negative reactions including crying, refusals to play particular games and calling them – as it did indeed happen more than once, according to the adults’ reports – “uninteresting”, “dull”, “stupid”, “incorrect”, “fake”, “incomprehensible”, or “for adults”. More than one in every five adults (parents of caregivers) provided no reply to the question which asked whether he or she was aware of the type of the games his or her child is playing. When randomly asked in oral form, they would typically tell that the

292

A. Voiskounsky

other parent is responsible for the choice of video games and for giving the child support in gaming; this separation of roles in families makes it possible for them to express no interest at all in videogaming practices of their children. Some parents and caregivers did not fill out the survey form carefully, limiting themselves to brief notes and said that their children were “OK”, since they showed no symptoms of either Internet addiction, or other negative after-effects of video gaming. When incapable of formulating positive results of video gaming, the parents and caregivers consider the results either neutral or negative. Marking no signs of visible negative effects, they refuse to provide any further information due to their belief that video games are neutral to the psychological development of their children. No regional differences in replies of parents, caregivers and teachers from the four locations have been revealed, probably due to the fact that the majority of respondents were from Moscow and the Moscow region. Now we turn to the results of the registered behavior of the participating children. Observation and registration of behavior of children show two major types of motivation: first, to achieve success, and the second to be ready to learn new information items. Thus, the major drives were achievements and cognition. Not surprisingly, the former corresponds to the entertaining video games and in a lesser degree to the games related to the animated movies. About thirty per cent of the participating children nevertheless show a vivid tendency to express cognitive interests while playing entertaining video games. Cognitive interests correspond mostly to the problems they face in the process of playing, and sometimes to the contents of the video games. In the both cases the cognitive interest is usually not very deep and does not last long. When they get tired, the participating children tend to lose concentration, express less interest in their achievements, fail to comprehend even the most simple instructions (although it is registered that before getting tired, they were fully aware of these instructions) and the adults’ support messages, preferring sometimes that the adults perform all the actions so that the game session could be finished and the winning result gained. Emotional reactions of the participating children can be characterized as follows: they show arousal and impatience during the whole gaming session; at the end of the sessions they tend to express even more arousal, asking the adults to let them go on playing until they win, and quite often – even after they have won. The children now and then shout out loudly, often cry if they do not get permission to go on playing from the adults. These reactions characterize most strongly – but not exclusively – the children who play entertaining games and express elements of achievement motivation. The intensity of participating children’s emotional reactions (the number of words and phrases spoken, the loudness of shouts, the number of cries) as a rule increased every next ten minutes, with the exception of those children who were not able to understand the rules of the games (the new ones, i.e. newly-bought for them) and got no advantage out of the adults’ support. More precisely, in about one third of the fiveyears-old participants the intensity of emotional reactions decreased after ten minutes of playing, and a little over half of the five-years-olds (including the previously mentioned one-third) and four six-years-olds showed a decrease after twenty minutes; none of them played for more than half an hour.

Preschoolers as Video Gamers

293

It is worth mentioning that the minority of children, in particular – less than ten per cent of the participants, were able to finish the gaming sessions calmly and without showing any arousal, regardless of the scores. Such an attitude towards video gaming might form the basis of children’s reasonable attitudes towards the use of computers and the Internet in general (not necessarily for gaming) in the future (i.e., in their teens and later, when they grow up); supposedly, they are likely to avoid any kind of the computer gamers’ Internet addiction. This latter supposition should be tested in longitudinal follow-ups. Generally speaking, video games resemble a sort of a projective psychological test since research assistants were able to estimate many psychological parameters such as the ability to overcome stress and manage the situations of both success and failure (even in long enough series of failures or winnings), as well as specific patterns of the achievement motivation, the level of development of the self-regulation mechanisms characterizing children’s behavior. Some of the participating children show different attitudes towards the computer as a tool for learning and entertainment, and towards different sorts of video games: they are able to classify the games reasonably enough – for example, as ‘useful’ and ‘recommended by adults’ vs. ‘highly speedy’, or ‘highly emotional’, or ‘for even younger children’. They often express a tendency towards animism (as described by Jean Piaget [2]), i.e. towards perceiving computers, video game disks or gaming consoles as fully alive entities; even more often they believe that particular characters within the game environment are alive (or ‘partly alive’, as three six-years-old participating children would repeatedly tell). More specifically, they tried to establish relations with the characters on the screen, to interact with them and ask to do favors, to praise or blame them for their particular actions. This tendency is stronger in the case of video games’ correspondence to the well-known animated movies, and less strong in the case of educational games. Several participants were able to differentiate a video gaming and an animated movie environment: they would now and then ask why the characters within the gaming environment act differently from what happened in the movies. The young children’s tendency for animism directed towards computers and electronic gadgets, as well as the characteristics given to them, namely being ‘partly alive’, was described earlier by Turkle [3]. The children who participated in the studies reported in Turkle’s book in the late 1970s and early 1980s could (considering their age) become parents of the preschool-aged children who participated in our study; thus unlike newer and newer generations of communication and information technologies (for example, computers, cell phones, web sites and gadgets), human generations differ much less radically, and the earlier gained pioneering results are still correct. Observation of the behavior of participating preschool-aged children shows that they sometimes acquire indirect ways of gaining competence in video gaming. Namely, some of the participating children were fond of looking at older children (in their families and/or in the kindergarten) playing video games. The elders readily replied to the ‘why’-type and ‘what’-type questions asked by the younger children; these replies were often very valuable, since the younger participants were able to use the replies and try to develop new effective (i.e., winning) strategies in their gaming behavior. These indirect types of gaining competence are especially characteristic of

294

A. Voiskounsky

those children whose parents or caregivers give them no support in video gaming and express little or no interest in the specifics of gaming behavior of their children. Lack of the parents’ interest and support give preschool-aged children chances to play video games which do not match their age, and gain competence in the games meant for older children or even adults, including warfare games with cruel episodes and elements of love scenes. Gender differences can be described as follows: boys prefer action games, arcadelike games, race-like games, and entertaining games, while many girls give equal preference to educational and entertaining games. When boys play educating games without external support, they often prefer to explore new strategies by probe and test procedures. Girls are more attentive, compared to the boys, to the instructions and elements of support provided by parents, caregivers or teachers. The adults believe that the boys are too often inattentive to their explanations and easily forget the given recommendations; in reality, however, the boys are fond of working out the strategies themselves, without any external instructions and recommendations more often than the girls. This way of learning to be successful in video gaming is only seldom pursued by preschool-aged girls, although two of the participating six-years-old girls turned out to be extremely successful in working out their own original patterns and strategies leading to a quick and safe success. Meanwhile, six boys of the same age and even one five-year-old boy tended to create original and safe winning strategies in the new games they were just introduced to without any adults’ support. It is worth mentioning that all nine of the abovementioned girls and boys were capable of reading the rules of the games. Age differences between five and six-year-old children should be mentioned as well. Younger (five-year-old) participants encounter a lot more problems in successful gaming than the children who are a year older. The younger ones have problems in comprehending instructions and rules, as well as in remembering and following recommendations given by their parents, caregivers or teachers. They get tired in as little as twenty minutes (while about one fifth of them even earlier, i.e. in ten minutes), and in general, they were only rarely able to play effectively for more than half an hour. The difficulties the younger children came across were partly dependent on the fact that they usually were not capable of reading, while video game producers now and then fail to duplicate in audio format essential explanations, which emerge on the screen. Having little success, the younger children soon tend to lose interest in the result, although they keep their interest in the gaming process, irrespectively of the likelihood of gaining success, winning, getting high scores, etc. This type of behavior was the most likely to occur when the children played entertaining games based on previously known to them animation movies. While the older children tend to imagine the game characters alive, the younger children often express verbal descriptions of what they observe on the screen; these descriptions resemble patterns of a kind of an egocentric speech and often refer to episodes and actions which are irrelevant to the processes of effective gaming. The egocentric speech which some of the participating preschool-aged children produced during videogame playing behavior was not directed at anyone in particular, even

Preschoolers as Video Gamers

295

when the children played in kindergarten groups and/or with an adult who registered the gamer’s behavior patterns nearby. This type of egocentric behavior is characteristic of an egocentric position taken by children irrespectively of their surrounding (i.e., of being alone or in a group of children), as it was described in the classical writings by first Jean Piaget [2] and later by Lev Vygotsky [4]. The latter suggested an alternative (compared to his predecessor’s, i.e. Piaget’ views) interpretation of children’s egocentrism: namely, Vygotsky explained and experimentally proved it to be an essential and necessary step toward the birth and proper development of the ‘inner’ (or internalized) speech and reflective thinking. Once again it is possible to state that the classic results which were first obtained long before the modern information and communication technologies became available, have been repeatedly observed in the current qualitative study.

5 Conclusions The goals of the qualitative study have been fully fulfilled. The results suggest that preschool-aged children who play video games depend to the largest extent on their parents, caregivers and/or teachers. The adults should not restrain from providing their preschool-aged children access to certain video games, but at the same, the adults are expected to give them support, i.e. explain and show in practice how the children should play, what the most needed rules and instructions are, alongside with the ways to use a keyboard, a joystick and/or a mouse, to change and adapt screen size, etc. The role of an adult can be effectively performed by an older child as well, though in this case, the preschoolers may have easy access to video games that are not appropriate to their age and/or capabilities, which may cause negative effects on the preschoolers’ moral and emotional development. The producers of video games for as well as educators of preschool-aged children must provide thorough and detailed instructions both for children (in the audio format) and adults. The adults should have easy access to the full information about the peculiarities of particular games in order to make sure that they are able to make their children aware of these details and thus provide children with effective and timely recommendations. At the same time, the adults should be aware of the advantages and disadvantages of video gaming at preschool age; they should fully realize all the positive and negative aspects of the gaming process. This knowledge may supposedly help adults make sure that their children are not in danger of becoming Internet addicts when they grow up. The younger (i.e., five-years-old) children meet more difficulties playing games aimed at the target group of preschoolers, compared to those children who are one year older. It is important to perform a profound quantitative investigation and to try to find more confirmations of this empirical fact. When and if this is done, it might be reasonable to recommend to producers of games for preschoolers to develop games that would better fit the five-years-old gamers –– maybe simpler versions of the currently existing games. These simpler versions may require a shorter time to beat: the younger children get tired more quickly than the six-years-old children. One more

296

A. Voiskounsky

recommendation which also needs to be empirically backed is that providing somewhat flexible rules for the games, so that the most creative children have a better chance to use such flexible rules in order to work out original strategies of playing these games and getting positive results, i.e. winning. It should be noted that adults must assist the preschool-aged gamers in acquiring certain habits of self-regulation; if this is achieved, the children will be able to end the gaming sessions in due time irrespectively of the gaming scores and the likelihood of winning or losing in a current game session. It is important that this is done when children play both the entertaining and the educational games; moreover, the study has shown that while playing both types of games, the children may find it possible to perform cognitive as well as competitive actions. It is important to note that the current study confirmed several effects that have been previously found – among them an effect that was discovered long before the birth of computers (such as animism and egocentrism). One of these effects (i.e., animism) was noted at the very start of studies of the effects relevant to the beginning of the era when for the first time children got a wide enough access to the use of information and communication technologies. Acknowledgements. The study was supported by the Russian Ministry for Education and Science. The author is grateful to the research assistants who participated in the study and collected data.

References 1. Petku, G.P., Novoselova, S.L.: Kompyuternyj mir doshkolnika (Computer World for Preschoolers). Novaya Shkola, Moscow, (1997) (in Russian) 2. Piaget, J.: The Origins of Intelligence in children. International University Press, New York (1952) 3. Turkle, S.: The second self: Computers and the human spirit. Simon and Schuster, New York (1984) 4. Vygotsky, L.S.: Thought and language. John Wiley and Sons, Cambridge (1962)

GAZO GAZO KUN: Photo-Sharing System Using an Anthropomorphic Photo Frame for Communication Support Takashi Yoshino and Tomoya Matsuo Faculty of Systems Engineering, Wakayama University, 930 Sakedani, Wakayama City, Japan [email protected]

Abstract. Many photo-sharing websites do not actively promote communication among users of the website. We believe that a website that promotes viewing and commenting on shared photographs will better support communication and relationships between people who do not know each other. Therefore, we have developed a photo-sharing system, named GAZO GAZO KUN. This system has an anthropomorphic photo function (a photo agent), which actively supports communication among users. Our experiments compare GAZO GAZO KUN and a photo-sharing website in Japan, similar to Flickr. We found that the effect of promoting viewing and commenting on photographs in GAZO GAZO KUN is higher than that of the photo-sharing website. Keywords: Communication support, photo sharing, agent.

1 Introduction Digital cameras are becoming more and more common around the world. As a result, people have a large amount of photographic data, which has led to an increase in the use of photo-sharing websites. An Internet-related association reported that fewer people use photo-sharing websites than other Internet services1. There are users and photographs that gather a following, and consequently, a high number of comments, but an enormous proportion of photographs are never seen on the photo-sharing websites. According to the research of the famous photo-sharing website Flickr2, a great many users and photographs are never seen or commented on [1]. We found that there are a small number of active exchanges between users in existing photo-sharing websites. However, these websites do not actively promote communication between users. If the website were to promote viewing and commenting on shared photographs, it would better support communication between people who do not know each other. In answer to this need, we have developed a photo-sharing system, named GAZO GAZO KUN. The purpose of this research is to verify the efficiency of our system in motivating users to view and comment on other users’ photographs. This system has an 1 2

CNET Japan (Japanese): http://japan.cnet.com/marketing/20366392/ Flickr: http://www.flickr.com/

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 297–305, 2011. © Springer-Verlag Berlin Heidelberg 2011

298

T. Yoshino and T. Matsuo

anthropomorphic photo function (a photo agent), which actively supports communication among users.

2 Related Work There are user-friendly photo-sharing websites. These sites have been designed to make it as easy as possible to comment on photographs, upload photographs, and make users aware of other website users. For instance, a photo-sharing website called JussPress [2] allows one to easily upload a photograph by dragging and dropping the picture. Then, the user is automatically notified if someone else comments on the photograph. Another photo-sharing website makes it possible to easily upload and comment on a photograph using a portable device [3]. Another example is a photosharing system, PhotoChat, that enables users to write comments directly on the photograph [4]. PhotoChat is a photo-sharing system that promotes communication between users by facilitating the sharing of photographs and comments. Our study aims to promote communication between users in the same way, although our photo agent does so in a different way to those described above.

3 GAZO GAZO KUN We have developed a photo-sharing system, named GAZO GAZO KUN 3 . This system has an anthropomorphic photo function (the photo agent), which actively supports communication among users of the system. The communication support in our photo-sharing system aims to increase the number of times users are viewed and the number of times people comment on photographs. 3.1 Design Policy 1. A photograph to an agent There is one agent per photograph. We call the agent the “photo agent.” As a result, the system is able to support communication relating to the photograph using the photo agent. For instance, we can give compatibility relationships between the appearances of photographs for each photo agent. 2. Promoting communication using the agent We give the interest and affection to users’ agents by the agent’s behavior. For example, each agent talks to the user in a friendly fashion. Moreover, each agent moves blithely. We believe that the agent’s behavior makes the communications support function more effective.

3

The origin of the name GAZO GAZO KUN is Japanese. “GAZO” is a Japanese word that means “image” in English. We repeat the word twice in the name to produce a rhythmic sound. The word “KUN” implies that we hold the person or object in the preceding word in high esteem.

GAZO GAZO KUN: Photo-Sharing System

299

3.2 System Configuration of GAZO GAZO KUN Figure 1 shows the configuration of GAZO GAZO KUN. The server manages all data in the system, namely the user list, the image list, and the comment list. The user uses a web browser to access the system. We have developed the system using PHP and Flash with ActionScript. GAZO GAZO KUN consists of the following pages. − My Page: the user’s own photo agents and links to other users’ photo agents. − Photo Square Page: a display of photo agents.

Fig. 1. The system configuration of GAZO ZAGO KUN

3.3 Photo Square Page Figure 2 shows a screenshot of the Photo Square Page. The page shows one photo agent belonging to the logged in user. The page also shows nine photo agents belonging to other users. The nine photo agents are photographs that the number of browsing are little. The photo agent can have one of five possible faces, each with a different color (red, blue, green, yellow, or purple). The choice of color is made based on the composition color of each photograph. This color is used by the compatibility check function. 3.4 Special Functions for a Photo Agent We developed special functions for a photo agent, which we describe in Table 1. From the results of our experiments in Chapter 4 of this paper, two of the functions were particularly effective. These were the comment request function and the stroking function. We explain each of these two functions in more detail below Table 1.

300

T. Yoshino and T. Matsuo

Login user’ s photo agent

Other users’ photo agents Fig. 2. Screenshot of the Photo Square Page Table 1. Function list for photo agents Functions

Explanation

(1) Photo praise function

The system praises the uploaded photo automatically.

(2) Comment request function

A photo agent (not the user) requests the comment.

(3) Comment report function

A photo agent that receives a comment from another user reports this to the owner of the photo. A photo agent that is stroked by other user reports this to the owner of the photo. A photo agent randomly selects and opens another user's page.

(4) Stroked report function (5) Random movement function (6) Compatibility check function

(8) Self-praise function

A photo agent moves to another user's page, based on the compatibility of the photo agents. The photo is evaluated by the photo agent when a user performs the stroking operation on the photo agent using a mouse. A photo agent praises itself (I’m a good photo!).

(9) User inviting function

The photo agent invites the user to view other users' photos.

(10) Other users’ photo agent display function (11) Few viewed photo display function

Other users’ photos are displayed on My Page.

(7) Stroking function

Photos that have not been viewed very often are displayed on the Photo Square Page.

Comment request function. Figure 3 (a) shows the screenshot of a comment request function. A photo agent asks other users to comment on the agent, not the owner. The purpose of this function is to encourage users to view photographs and contribute comments. After receiving a comment, the photo agent reports this to the owner of the photograph.

GAZO GAZO KUN: Photo-Sharing System

301

Stroking function. Figure 3 (b) shows the screenshot of a stroking function for evaluation. When stroking a photo agent, the shape of the mouse cursor changes to a hand-shaped mouse cursor. After stroking, the stroked photo agent reports this to the owner. (1) Photo agent

Comment me! Please

(a) Comment request function

(2) Hand-type mouse cursor

Please stroke me

(b) Stroking function for evaluation

Fig. 3. Screenshots of comment request function and stroking function for evaluation in GAZO GAZO KUN

4 Comparative Experiments between GAZO GAZO KUN and an Existing Photo-Sharing Website 4.1 Procedure of the Experiments We conducted experiments in order to compare GAZO GAZO KUN and an existing photo-sharing website. The purpose of the experiment is to verify whether GAZO GAZO KUN effectively promotes communication between the users of the system. The subjects are 20 students at Wakayama University. We used livedoorPICS4 as the existing photo-sharing website in Japan. The livedoorPICS is functionally similar to Flickr, a very popular photo-sharing website. We chose livedoorPICS rather than Flickr because the display language of livedoorPICS is Japanese. In the experiment, we divided the 20 subjects into two equal groups, Group A and Group B. − Group A: use livedoorPICS in the first week, and use GAZO GAZO KUN in the second week. − Group B: use GAZO GAZO KUN in the first week, and use livedoorPICS in the second week. We ensured that the subjects in each group did not know each other prior to the experiment. We did this in order to imitate the communication between people using photo-sharing in a real-world environment. We asked each subject to perform the following tasks in both experiments: 4

livedoorPICS: http://pics.livedoor.com/

302

T. Yoshino and T. Matsuo

− Upload five photographs. − Log on to the system once every day. − Upload one or more photographs every day. After the experiment, each subject answered a questionnaire on each system. 4.2 Result of the Experiments and Discussion We show each user's log data and questionnaire. We used the Wilcoxon signed-rank test, with a significance level of 5%. Log data. Table 2 shows the users’ operation status in both GAZO GAZO KUN and livedoorPICS. The average time they spent on GAZO GAZO KUN is 100.8 minutes, and 84.2 minutes on livedoorPICS. The significance probability is 0.22, which indicates that there is no significant difference between the amount of time spent on GAZO GAZO KUN and livedoorPICS. There was also no significant difference in the number of photographs uploaded, the number of visits on My Page, and the number of visits to other users’ pages. In addition, we found that there was no significant difference in the users’ operation status between GAZO GAZO KUN and livedoorPICS. Table 3 shows the number of viewed photographs on GAZO GAZO KUN and livedoorPICS. We found the number of viewed photographs in GAZO GAZO KUN to be more than that of livedoorPICS in both groups. However, the standard deviation of both groups is large. Therefore, we need to encourage inactive users to view more photographs. Table 4 shows the number of photographs commented on in both GAZO GAZO KUN and livedoorPICS. We found the number of comments in GAZO GAZO KUN to be more than that in livedoorPICS. However, once again, the standard deviation of both groups is large. Table 2. The users’ operation status in GAZO GAZO KUN and livedoorPICS GAZO GAZO KUN

livedoorPICS

Significance probability

Average

Standard deviation

Average

Standard deviation

(1) Time spent

100.8

50.5

84.2

43.8

0.220

(2) Number of uploaded photo

15.1

5.9

15.8

6.2

0.938

(3) Number of moving My Page

24.9

14.9

26.1

23.2

0.990

(4) Number of moving other users’ pages

29.6

29.7

24.5

27.1

0.076

GAZO GAZO KUN: Photo-Sharing System

303

Table 3. Number of viewed photographs on GAZO GAZO KUN and livedoorPICS Group A

Average Standard deviation

Group B GAZO GAZO KUN

livedoorPICS

49.2

12.1 9.3

18.9

Significance probability

GAZO GAZO KUN

livedoorPICS

Average

190.8

21.8

Standard deviation

70.7

9.9

Significance probability

0.005

0.005

- Group A: the fist week: livedoorPICS, the second week: GAZO GAZO KUN - Group B: the first week: GAZO GAZO KUN, the second week: livedoorPICS Table 4. Number of contributed comments of photographs on GAZO GAZO KUN and livedoorPICS Group A

Group B GAZO GAZO KUN

livedoorPICS

Average

4.4

2.0

Standard deviation

3.1

1.6

Significance probability

0.027

GAZO GAZO KUN

livedoorPICS

Average

15.3

7.6

Standard deviation

6.3

5.4

Significance probability

0.013

- Group A: the fist week: livedoorPICS, the second week: GAZO GAZO KUN - Group B: the first week: GAZO GAZO KUN, the second week: livedoorPICS

Results of the Questionnaire Table 5 shows the evaluation reasons for “I thought that I want to browse photographs by the action of a photo agent.” We found the photo agents to be effective as a way to browse photographs. Some users were more interested in the photographs than the photo agents. However, we consider this to be a positive tendency, rather than a problem. Table 6 shows the evaluation reasons for “I thought that I want to comment to photographs by the action of a photo agent.” We found that the photo agent aids the commenting process, but that there are some problems in encouraging users to continue to use this feature. Table 7 shows the questionnaire results for encouraging browsing and commenting using the photo agent. The most effective aspect of the photo agent when encouraging browsing is “Other users’ photo agent display function” (Table 7 (10)). The most effective aspect of the photo agent when encouraging users to comment is the “Comment request function” (Table 7 (2)). We also found that the “Stroked report function” to be effective (Table 7 (4)).

304

T. Yoshino and T. Matsuo

Table 5. Evaluation reasons for “I thought that I want to browse photographs by the action of a photo agent” Positive reasons - Because I was interested in the photo agent of other users who came to My Page. - Because I become curious about the photograph chosen by the compatibility check. - I browsed others’ photograph by the function that moved to other user pages. - Because the photo agent had said me, “User X seems to be lonely,” I thought that I went to the user page. Negative reasons - I wanted to see the photograph simply. The movement of the photo agent is unrelated for me. - I don’t have a warm feeling for a photo agent. - It was necessary to see the photograph in the agent clicking because it was small. - I do not think that there was especially an action of the agent who motivates browsing.

Table 6. Evaluation reasons for “I thought that I want to contribute a comment to photographs by the action of a photo agent” Positive reasons - Because I felt that the agent is appealing keenly to me. - Because the photo agent said, “Please comment”, that tempted me to comment. - Because I am glad when it is said by me that there was a comment in my photograph. - Because I have a warm feeling for a photo agent. Negative reasons - I am not controlled whether to comment to the photograph by the action of the photo agent. - I have not understood who other users are. Therefore, I feel uncomfortable to comment the photos. - The agent’s behavior becomes a chance. However, I commented by the content of a photograph. - I commented because the action of the photo agent was unusual at first. Afterwards, I commented only to an interesting photograph.

5 Conclusion We found that there are existing photo-sharing websites that provide a facility for active exchanges between users. We believe that a system that promotes viewing and commenting on photographs facilitates communication between people who do not know each other. We have developed a photo-sharing system, named GAZO GAZO KUN. From the results of our experiments, which compare GAZO GAZO KUN and livedoorPICS (a photo-sharing website in Japan, similar to Flickr), we found the following: 1. The number of viewed photographs in GAZO GAZO KUN is higher than the photo-sharing website. We found that the effect of promoting the viewing of photographs in GAZO GAZO KUN is higher than in the photo-sharing website.

GAZO GAZO KUN: Photo-Sharing System

305

2. There were more comments in GAZO GAZO KUN than the photo-sharing website. We found that the effect of promoting comments in GAZO GAZO KUN to be higher than that in the photo-sharing website. 3. The function of the comment request function by a photo agent is more effective for encouraging users to contribute comments. Moreover, the function of stroking the photo agent was used as an evaluation function of other user’s photographs by the user. Table 7. Questionnaire results for encouraging browsing and commenting when using the photo agent Functions

Stimulation of browsing 0

Stimulation of contributing comment 0

(2) Comment request function

2

11

(3) Comment report function

4

6

(4) Stroked report function

4

4

(5) Random movement function

2

0

(6) Compatibility check function

2

1

(7) Stroking function

0

0

(8) Self-praise function

1

1

(9) User inviting function

3

1

(1) Photo praise function

(10) Other users’ photo agent display function

9

3

(11) Few viewed photo display function

6

0

- The total number of answers is 20.

References 1. van Zwol, R.: Flickr: Who is Looking? In: IEEE/WIC/ACM International, Conference on Web Intelligence (WI 2007), pp. 184–190 (2007) 2. Sit, R.Y., Hollan, J.D., Griswold, W.G.: Digital Photos as Conversational Anchors. In: HICSS 2005, pp. 109–119 (2005) 3. Counts, S., Fellheimer, E.: Supporting Social Presence through Lightweight Photo Sharing On and Off the Desktop, Conference on Human Factors in Computing Systems. In: Proceedings of the SIGCHI Conference on Human factors in Computing Systems, pp. 599– 606 (2004) 4. Sumi, Y., Ito, J., Nishida, T.: Photochat: communication support system based on sharing photos and notes. In: Conference on Human Factors in Computing Systems, CHI 2008 Extended Abstracts on Human factors in Computing Systems Table of Contents, pp. 3237– 3242 (2009)

Is Your Social Networking Privacy Reliant on Intuitive Interfaces? Deborah S. Carstens and Veronica Giguere Florida Institute of Technology, 150 W. University Blvd. Melbourne, Florida 32901, USA {carstens,vgiguere}@fit.edu

Abstract. A study was conducted to assess the level of awareness that users have with regard to the privacy account settings within Facebook, a popular social networking site. A questionnaire was developed and disseminated. The questionnaire findings are discussed and recommendations for improvement to the interface in terms of creating more awareness of privacy settings were identified. The conclusion section discusses improvements to enhance users’ awareness of their privacy settings. Future research was identified in pursuit of increasing intuitive interfaces related to privacy settings. Keywords: Usability, Social Networks, Privacy, Intuitive Interface.

1 Introduction A study was conducted to assess the level of awareness that users have with regard to the privacy account settings within Facebook, social networking site. As Facebook continues to expand and grow, concerns about privacy drive research into user awareness of settings and profile privacy settings intuitiveness. The study consisted of developing and disseminating a questionnaire which was developed through analysis of privacy literature and a careful review of the Facebook privacy account settings. The questionnaire focused on the degree of personal information users share on their profile. The research methodology consisted of developing the questionnaire to assess the intuitiveness of different privacy settings within Facebook, disseminating the questionnaire to survey participants, analyzing the data collected from the questionnaire and summarizing the results for enhancing protection of personal data within social networks. Over 392 participants were surveyed, with 366 of those surveyed having a Facebook account. Recruiting of the participants took place on a University campus offering extra-credit for their participation. The questionnaire findings were gathered and are discussed in the results section. Recommendations for improvement to the interface in terms of potential usability barriers were identified. The conclusion section discusses improvements to enhance users’ awareness of their privacy settings. Future research was identified in pursuit of increasing intuitive interfaces specifically related to privacy settings within Facebook. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 309–318, 2011. © Springer-Verlag Berlin Heidelberg 2011

310

D.S. Carstens and V. Giguere

2 Social Networking Privacy Social network sites (SNSs) such as Facebook are defined by the functions permitted to the users of the community. By creating an individual user profile, the user may establish connections with others in the community and explore these connections as well as connections made by others in the system [1]. As opposed to communities where a user may create a profile and post to a common area such as a message board or a blog, SNSs allow a user to demonstrate the breadth of their social networks through the types of connections that they make with other users. Users do not necessarily seek out strangers with whom to connect; instead, the online interactions tend to reinforce connections that users make offline. In its inception in 2004, Facebook limited its social network to the Harvard community where it was conceived. Soon afterward, it expanded to include colleges and universities. In 2006, Facebook opened membership to include high schools and businesses; being part of a pre-established network was necessary in order to join. More than six years after its beginnings, Facebook now allows users to join, although they often require approval to list themselves as part of a given network. In joining Facebook, a user creates a profile that requires a name, email, gender, and birth date. This is the minimum amount of information that a user may provide, but it alone is insufficient in utilizing Facebook as a social network site. SNSs by definition require that the user establish and connect with friends; therefore, the user must designate a list of friends. The user becomes the center of his or her personal network, with the connections often defined by common interests, beliefs, or past experiences. These friends are then able to post comments on a user’s personal page and view each other’s profiles. With Facebook, groups of users can come together under a common interest or cause, thereby further establishing connections outside of an offline network. University students, the first group to establish a collective identity on Facebook, have used the SNS to establish and maintain social capital that bridges the offline and online worlds. Ellison, Steinfield, and Lampe examined the role that an individual’s Facebook “friends” played in establishing and maintaining a university student’s well-being, concluding that individuals use Facebook to solidify already existing relationships [2]. In a time before the inception of Facebook, Bargh, McKenna, and Fitzsimons concluded that for some individuals, the Internet medium facilitated a representation of one’s true self [3]. During the process of profile creation, the new user is prompted to voluntarily submit information tied to identity, including religious and political affiliation as well as favorite books, movies, and movies. This information made available through a Facebook connection then allows individuals to maintain weak relationships, such as by sending birthday greetings or commenting on a recently posted photo. For this to take place, however, the user must make such information public and available to members of his or her network. Around the same time that Facebook expanded to include users beyond colleges and universities, studies emerged that focused on users’ familiarity and comfort with issues related online privacy. Facebook offers its users the ability to share a wealth of itemized information about themselves; hometown and present locations, academic affiliations from high school and beyond, religious and political affiliations, and employment status can all be listed in carefully controlled fields in their profiles. In

Is Your Social Networking Privacy Reliant on Intuitive Interfaces?

311

these earlier years, Facebook users often joined the SNS in spite of personal privacy concerns; those users who were aware of the potential privacy threats on a large social network relied upon Facebook’s privacy controls to fine-tune the availability of the information, as opposed to simply not listing the information at all [4]. Facebook users have been shown to reveal more information about themselves in profiles when compared to users of another popular SNS, MySpace [5, 6]. These connections are further maintained and segregated through the ability of control that a user can have over the wealth of information that can exist in a given user profile; however, the new Facebook user encounters a default privacy setting that is public. Only by deliberately changing settings can the user opt out of searches among networks and friends as well as larger search engines. Privacy on a large social network site such as Facebook is relative; simply by joining the community, the user already provides a real name, gender, and birthday [7]. Users agree to the Statement of Rights and Responsibilities which includes granting Facebook the “non-exclusive, transferable, sub-licensable, royalty-free, worldwide license to use any IP content” that a user may post on or in connection with Facebook [8]. The permissions granted are subject to the user’s privacy and application settings; at the same time, the Statement of Rights and Responsibilities reminds users that posting and publishing content under the ‘everyone’ setting permits everyone, including persons who may not use Facebook, to view and utilize the information and associate it with the user’s name and likeness. In studying privacy settings of college students on Facebook, Lewis, Kaufman, and Christakis considered a profile to be private if a student altered the default settings to restrict a personal profile so that only a partial profile is available to a nonfriend, same network users or so that no part of the profile may be searched by a nonfriend, same network user [9]. In either case, the user must have taken deliberate action to ensure some measure of privacy within the Facebook community. Results from this study indicated that a student’s utilization of privacy settings was influenced by the number of friends who also maintained private profiles as well as how frequently a user updated his or her profile. As Facebook continued to expand and grow, concerns about privacy drive research into user awareness of settings and profile availability. Less than two years after Facebook’s push onto college campuses, studies showed that while three-quarters of users were aware of privacy settings, between one-third and one-half of users actually used those privacy settings [8, 10]. As recently as 2009, research into privacy issue awareness and the risks versus benefits of using Facebook demonstrated that the majority of users who report restricting their profile still provides a wide range of detailed personal information to friends they may or may not know from offline encounters [11]. This same study indicated that users were more likely to alter privacy settings if they had experienced a negative encounter with Facebook, such as having a profile hacked or an unwanted posting on a profile wall, and concluded that concerns about user privacy should be countered with a drive toward technical, sociocultural, and ethical education about social network site privacy.

312

D.S. Carstens and V. Giguere

3 Methodology A study was conducted to assess the usability of privacy account settings within Facebook, the social networking site. The study consisted of the following steps in planning, conducting, and assessing the interface affiliated with the privacy settings. The form of assessment was through development and dissemination of a questionnaire during the fall of 2010. Step 1: Develop the questionnaire to assess the intuitiveness of different privacy settings within Facebook. The questions were developed through two means. First, a review of literature affiliated with concerns of privacy on social networks provided insight into areas of privacy that can be problematic for users. Furthermore, the researchers went through the Facebook account settings to identify all settings that fall under the category of privacy settings. From the review of the literature and of the privacy settings, the questions for the questionnaire were identified. The first part of the questionnaire focused on the degree of personal information users share on their profile such as posting a picture of themselves as their profile picture, address, birthday, home phone number, work phone number, email address and other pictures of themselves in photo albums. The second part of the questionnaire focused on the type of setting that users have selected when posting their personal information such as selecting which audiences can view their personal information. The different audience settings are everyone, friends of friends, friends only, and other (user has customized their setting so that different friends have different permissions in terms of viewing a user’s personal information including the option to not allow anyone to view personal information). Step 2: Disseminate the questionnaire to survey participants. Extra-credit was given to students enrolled in a freshman level university course. An email was sent out to these students informing them of the opportunity to participate in the survey if they desired resulting in extra-credit being provided. The link to the survey was also provided so that the survey was taken online through the university learning management system. Step 3: Analyze the data collected from the questionnaire. From the analysis of the survey data, the responses for each question were calculated as well as the mean, standard deviation, standard error and 95% Confidence Interval data. The questionnaire findings were gathered and are discussed in the results section. Recommendations for improvement to the interface in terms of potential usability barriers were identified. Step 4: Summarize the results for enhancing protection of personal data within social networks. The conclusion section discusses usability improvements in pursuit of increasing intuitive interfaces specifically related to privacy settings within Facebook. Future areas of research were also identified.

4 Results The results for each question listed in the questionnaire to assess the usability of the Facebook interface in terms of how intuitive is the privacy account settings for users

Is Your Social Networking Privacy Reliant on Intuitive Interfaces?

313

are presented. The range in age of the 392 participants surveyed was between 18-25 years of age. All participants surveyed are college students. Do you have a Facebook account? Of the 392 participants surveyed, 366 or 93.4% of those surveyed, have Facebook accounts. Do you use a picture of yourself as an identifying profile picture? Of the 366 participants surveyed that have a Facebook account, 94.54% use a picture of themselves as their profile picture. Do you post pictures on your Facebook account? Of the 366 participants surveyed that have a Facebook account, 90.16% post pictures such as posting pictures in a picture album or wall photos. Do you post information regarding the address where you live? Of the 366 participants surveyed that have a Facebook account, 19.95% identify their address on their profile. What are your settings for your birthday? Table 1 displays the results broken out by response category. Table 1. Birthday Setting Birthday Setting Response My birthday shows the month, day and year. My birthday shows the month and day, but not the year. I do not show my birthday.

% 55.19 37.16 7.65

Do you post your phone number (home or cell)? Of the 366 participants surveyed that have a Facebook account, 26.23% post their home or cell phone number on their profile. Do you post your work phone number? Of the 366 participants surveyed that have a Facebook account, 3.83% post their work phone number on their profile. Do you post your email address? Of the 366 participants surveyed that have a Facebook account, 59.56% post their email address on their profile. Has your Facebook account ever been hacked? Of the 366 participants surveyed that have a Facebook account, 9.29% have experienced their account being hacked. Keeping in mind the Privacy Settings available in Facebook, what permissions have you set for your own profile for the following nine areas? The nine areas that Facebook allows users to set privacy settings for as well as the responses in terms of percentages of the 366 participants surveyed that have a Facebook account is displayed in Table 2. Keeping in mind the Basic Directory Information setting in Facebook, what permissions have you set for your own profile in Facebook? The seven areas that Facebook allows users to set privacy settings for their basic directory information as well as the responses in terms of percentages of the 366 participants surveyed that have a Facebook account is displayed in Table 3.

314

D.S. Carstens and V. Giguere Table 2. Privacy Settings Permissions

Privacy Settings Permissions exactly as shown in Facebook My status, photos and posts. Bio and favorite quotations. Family and relationships. Photos and videos I’m tagged in. Religious and political views. Birthday. Can comment on posts. Places I check in to. Contact Information.

Everyone %

Friends Only %

Other %

20.77% 25.68% 21.31% 12.30%

Friends of Friends % 11.20% 11.75% 10.38% 17.76%

61.20% 56.28% 59.56% 59.02%

6.83% 6.29% 8.75% 10.92%

20.22%

15.85%

54.37%

9.56%

18.31% 13.11% 10.11% 10.93%

17.21% 10.11% 7.92% 7.10%

56.28% 69.40% 67.49% 63.11%

8.2% 7.38% 14.48% 18.86%

Table 3. Privacy Settings for Basic Directory Information Permissions Privacy Settings for Basic Directory Information Permissions exactly as shown in Facebook Search for me on Facebook. Send me friend requests. Send me messages. See my friend list. See my education and work. See my current city and hometown. See my interests and Pages.

Everyone %

Friends of Friends %

Friends Only %

Other %

66.94%

10.11%

17.49%

5.46%

69.95% 58.74% 42.35% 34.70%

9.56% 12.84% 16.67% 14.75%

16.67% 25.68% 36.07% 45.63%

3.82% 2.74% 4.91% 4.92%

34.97%

12.84%

46.72%

5.47%

32.24%

12.84%

46.72%

5.47%

Do you use the “Places” application in Facebook that allows users to see where you have ‘checked in,’ such as local restaurants, airports or entertainment venues? Of the 366 participants surveyed that have a Facebook account, 11.20% utilize this application. Do you use security questions to help retrieve passwords? Of the 366 participants surveyed that have a Facebook account, 81.42% set up a security question such as the name of a childhood teacher to help these users retrieve a forgotten password. Please select the questions that you commonly use for password retrieval (select all that apply). Table 4 displays the percentages of responses based on the 366 participants surveyed that have a Facebook account. Do you display information that could be linked to a security question (name of pets, mother’s maiden name, hometown, etc.)? Of the 366 participants surveyed that have a Facebook account, 12.57% display information on their profile which could be used to hack into their Facebook (or other) account.

Is Your Social Networking Privacy Reliant on Intuitive Interfaces?

315

After completing this survey, do you plan to alter your Facebook profile settings? Of the 366 participants surveyed that have a Facebook account, 24.86% plan to change some aspect of their Facebook profile settings. Table 4. Questions the 366 Participants Commonly Use for Password Retrieval

Use of the below Questions Commonly Used for Password Retrieval Mother’s maiden name. Name of first pet. Name of childhood teacher. Favorite Sports Team. Name of town where you born. I do not use any of these questions.

% 40.16% 40.44% 9.84% 14.21% 39.89% 36.07%

5 Conclusion The objective of evaluating users’ privacy account settings was to determine if users adjust their privacy account settings within Facebook with the overall objective being to gain insight on whether users find the interface for privacy account settings to be intuitive enough to be making adjustments to their privacy controls. The questions relating to the permissions set for different information on individual profiles varied. The permission settings consist of “everyone” where everyone on Facebook can view particular information about a user. The setting “friends of friends” signifies that all of your friends’ friends can have full access to information where this setting is utilized. The setting “friends only” refers to only those friends of a particular Facebook user can view information where this setting is utilized. Lastly, there is a custom setting where a Facebook user can block certain friends or even nonfriends from seeing any or all aspects of their profile. The custom setting also has the option of posting information for personal viewing only as there is an option where a user can set information to only be viewable only to their self. With regard to birthday settings, over 55% of participants allow for the default setting of showing their birthday month, day and year to be displayed on their profile. Furthermore, over 18% of the survey participants utilize the default setting of “everyone” for viewing the birthday information that user’s display on their profile. Over 20% of survey participants utilize the default setting of “everyone” for viewing their profile status, photos and posts. Over 25% of the survey participants utilize the default setting of “everyone” for viewing their profile biography and favorite quotations. Over 21% of the survey participants utilize the default setting of “everyone” for viewing their family (displays information such as names of your siblings, cousins, etc. on Facebook) and relationship status (married, single, etc.). Over 12% of the survey participants utilize the default setting of “everyone” for viewing their photos and videos where they are tagged which could include those photos or videos posted by a Facebook friend. Over 20% of the survey participants utilize the default setting of “everyone” for viewing their religious and political views. Over 13% of the survey participants utilize the default setting of “everyone” for allowing everyone on

316

D.S. Carstens and V. Giguere

Facebook to comment on posts on their profile. The “everyone” setting means that a stranger can post a comment on an individual’s profile. Over 11% of survey participants utilize the “places application” and over 10% of the survey participants allow for the “places application” to be viewable to everyone on Facebook. Therefore, any person with a Facebook account can view profiles set to “everyone” resulting in a stranger being able to view exactly where a person is throughout the day because the “places application” alerts on a user’s profile that the particular user has checked into a particular airport, restaurant, store or even showing that a particular user has checked into their home. Over 10% of survey participants utilize the default setting of “everyone” for allowing everyone on Facebook to view their contact information which could include any combination of information such as their home address, email address and/or phone number (work, home and/or cell). Therefore, depending on the type of information (photos, contact information, etc.), anywhere from 10% to 25% of the participants surveyed utilize the default setting for information sharing. Taking the lower range of 10% of survey participants suggest that of the 366 participants that approximately 36 of those participants either are not concerned about their information privacy or that these participants are not comfortable in exploring the interface where a user can control their privacy settings. A similar question asked participants about their permission settings with regard to basic directory settings. The results showed that over 66% of survey participants utilize the default setting of “everyone” for being able to be found or rather searched on by other Facebook users. Over 69% of survey participants utilize the default setting of “everyone” for being able to be sent “friend requests” by any person with a Facebook account. Over 58% of survey participants have the default setting of “everyone” for allowing any person on Facebook to send a message. Over 42% of survey participants have the default setting of “everyone” on Facebook to be permitted to view their entire list of friends. Over 34% of survey participants have the default setting of “everyone” to view their education and work information provided. The amount of information users’ list can vary from the name of their school or company, name of degree(s), etc. Over 34% of survey participants utilize the default setting of “everyone” to view their current city and hometown. This could also be an area of concern since some password reset options on email accounts and online marketplaces utilize a question regarding a person’s hometown for password reset options. Over 32% of survey participants utilize the default setting of “everyone” for viewing their interests and pages consisting of fan pages possibly to their favorite restaurants or organizations/clubs. The directory setting information doesn’t reveal home addresses but can also reveal private information such as where a person works. The topic of security questions to help retrieve passwords within Facebook was covered on the questionnaire. Over 81% of survey participants utilize the option to set a security question to retrieve their password. The options for a user is to be asked for their mother’s maiden name, name of their first pet, name of a childhood teacher, name of their favorite sports team, town where they were born, etc. Over 40% of survey participants use their mother’s maiden name as the security question. However, a person might utilize the capability to list out family members on their profile. If their mother uses their maiden name as part of their name display on Facebook, then a person’s mother’s maiden name is no longer private. Over 40% of survey participants also use the name of their first pet as the security question.

Is Your Social Networking Privacy Reliant on Intuitive Interfaces?

317

Individuals on Facebook can post old pictures of their old pets and reveal the name of their first pet which would make that information no longer private. The same holds true for any of the security questions as users may accidentally reveal the information needed to answer their security question which could grant a person access to their profile and ultimately their personal data. Over 12% of the survey participants revealed that they have displayed information on their profile which is directly linked to their security question. As Facebook continues to expand and grow, concerns about privacy drive research into user awareness of settings and profile privacy settings intuitiveness. An interesting response from the questionnaire was that 24.86% of the participants plan to change some aspect of their Facebook profile setting after completing the questionnaire suggesting that the questionnaire itself provided awareness on privacy settings possibly outside of the survey participants’ existing knowledge base. Perhaps a short video clip could be put together so that users when signing up for an account can be provided with information pertaining to the importance of fully utilizing the privacy settings within Facebook. The video clip could also utilize a quick demonstration/tutorial so assist a user with adjusting their privacy setting. Another idea could be the use of a wizard to assist a user as they adjust their privacy settings. Also, an alert of some kind could be given to users when setting their security question regarding the importance of not displaying the answer to their security question on their profile. Future research will be conducted to explore different interface displays where usability testing will be conducted to determine which interface features are the most intuitive for Facebook users. An additional questionnaire will be developed also to gain further insight into the perceived ease of use that users have when exploring the privacy account settings within Facebook. Acquiring more information on the reason users do not utilize the privacy account settings would be useful in driving the design of privacy account settings not only for Facebook but also for other social networking sites.

References 1. Boyd, D., Ellison, N.B.: Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication 13, 210–230 (2008) 2. Ellison, N.B., Steinfield, C., Lampe, C.: The benefits of Facebook “friends:” Social capital and college students’ use of online social network sites. Journal of Computer Mediated Communication 12, 1143–1168 (2007) 3. Bargh, J.A., McKenna, K.Y.A., Fitzsimons, G.M.: Can you see the real me? Activation and expression of the “true self” on the Internet. Journal of Social Issues 58(1), 22–48 (2002) 4. Acquisiti, A., Gross, R.: Imagined Communities: Awareness, information sharing, and privacy on the Facebook. In: PET 2006 (2006), http://privacy.cs.cmu.edu/dataprivacy/projects/facebook/face book2.pdf (retrieved January 12, 2011) 5. Dwyer, C., Hiltz, S.R., Passerini, K.: Trust and privacy concern within social networking sites: A comparison of Facebook and MySpace. In: Proceedings of AMCIS 2007, Keystone, CO., http://csis.pace.edu/~dwyer/research/DwyerAMCIS2007.pdf (retrieved January 12, 2011)

318

D.S. Carstens and V. Giguere

6. Fogel, J., Nehmad, E.: Internet social network communities: Risk taking, trust, and privacy concerns. Computers in Human Behavior 25(1), 153–160 (2009), http://linkinghub.elsevier.com/retrieve/pii/ S0747563208001519 7. Facebook. Facebook’s Privacy Policy (2010), http://www.facebook.com/policy.php (retrieved January 19, 2011) 8. Facebook. Statement of Rights and Responsibilities (2010), http://www.facebook.com/terms.php (retrieved January 19, 2011) 9. Lewis, K., Kaufman, J., Christakis, N.: The Taste for Privacy: An Analysis of College Student Privacy Settings in an Online Social Network. Journal of Computer-Mediated Communication 14, 79–100 (2008) 10. Govani. T., Pashley, H.: Student awareness of the privacy implications when using Facebook. Carnegie Mellon (2005), http://lorrie.cranor.org/courses/fa05/tubzhlp.pdf (retrieved January 19, 2011) 11. Debatin, B., Lovejoy, J., Horn, A., Hughes, B.: Facebook and Online Privacy: Attitudes, Behaviors, and Unintended Consequences. Journal of Computer-Mediated Communication 15, 83–105 (2009)

A Study on Social Network Services Visualization Based on User Needs Young Suk Han, Jong Kyu Choi, and Yong Gu Ji Dept. of Information and Industrial Engineering, Yonsei University, Seoul, Korea {youngsuk,jk.choi,yongguji}@Yonsei.ac.kr

Abstract. In this research, we aim to improve Bridging Social Capital through visualization based on the needs and cultural features of Korean SNS users. According to previous researches, functions such as Expert Search, Connection, and Communication influence on forming Korean SNS users’ Bridging Social Capital. Thus, in this research, by conducting the FGI (Focus Group Interview), users’ needs and opinions were collected. Then, we developed visualization system through FGI results and observed usage in laboratory setting. The usage observation revealed that our visualization system contributes in the improvement of Bridging Social Capital. These results show that Bridging Social Capital could be improved through visualization based on user needs and reveal that features have to be considered in Korean SNS. Thus we expect that this study has academic and practical significance. Keywords: Social Network Services, Visualization, Social capital, Cultural features.

1 Introduction Recently, the development of Information and communication technology (ICT) and increase on the users of smart phone, lead to an accompanied increase on Social Network Sites (SNS) use such as Facebook, MySpace, Twitter, YouTube and Blog Services. With the growth of SNS usage, there are various academic approaches conducted on users’ motivation, users’ behavior, cultural differences[1,2,3], Social Network Analysis (SNA)[4,5,6], and visualization of users’ network[7,8,9]. Particularly, we focused on researches related to cultural difference in SNS use and visualization of users’ network and proceed based on these researches. Previous work related to cultural difference in SNS use revealed that using SNS influence more in forming Bridging Social Capital than Bonding Social Capital in countries like Korea, China, and the United States. According to the study on visualization of users’ network, visualization provides improvements of users’ convenience, playfulness and opportunities to form new relationships. Thus, in this research, we aim to improve the creation of Bridging Social Capital through visualization based on users’ needs and cultural features of Korean. TreeMap layouts were adjusted to visualization system. We conducted observation and interview with participants in our system then, we could find out usefulness, usability and enhancements. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 319–325, 2011. © Springer-Verlag Berlin Heidelberg 2011

320

Y.S. Han, J.K. Choi, and Y.G. Ji

2 Background 2.1 Researches on Relationship between Social Network Service and Social Capital Social capital has been defined as the resources embedded in a social structure which are accessed and/or mobilized in purposive actions[11]. Social capital consists of Bonding Social Capital and Bridging Social Capital. Bonding social capital refers to trusting and co-operative relations between members of a network who see themselves as being similar, in terms of their shared social identity. Bridging social capital, by contrast, comprises relations of respect and mutuality between people who know that they are not alike in some sociodemographic (or social identity) sense (differing by age, ethnic group, class, etc.) [12,13]. Researches applying Social Capital Theory to off-line as well as on-line are being conducted. Recently, there are researches linking Social Capital Theory and SNS. The results from these researches show that using SNS can improve individual Bridging Social Capital in terms of forming and maintain bigger network[14]. Hwangbo, H.[10] defined the cultural differences in SNS use based on relationship between Social Capital Theory and SNS among the countries of Korea, China, and the United States. The functions of SNS were categorized into five types: 1) Expert Search, 2) Communication, 3) Connection, 4) Content Sharing, 5) Identity then, he studied influence of functions in Bonding and Bridging Social capital. As a result, all of the users were more influenced by SNS use when forming Bridging Social Capital, and especially, Expert Search, Connection, and Communication functions have an effect on forming Bridging Social Capital in Korean. 2.2 Researches on Network Visualization Visualization is any technique for creating images, diagrams, or animations to communicate a message. Visualization through visual imagery has been an effective way to communicate both abstract and concrete ideas since the dawn of man[16]. Visualization techniques are the followings: TreeMap, Venn diagram, table, timeline, charts, graphs etc. among them Node-Link layouts is widely used to visualize network structure[6], thus, various researches on SNS visualization adopted this layouts [7,8,9]. Heer, J. and Boyd, D.[7] designed and implemented a visualization system based on Node-Link layouts for end-user exploration and navigation of large-scale online social networks by visualizing users’ network of Friendster.com. It aims to improve playfulness of users’ network exploration and comprehension aspects through supporting search and analysis in Friendster. They conducted observation of system usage in two environments: a public installation at a large party and an informal laboratory setting and the results provide evidence of the system’s usability, capacity for facilitating discovery, and potential for fun and engaged social activity. TreeMap layouts utilize a space-filling algorithm that fills recursively divided rectangle areas with components of a hierarchy[15] and have been adopted in several visualization projects. It has mainly 2 advantages: 1) space utilization is more

A Study on Social Network Services Visualization Based on User Needs

321

effective than tree structure, 2) size and colors of space make possible to compare and distribute immediately 3) It is less complex than Node-Link even though amount of objects were increased[17,20]. A visualization of system that has hierarchical structure like computer file system adopted this layout for these reasons[18,19]. Table 1. Definition of Five Functions of Social Network Service Usage. Hwan Hwangbo (2011).

Function Expert Search Communication Connection Content Sharing Identity

Definition Function to search for people whom the person cannot meet usually and who has professional knowledge by using SNS. Function to have conversation with friends and to convey his/her opinion through SNS. Function to maintain relationship made offline through SNS. Function to share or distribute contents like information that he or she wants to know, music, and video through SNS. Function to express own characteristic and emotion, mood, or recent status through SNS.

2.3 Limitation Most of the researches on users’ network visualization are based on Node-Link layouts for familiarity reasons, and their systems have not reflected users’ needs or opinion. There are scarcely of researches where user testing was conducted. Thus, it is necessary to carry out research of these aspects.

3 System Development 3.1 Focus Group Interview (FGI) Before the development of visualization prototype, we conducted FGI to collect qualitative data of users’ needs and opinions about visualization. It consisted of 18 participants (10 men and 9 women) which were SNS users between 20 and 30 years old and without limitations of their SNS usage in order to remove bias. They were divided into 2 groups of 9 persons. First, we explained concepts of visualization and then, provided instances to capture sufficient background knowledge. The issues of FGI and users’ important comments are the following. • Useful information of SNS to visualize − “It would be useful, if information about not only my SNS activities but also friends’ activities is provided.” − “If search results about specialist in a specific field are presented by visualization, it would be more convenient.” • Visualization layouts − “It is hard to recognize Node-Link layouts due to its complexity.” − “I would like to recognize the information about various SNS activity at a glance.” − “I prefer simple graph types as it is easy to understand”

322

Y.S. Han, J.K. Choi, and Y.G. Ji

3.2 System Structure Layouts. Most FGI participants disliked Node-Link layouts thus, we reviewed other visualization layouts. Moreover, we paid attention to comments that participants want to see information about various SNS activity at a glance and prefer simple graph types and selected TreeMap layouts. TreeMap has three advantages: 1) space utilization is more effective than tree structure, 2) size and colors of space make possible to compare and distribute immediately, 3) It is less complex than Node-Link even though amount of objects were increased[17,20] therefore, TreeMap is appropriate for reflecting users’ comments. Data Set. Twitter in particular, provides poor support for discovery of communities of interest beyond following a public user that represents a company, event or group of interest [22], thus we singled out data set of Twitter. We conducted usage observation and discovered that our system could contribute to improve Bridging Social Capital by supporting these aspects. Visualization system provides these data to users. • Number of Followings / Followers • Number of Tweets / Retweets In order to reflect comments that participants want to compare their SNS activity, our system provides number of followings, followers, tweets and retweets which provide information about users’ Twitter activity. Moreover, our system support to find specialists in users’ interesting fields by search box function (Fig. 1). These data were stored at our server and processed for visualizing speed.

Fig. 1. Main screen of system provides information about Followings, Followers, Tweets, and Retweets in Twitter

A Study on Social Network Services Visualization Based on User Needs

323

4 Usage Observation We observed users’ usage of our visualization system for evaluating in laboratory environment. The observation focused on usefulness and usability also, aimed to finding usage patterns. Participants consisted of 7(6 men and 1 woman) which were active twitter user. First, we did not provide users with any tutorials and asked them to think-aloud while using the system as they saw it. After 10 minutes, we provided explanation for giving functions that they couldn’t’ find. Second, they performed following tasks which provide evidence for improvements of Bridging Social Capital through supporting functions such as Expert Search, Connection, and Communication and are based on Bongshin Lee at el[23]. Then, they expressed their opinions about system through interview. • By keyword searching about one’s interest, ranking of the 5 people who have most tweets. • Following the person who has most tweet with respect to the keyword • Visit the page of the person who has left most tweets • From the followers I have, find the one with most followers. Participants use functions of our system that supporting Expert Search, Communication, and Connection as performing these tasks.

5 Result Twitter has poor support that provides expert search and seeing users’ activity, however, our system supplements these aspects. For that reason, the results of usage observation and interview show that participants are satisfied with the support of our system. As a result of free usage observation, 3 participants among a total of 7 participants said that it is hard to understand the meaning which is provided by space of TreeMap. 5 participants wondered criteria of following and follower space in result screen of interesting keyword search. Moreover, other functions of our system were used without a hitch, also functions such as visiting or following of other users who have a connection with participant or specific keyword were used frequently. In the result of task performing observation and interview, most of the participants replied that visiting and following functions are useful. They performed well without any problem. They also gave some suggestions for the improvement of TreeMap by saying that space filling algorithm is not divinatory and system did not present real time data.

6 Conclusion and Future Study In this research, users’ opinions were collected by FGI and, visualization system was developed based on FGI results, in order to improve Bridging Social Capital of Korean. Then, usage observation conducted while users performed specific tasks in

324

Y.S. Han, J.K. Choi, and Y.G. Ji

our system and we could discover these things. Most users thought that providing information about SNS activity through visualization is useful. Particularly, they enjoyed comparing themselves with others with respect to their community activity. Moreover, they answered their comprehension of community was improved during the performance of various tasks that were proposed for usage observation. We also observed that participants highly utilize functions such as visiting or following based on visualization results. Therefore, we confirmed partially that Bridging Social Capital was improved by our visualization system. It was based on users’ needs and cultural features of Korean. Moreover, this research has other significance as users’ opinions were collected and reflected to system development process for solving problem of previous works. However, there are some limitations in the number of participants and technical issues in visualization system. Thus, in future research we will reflect improvements and conduct comparative study on our system and Node-Link layout which is widely adjusted in visualization system for complementing these aspects. Through comparative study, we will gain and analyze quantitative data then, we will attempt to assess the validity and effectiveness of our visualization system.

References 1. Ellison, N.B., Steinfield, C., Lampe, C.: The Benefits of Facebook “Friends:” Social Capital and College Students’ Use of Online Social Network Sites. Journal of ComputerMediated Communication 12, 1143–1168 (2007) 2. Subrahmanyam, K., Reich, S.M., Waechter, N., Espinoza, G.: Online and offline social networks: Use of social networking sites by emerging adults. Journal of Applied Developmental Psychology 29, 420–433 (2008) 3. Joinson, A.N.: Looking at’, ‘Looking up’ or ‘Keeping up with’ People?. Motives and Uses of Facebook. In: CHI 2008 Proceedings (2008) 4. Mislove, A., Marcon, M., Krishna, P., Druschel, G.P., Bhattacharjee, B.: Measurement and Analysis of Online Social Networks. In: IMC 2007 Proceedings (2007) 5. Henry, N., Fekete, J.D., McGuffin, M.J.: NodeTrix: a Hybrid Visualization of Social Networks. Visualization and Computer Graphics 13, 1302–1309 (2007) 6. Henry, N., Fekete, J.D.: MatLink: Enhanced Matrix Visualization for Analyzing Social Networks. In: Baranauskas, C., Abascal, J., Barbosa, S.D.J., et al. (eds.) INTERACT 2007. LNCS, vol. 4663, part II, pp. 288–302. Springer, Heidelberg (2007) 7. Heer, J., Boyd, D.: Vizster: Visualizing Online Social Networks. In: Proceedings of IEEE Symposium on Information Visualization, InfoVis 2005 (2005) 8. Rodrigues, D., Oakley, I.: Social Circles: A 3D User Interface for Facebook. In: Gross, T., Gulliksen, J., Kotzé, P., Oestreicher, L., Palanque, P., Prates, R.O., Winckler, M. (eds.) INTERACT 2009. LNCS, vol. 5727, pp. 838–839. Springer, Heidelberg (2009) 9. Okamoto, T., Buza, K.: Visualization of Social Interactions in Facebook. In: Demonstration, AAAI International Conference on Weblogs and Social Media, Chicago (2008) 10. Ji, Y.G., Hwangbo, H.: The Influence of Cultural Differences on the Use of Social Network Services and the Formation of Social Capital. International Journal of HumanComputer Interaction 26, 1100–1121 (2010)

A Study on Social Network Services Visualization Based on User Needs

325

11. Lin, N.: Social capital: a theory of social structure and action. Cambridge University Press, New York (2001) 12. Putnam, R.D.: Bowling alone: the collapse and revival of American community. Simon and Schuster, New York (2000) 13. Szreter, S., Woolcock, M.: Health by association? Social capital, social theory and the political economy of public health. Int J. Epidemiol. 33, 650–667 (2004) 14. Donath, J., Boyd, D.: Public Displays of Connection. BT Technology Journal 22, 7–82 (2004) 15. Chen, C.: Information Visualization. Springer, London (2006) 16. Visualization techniques, http://en.wikipedia.org/wiki/Visualization_computer_graphics 17. Mazza, R.: Introduction to Information Visualization. Springer, London (2009) 18. Jarke, J., Wijk, V., Wetering, H.: Cushion Treemaps: Visualization of Hierarchical Information. In: Proceedings of the 1999 IEEE Symposium on Information Visualization, San Francisco (1999) 19. Johnson, B., Shneiderman, B.: Tree-Maps: a space-filling approach to the visualization of hierarchical information structures. In: Proceedings of the 2nd Conference on Visualization 1991, San Diego (1991) 20. Turo, D., Johnson, B.: Improving the visualization of hierarchies with treemaps: design issues and experimentation. In: Proceedings of the 3rd Conference on Visualization, Boston (1992) 21. Rodrigues, J., Ferreira, V., Ferreira, D., Freitas, M.: Twitviz- Exploring Twitter network for your interests (2009) 22. Lee, B., Plaisant, C., Parr, C.S., Fekete, J.D., Henry, N.: Task taxonomy for graph visualization. In: Proceedings of the 2006 AVI Workshop on Beyond Time and Errors: Novel Evaluation Methods for Information Visualization, Venice (2006)

A Virtual World Prototype for Interacting with a Music Collection Jukka Holm1 and Arto Lehtiniemi2 1

Tampere University of Technology, Human-Centered Technology, P.O. Box 589, 33101 Tampere, Finland 2 Nokia Research Center, Visiokatu 1, FIN-33270, Tampere, Finland [email protected], [email protected]

Abstract. This paper studies the idea of using virtual worlds as an interface to access music collections or music recommendation services. In the evaluated 3D prototype, the user moves around a city and interacts with various characters, the looks of which reflect the musical style that they are recommending. In a user study with 41 participants, it was learned that the concept did not fit well to most users’ music consumption habits, and using the prototype for only browsing music was considered to be too tedious and boring. The application should contain more activity and interactive elements, and there should be a clear goal to strive for. To develop the concept further, a better solution would be to modify an existing game or virtual world application by adding music recommendation capabilities. Keywords: Music, visualization, character, avatar, virtual world, playlist, music recommendation, musical genre, user experience, AttrakDiff, Demola.

1 Introduction A virtual world is a persistent and shared online environment, where the users can interact and socialize with each other in real-time [3]. Virtual worlds are used for various purposes including gaming, social networking, education, marketing, and advertisement. In these applications, the user is typically represented with a humanlike avatar that (s)he can move around the world interacting with other characters and different objects. The potential of virtual worlds has also been noticed in the music industry. Second Life [12] includes various clubs and radio stations playing different musical styles, and it has become a popular place for live concerts where the performance of a reallife artist is streamed to the virtual world and performed by representative avatars. Since 2006, major artists such as U2 and Duran Duran have performed in Second Life [1]. In the academia, avatars have been used to access music recommendation services and to summarize users’ musical preferences [5, 9, 10]. In this paper, we propose the concept of using virtual worlds as an interface for accessing music collections or music recommendation services. The virtual world can be, e.g., a city consisting of numerous suburbs, buildings, vehicles, and characters. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 326–335, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Virtual World Prototype for Interacting with a Music Collection

327

Each city element works as an interface to a song or album in the music collection or some specific music player function. For example, different looking characters could recommend different type of music from the service, speeding a car could fastforward a song, shooting a character could delete a song, and each bar or discotheque could represent a different musical genre. The world could also be dynamic; old buildings (representing rarely listened albums or genres) could be torn down, new bars open, and so on. To study the idea in practice, a simplified off-line version of such an application was developed for the PC platform. In the prototype, the user moves around a city and interacts with various characters to receive new music recommendations. The look of each character reflects in some way the musical style that it is recommending. This paper discusses the prototype implementation in more detail, presents the results of a user study with 41 participants, and proposes a broad range of ideas for the future development of the concept.

2 Prototype The virtual world prototype was implemented as a Demola [4] student project for the PC platform. As creating an extensive virtual world (such as the one described in the Introduction) would have required significant resources, our approach was to make a simplified off-line prototype to test the feasibility of the overall concept. The prototype was implemented as a Flash application to enable the use of different platforms including personal computers and mobile devices such as Nokia N900. As suitable mobile devices or Apple iPad were not available at the time of implementing the prototype, the main test platform was a HP Pavilion tx2500 touch-screen laptop. In the prototype (Fig. 1), the user moves around a city using the arrow keys and interacts with various characters by pressing the enter key. While each character can move freely within the city limits, they spend most of their time in their natural surroundings (e.g. the heavy metal character is typically located close to the black and

Fig. 1. Screenshot from the prototype

328

J. Holm and A. Lehtiniemi

metallic buildings). When interacted with, each character recommends new music from a genre that is somehow related to the looks of the character. The user is able to listen to the songs and add them to his/her playlist by pressing the tab key. The user can also enter any house in the city, but due to project’s time limitations all houses contain the same room (a game hall) and the user cannot do anything in the room. To simplify the prototype implementation, the number of characters was limited to eleven. From our research’s point view, the biggest problem with the original Demola prototype was that several characters had visual characteristics that represented other musical genres than the playlist linked to the character. Thus, we decided to modify some of the characters and change most of the music. Due to implementation details, only minor graphical modifications such as changing colors and adding symbols on top of the characters were possible. Unfortunately, we were not able to change the body shapes, clothing, and other such features selected by the students. Fig. 2 shows the eleven modified characters. These included (from top left to bottom right): 1. female with a cyan dress (representing dance music), 2. female with a red and black dress (grunge music), 3. female with a pink dress (pop music), 4. female with black leather clothes, strong make-up and skull decorations (heavy metal music), 5. “Raiden” – male character with Eastern and ninja elements (world music), 6. main character from the Blade movie (soundtrack music), 7. Homer Simpson (soundtrack music), 8. Megaman (game music), 9. Astroboy (cartoon music), 10. “Princess Li” – female character decorated with Eastern symbols (world music), and 11. “Question mark blade” – Blade with different clothes and question marks on the chest and the forehead (random or miscellaneous music).

Fig. 2. Characters used in the prototype

3 Research Method To study how well the prototype works in practice, we arranged a user study with 41 participants. The selected research method was a combination of observation, semistructured interview, and questionnaires. The prototype was tested in various locations such as participants’ own homes, authors’ homes, as well as the premises of Tampere University of Technology and Nokia Research Center. Each session lasted

A Virtual World Prototype for Interacting with a Music Collection

329

for about 30 minutes. All participants tested the software for the first time, and they were not informed about how the prototype works or what they should do with it. In the beginning of a session, the users had to fill in a short background information questionnaire. After that, the users were able to use the software freely, move within the city, and interact with the characters. When meeting a character, the users were asked to describe it with a couple of words and describe what type of music comes to their mind when looking at the character. After the user had listened to the associated playlist, (s)he was asked to rate how well the music corresponded with the looks of the character on a seven-point Likert scale. Due to the time limitations of the user study, only four characters per user were studied. After rating the characters, a short, semi-structured interview was conducted. The participants were asked to describe their first impressions on the prototype, and various quality aspects of the prototype were studied by asking questions and filling in a questionnaire. Both hedonic and pragmatic aspects [6] were studied, and nine users also filled in the AttrakDiff questionnaire [1]. The Finnish participants were selected using convenience sampling from varying age groups and education levels. 78% were male and 22% female. 7% (3 participants) were between 12 and 17 years old, 12% (5) were 18-30, 56% (23) were 31-40, and 25% (10) were 41-55. All participants from the 12-17 age group were male, and the average age was 34.5. 61% were university graduates and the rest were split between various types of education such as elementary school, college, and university students. 7% were professional and 49% hobby musicians. 51% of the participants considered listening to music as their hobby. All except two listened to music at least once a day, and the average time per day was 2.1 hours. During listening, 20% did not usually do anything else but concentrate on the music. Popular listening contexts included home (55%), car (45%), school, work, sports, and the like. The most popular genres included pop (liked by 85% of the participants), rock (81%), metal (63%), soul, rnb & funk (61%), and alternative (56%), while the least popular genre was gospel (5%). 43% of the participants searched actively for new music to listen to. The most common sources included the Internet (65%), friends (63%), magazines (33%), radio (33%), and Spotify (20%). 88% had used YouTube for listening to music, 51% had used Spotify, 32% Last.fm, and 29% iTunes Genius.

4 Results Based on the user study, we were able to answer our initial research questions and also received several ideas for the future development of the concept. In the following, the results are discussed in a detailed manner. 4.1 Quality Aspects In the end of the interview, various quality aspects of the prototype were studied by asking questions and filling in a questionnaire. In the questionnaire, the participants had to rate nine different aspects of the software on a seven-point Likert scale (1=totally disagree, 7=totally agree). Fig. 3 shows the results using boxplots. Unless otherwise noted, the ratings did not correlate with participants’ age, gender, or level of experience with music recommendation systems. We also experimented with the Chi-Square test, but the conditions for the test were not met in any of the cases.

330

J. Holm and A. Lehtiniemi

Fig. 3. Boxplots for the quality aspects of the prototype

On the positive side, the participants considered the concept to be a novel way to present music (median 6.0). During the interview, comments such as “Great idea” and “A promising concept but very draft implementation” were common. Most participants also felt that the prototype was easy to use (median 5.0). The features were easy to find, and in general help files or tutorials were not needed. However, some participants did not fully understand the basic idea before it was explained to them. Two persons thought that the application was a first-person shooter game and were wondering where all the action was. 44% of the participants considered the prototype to be more fun than the traditional music players. However, for the statement “The program was fun to use”, the median fell on the “negative” side (3.0). There was moderate correlation with age (Spearman’s rho=-0.44, significant at the p<0.01 level (2-tailed)), i.e., older participants did not consider the prototype to be as fun as the younger participants. The graphics were considered to be old-fashioned (“retro feeling”) or simply bad. Many participants also felt that the female characters were too sexual. As a result, the median for “The program looked impressive” was as low as 2.0. For the statement “The program had an extensive set of well-implemented features”, the median was also 2.0. When asked about the negative sides of the prototype and/or the idea, the following issues came up: difficulties and required time in finding characters representing desired music, “nothing to do” while moving in the city, and the lack of textual search function (requested by 44% of the participants). In addition, 88% felt that this type of application does not fit their current music consumption habits. When listening to music, they either just want to concentrate on the music and do nothing else, or the music plays in the background while something else (e.g. reading or driving) is being done. Several participants commented that they would not use this type of music player even if the implementation was much better. Finding music using characters was considered to be too tedious (“It takes a lot of time to listen music with this approach: too much running, too little to do, too few people”). As a result, the median rating for the statement “I was able to start the playback of desired music quickly” was only 2.0. During the interview, it became clear that people were not willing to spend extra time looking for music. Instead, the process should be as fast and easy as possible.

A Virtual World Prototype for Interacting with a Music Collection

331

While the median for lasting appeal was as low as 2.0, the results correlated clearly with age (Spearman’s rho=-0.55, significant at the p<0.001 level (2-tailed), i.e., younger participants were more interested in using the prototype longer than the older ones. Only 22% mentioned that they would like to use the prototype for a longer time. When asked the question “Would you like to use this type of application rather than traditional music player applications?”, only one participant gave a positive reply. 37% answered that they could use a virtual world to complement other applications. As this is in conflict with the 22% of lasting appeal, some participants may have referred to an imaginary future product instead of our prototype. The interest in using virtual worlds to complement traditional music players decreased with age (Spearman’s rho=-0.55, significant at the p<0.001 level (2-tailed)), and some participants even stated that they were too old for using this type of software. The medians for “My overall grade for the program” and “My overall experience on using characters to listen to music” were both 3.0. However, the mean for the latter was 0.4 higher as some participants considered the idea to be slightly better than the actual implementation. The overall grade correlated clearly with age (Spearman’s rho =-0.52, significant at the p<0.01 level (2-tailed)) and somewhat with gender (rho=0.32 with p<0.05). In other words, older participants and women did not like the prototype as much as younger participants or men. Moderate correlation between the overall experience and age was also found (rho=-0.45 with p<0.05). In addition to our own questionnaire, nine users also filled in the AttrakDiff questionnaire [1]. AttrakDiff is a tool for evaluating the pragmatic and hedonic quality of products, and it consists of 28 evaluations on a seven-point semantic differential scale. According to the used online tool, our prototype was considered to be simple, inventive, creative, and innovative. In terms of pragmatic and hedonic quality, the user interface was rated to be neutral and it just met ordinary standards. The results also suggested that “product’s attractiveness value is located in the average region” and “the overall impression of the product is moderately attractive.” The results seemed to be well in line with the results derived from the interview and own questionnaire. As the use of the AttrakDiff method did not seem to bring any extra benefit to our study, other participants did not fill in the questionnaire. Also, we found it quite tedious and unpractical to translate the terms used in the tool to those participants who were not fluent with English. 4.2 Characters After the users had listened to some of the songs linked to a character, they were asked to rate how well the looks of the character matched with the music using a seven-point Likert scale (1=no match, 4= neutral, 7=perfect match). As mentioned earlier, we had very limited means to affect the looks of the characters in the development phase. This was also reflected in the results (Fig. 4), which were not as promising as in our two avatar related studies [9, 10]. Before listening to the music, the users were also asked to describe the character with a couple of words and guess what type of music it could possibly represent. These first impressions revealed many reasons for the poor ratings of some characters.

332

J. Holm and A. Lehtiniemi

Fig. 4. Boxplots for how well the music and the characters match. In the case of each character, N refers to the number of the participants who met and rated the character.

The winner of the study was the heavy metal character, which relied on findings from our earlier research [7, 8]. Most participants (39/41) met the character immediately after starting to use the prototype, and felt that the looks matched well with the songs that the character was mapped to (median 6.0). Ratings less than four were explained by some participants who considered the character to be quite futuristic (or even a cyborg), and thus associated it with industrial, electro, or gothic music. Artist Marilyn Manson was mentioned quite a few times. Several participants would have preferred a male character to represent metal. To improve the design, the character should have looked more down-to-earth or like a male black metal fan. In fact, the latter approach was taken in [9] and [10] and proved to be very successful. Another successful character was the pink woman representing pop music (median 5.0). One reason for this might have been the color selection, as in [7] it was found out that 48% of the participants associated pink with pop. The single outlier value is explained by one person who associated the character with clubs and trance music. For unknown reasons, the cyan female character was often associated with pop music and the grunge woman with dance, techno, and hip-hop. The single outlier value for grunge is explained by a 12-year participant who recognized that grunge bands have often worn such shirts. The short dress was also associated with supermodels, movie stars, and RnB videos. A major complaint for the pink, cyan, and grunge female characters was the body shape, which gained undesired attention from both genders. Many participants felt that the characters were too sexual (“Why does she have such big tits?” female 38), had silicone implants, and/or looked like prostitutes. Due to the exaggerated sexuality, many participants did not pay enough attention on the dress colors and other visual cues, and some even thought that the same character had just changed her dress. To improve the recognition rates, the body shapes should be less extreme and differ clearly from each other. Also, the characters should stay closer to their natural surroundings (e.g. pop fans outside a pop club) instead of moving around the city. The main problem with the movie, cartoon, and game characters was that many participants did not associate them with soundtrack music. Instead, the answers were based on the looks of the character, or the participants imagined what type of music the character could listen to. Thus, Homer Simpsons was associated with reggae, funk, adult rock, 70’s hits, children’s music, country, and “funny” music, Astroboy was mapped to techno, punk, and funk, and Megaman to techno, pop, and metal. Blade was associated with genres such as metal, gangsta rap, hip-hop, and punk. The

A Virtual World Prototype for Interacting with a Music Collection

333

main reasons for the metal association were the leather clothes and the black color, which is typically associated with metal music [7]. Raiden was commonly described with terms such as martial arts, ninjas, kung-fu, East, Asia, Japan, and Vietnam. However, as Raiden’s world music playlist included also African songs, some participants decided to give it a negative rating. One participant (male, 33) commented that “I understand what you are after here, but I have to give a bad rating as the playlist includes African music.” On the other hand, many participants considered Princess Li to be quite confusing because the looks were a mixture of various cultures. In addition to world music, Li was associated with techno, female rock, hip-hop, and blaxpoitation soundtracks. As a conclusion, one should not try to represent world music with a single character but there should be a different character for each country or region. The “Question mark Blade” character was a complete failure (median 2.0). While the question mark is often associated with unknown music [8], it did not work well in this context. Participants focused more on the other properties of the character, and many did not even notice the question mark. The character was associated with various music styles such country, blues, rap, and funk, and those two participants who recognized the character just thought that it represented Blade as an old man.

5 Ideas for Further Development During the interview, we also received several ideas for the future development of the concept. To start with, most participants felt that there should be more interactive content, characters, and things to do. Examples included fighting, destroying, driving, entering buildings, customizing one’s own avatar, hidden features, and playing minigames. Many participants were a little confused regarding to what to do in the city, and it was said that here should be a clear goal to be strived for while listening to the music. The virtual world could also be extended to include areas outside the city. The city could be enhanced with different types of ambient sounds (e.g. speech and traffic sounds) and background music. These sounds could change when the user moves around the city and comes closer to the points of interest. Music played inside a club could get louder as the user walks closer, and there could be background music in shopping centers, elevators, and other places where such music is typically played. The relationship between the characters and the music should be clearer, and the characters could show more hints on which musical style they represent. There could also be a setting for adjusting which musical genres are visible in the city. By default, different suburbs could represent different musical styles and characters could hang around outside the clubs. The desired characters should be easier and faster to find. One way to accomplish this would be a map indicating the current location of the user and the movement of selected characters. Quick shortcuts were also seen as important for jumping to the next character in the city and for finding certain characters. While moving in the city, the user could toggle different view modes such as FPS or birds-eye. Character playlists should be dynamic and change (for example daily) to provide different songs from the same genre. Certain areas of the city (such as the central square) could provide dynamically changing top lists of music listened to in the city.

334

J. Holm and A. Lehtiniemi

By mixing different characters, users could create cross-genre playlists. The characters could also represent real people such as known artists with their own music recommendations. The interaction with the characters should be improved. There could be a possibility to talk with the characters, and speech bubbles could appear on the top of them. Some participants were also interested in chatting with the other users of the system. It was proposed that the interaction could resemble the interaction between the characters in Habbo Hotel and Sims. Some participants felt that the prototype did not include enough information on the songs. Thus, when a song/artist has been selected and is being played, the system could retrieve additional information from Wikipedia or other web page. The virtual world could also give access to exclusive content such as remixes of popular songs.

6 Conclusions and Future Work In this paper, we studied the idea of using a virtual world as an interface for accessing music collections or music recommendations services. In the simplified prototype implementation, different characters acted as an interface to access the music. The looks of each character reflected the musical genre that they were mapped to. The prototype was evaluated by 41 persons. Due to the poor implementation of the prototype and the use of only Finnish participants, the results are only indicative. Still, based on our understanding, it is not worth the time and money to start implementing game or virtual world –like features on top of an existing music application. Firstly, exploring music collections using the prototype did not fit well with the music consumption habits of the participants. Most people prefer to concentrate only on the music, or the music plays in the background while something else is being done. Finding new music should be as fast and easy as possible, and thus using the prototype was considered to be too tedious. Secondly, providing only music recommendations in a virtual world environment does not provide enough entertainment and value for the users. To make a musicrelated virtual world application more appealing, it should also include a lot of action (e.g. fighting and driving) and interactivity in addition to the basic music player features. There should also be a clear goal to be strived for while listening to the music. During the user study, we received several ideas for the further development of the concept. Should they be implemented, the prototype would start to resemble more and more games such as Grand Theft Auto, virtual worlds such as Second Life, or social networking services such as Habbo Hotel. Thus, rather than building the virtual world from scratch on top of a music player application, a better solution would be to enhance an existing game or virtual world with music recommendation and searching capabilities, This solution would fulfill the users’ need for more action and interactivity, and minimize the financial risks of developing a new type of service. The resulting application could also evoke the interest of gamers and people that are not active music listeners. In addition to the music related features, the user experience of such an application would depend on the genre and implementation details of the original application. As

A Virtual World Prototype for Interacting with a Music Collection

335

one cannot speculate how such a combination would be perceived by the users, this would be an interesting topic for a future user study that should preferably be done in collaboration with the developer of the original application.

References 1. Andrews, R.: Second Life Rocks (Literally), http://www.wired.com/techbiz/it/news/2006/08/71593 2. AttrakDiff, http://www.attrakdiff.de 3. Book. B.: Moving Beyond the Game: Social Virtual Worlds. In: State of Play 2 Conference (2004) 4. Demola, http://www.demola.fi 5. Haro, M., Xambo, A., Fuhrmann, F., Bogdanov, D., Gomez, E., Herrera, P.: The Musical Avatar – A Visualization of Musical Preferences by Means of Audio Content Description. In: Audio Mostly. ACM Press, New York (2010) 6. Hassenzahl, M., Platz, A., Burmester, M., Lehner, K.: Hedonic and Ergonomic Quality Aspects Determine a Software’s Appeal. In: CHI 2000 Conference on Human Factors in Computing Systems, pp. 201–208. ACM Press, New York (2000) 7. Holm, J., Aaltonen, A., Siirtola, H.: Associating Colors with Musical Genres. Journal of New Music Research 38, 87–100 (2009) 8. Holm, J., Holm, H.: Associating Icons with Musical Genres. In: International Conference on Internet and Multimedia Systems and Applications. ACTA Press, Calgary (2008) 9. Holm, J., Lehtiniemi, A.: Evaluating an Avatar-Based User Interface for Discovering New Music. In: 9th International Conference on Mobile and Ubiquitous Multimedia. ACM Press, New York (2010) 10. Holm, J., Siirtola, H., Laaksonen, L.: Associating Avatars with Musical Genres. In: 14th International Conference on Information Visualisation. IEEE, Los Alamitos (2010) 11. Lehtiniemi, A.: Evaluating SuperMusic: Streaming Context-Aware Mobile Music Service. In: International Conference on Advances in Computer Entertainment Technology, pp. 314–321. ACM Press, New York (2008) 12. Second Life, http://secondlife.com

A Supervised Machine Learning Link Prediction Approach for Tag Recommendation Manisha Pujari and Rushed Kanawati LIPN, CNRS UMR 7030, Universit_e de Paris13, F-93430 Villetaneuse, France [email protected]

Abstract. One of the primary goals of tag recommendation approaches is to deal with the problem of ambiguity of tags in a folksonomy by helping users to select the most appropriate tag to annotate a resource. We propose in this work, an original approach for tag recommendation applying a link prediction using supervised machine learning. Given a user (target user) and a resource (target resource) the proposed algorithm computes a list of tags best suited for recommending target user to annotate the target resource. It first searches for users similar to the target user. Then a link prediction approach is applied on a temporal sequence of bipartite graphs coding the history of tagging of retrieved similar users. This results in obtaining one or more lists of tags for the target resource or similar resources. These lists are then merged using a list aggregation method to get a single list of tags for recommendation. The first prototype of this approach is described in this article. Preliminary results of applying the proposed approach to real dataset extracted from the bibliographical folksonomy CiteULike1 show the validity of the approach. Keywords: Tag recommendation, folksonomy, Folksonomy, Tag recommender, Tag Prediction, Social Network Analysis, Supervised Machine Learning.

1 Introduction A Folksonomy is a social resource sharing system where users can store and annotate resources by keywords (or tags) that can be used later to access tagged resources. The practice of tagging has become very popular in recent years. May it be Flickr2 for sharing of images or web sites like CiteULike, Delicious3 or Bibsonomy4 etc. where shared resources are bibliographic references, the users all over the world are very active to annotate resources with tags of their choice. One of the key success factors of folksonomies is the freedom of users to choose their own tags for resources they annotate. However, since tags are entirely user defined, most folksonomies suffer from the tag ambiguity problem where the same tag is used differently to index 1

http://www.citeulike.org http://www.flickr.com 3 http://www.delicious.com 4 http://www.bibsonomy.org 2

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 336–344, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Supervised Machine Learning Link Prediction Approach

337

different resources or same resources can be indexed by different tags [17]. Figure 1 shows a snapshot from Flickr site where we can see the presence of multiple tags for a single image. This may have resulted due to the fact that different users annotated the image with different tags of their own choice or one user used different tags at different point of time, to annotate the same image.

Fig. 1. A snapshot from flickr

Tag recommender approaches try to cope with tag ambiguity problem by helping a user to select the most appropriate tag to annotate a given resource. Different approaches are presented in the scientific literature for tag recommendation computation [11, 12, 25, 19]. These can broadly be classified into two main classes: content-based approaches and topological approaches. Roughly speaking, in the first type of approaches tags are induced from the content of the target resource (or similar resources), while in topological approaches, tags are computed by analyzing the tripartite graph describing a folksonomy. As far as we are aware, all topological approaches consider only static graphs; aggregating the history of the evolution of the folksonomy in one graph. In this work we propose an original approach taking into account the dynamic nature of a folksonomy. The proposed approach is based on using a link prediction approach developed for link prediction in bipartite graphs applying supervised machine learning approaches [4].

Fig. 2. A folksonomy structure

338

M. Pujari and R. Kanawati

Actually, a folksonomy can be represented as a tripartite graph linking objects from three different sets: users, tags and resources. Three different bipartite graphs can then be deduced by projecting the initial graph on each possible couple of sets. Figure 2 shows a graphical structure of a folksonomy. The reminder of this paper is organized as follows. In section 2, we present a short state of the art of tag recommenders in folksonomy. Our approach, named Liptar5 is presented in section 3. Experimentations and preliminary results are given in section 4. Finally we conclude in section 5.

2 Related Work Current approaches for tag recommendation can be classified into two broad categories: content-based approaches and topology-based approaches. One example of content-based approaches is the work proposed in [20]. In this work recommended tags are generated from the content of the resources. Every resource provides different information so the individual scores are calculated based on that. All the scores are then aggregated to get a global score for each tag. In other words, the results are combined based on weighting factors which are determined experimentally by comparing generated and expected tags based on the available training data. Tags with top five highest scores are selected for recommendation. Another content-based approach is proposed recently in [19]. The experiments are done on the data from social bookmarking service Delicious. In such services, tag recommendation is available only for pages with tags specified by other users. The proposed approach is based on the observation that similar web pages usually have same tags. Therefore, each web page can share their tags with similar ones. The propagation of a tag depends on its weight in the originating web page and the similarity between the sending and receiving web pages. The similarity metric between two web pages is defined as a linear combination of four types of cosine similarities, taking into account both tag information and page content. In [18] authors propose yet another content-based approach where recommendation computation is made in a three-step process: tags are first extracted from resource titles. The set of potential recommendations is then extended by related tags proposed by a lexicon based on concurrences of tags within the resource posts. In the third and final step tags are filtered by user's personomy: a set of tags previously used by the user. The main advantage of a content-based approach is that it allows recommending new tags: tags that have not yet been used by the users before. The major limitation of this kind of approach is that, although they may provide very relevant tags, but this relevance is only with respect to involved resources. They do not take into account user's choice. Moreover these approaches are only effective if the resources contain rich content of information that can be automatically extracted. In the category of topology-based approaches, one of the prominent works is given in [12]. In this work authors compare a number of recommendation techniques like collaborative filtering, PageRank and its modified version for folksonomy known as FolkRank. They show that the FolkRank based recommender outperforms the other 5

LInk Prediction based Tag Recommender.

A Supervised Machine Learning Link Prediction Approach

339

two approaches. They propose two tag recommendation algorithms: an adaptation of user-based collaborative filtering and a graph based recommender built on the top of FolkRank. Tests were performed on the dense core of folksonomy, so it might not be very representative. Moreover they do not take into account the dynamic nature of a folksonomy. Another work is given in [26] in which authors propose a tag recommendation algorithm based on an integrated diffusion on user-item-tag tripartite graphs. This work gives a start point for design of hybrid algorithms making use of tag information. Authors propose an algorithm using both user-resource relations and the collaborative tagging information. They emphasize on the fact that two resources, sharing many common tags, have greater probability of being closely related in content. The algorithmic performance was measured using recall, area under ROC curve, diversity and a metric for novelty that quantifies the capacity of an algorithm to generate novel and unexpected results. They conclude that the use of tag information can significantly improve the accuracy, diversification and novelty of recommendation.

3 The Liptar Approach Figure 3 illustrates the general outlines of the tag recommendation cycle applied in the Liptar system. The system takes as an input a target user ut and a target resource rt. The goal is to compute a list of tags that suits best the user ut to tag resource rt. The cycle is structured in three steps: 1. First, the approach computes a set of k top similar users U that are similar to ut. A variety of user similarity metrics can be used for this purpose. In the current prototype the similar users have been found by simple application of k-nearest neighbors in the bipartite graph connecting users to tags used by them. 2. Each user u ∈ U is associated with a sequence of temporal bipartite graphs relating resources added by user u to tags used by u. We apply the link prediction approach proposed in [5] to these bipartite graphs for obtaining one or more lists of tags for annotating the resource rt and other similar resources. At this point, we make an assumption that the tags used by the similar users for resources that are somehow similar to the target resource can also be useful for recommendation. The approach of link prediction is described in next section. 3. At the end of step 2, we obtain k different ranked lists of tags, each obtained from data related to one of retrieved similar users. We apply a method to merge these lists using some list aggregation method [8]. To sum up, the Liptar approach is conceived as a framework offering three main hotspots to be adapted: a) the user similarity metric, b) the link prediction approach to be applied to infer tags to be recommended from the point of view of each retrieved similar user and c) the rank aggregation method to be applied to merge all obtained list of tags computed in step b). Next, we give a detailed description of the link prediction approach we have applied for prediction of tags for recommendation.

340

M. Pujari and R. Kanawati

Fig. 3. Liptar general cycle

3.1 Link Prediction Approaches We classify link prediction approaches according to the following criteria: − Dyadic vs. Structural approaches. A dyadic approach consists in computing a link score for each couple of unlinked vertices in graphs Gi (target vertices should be in the same connected component). The majority of link prediction approaches apply a dyadic approach. Different systems apply different score functions: topological scores [16], node features based scores [24, 23]. In contrast, structural approaches mine rules of evolution of (frequent) sub-graphs in the whole network. Frequent sub-graph mining approaches can be used for that purpose [15]. Such approaches can predict the appearance of multiple links at once. − Topological vs. nodes features based approaches. Score function of a link can be based on mere topological features such the degree, the common neighbors, the distance between involved nodes, etc. Or can be dependent of the node features itself. Topological approaches present the advantage of being applicationindependent. − Temporal vs. non temporal approaches. Most approaches ignore the temporal aspects of network formation. In other words, a link established at time t – i is considered having the same effect, for the prediction task, as a link established at time t. Recently, several works propose to take the temporal issue into consideration [7]. In our work, we apply a dyadic topological approach. This approach consists of reformulating the link prediction problem in terms of a two-class discrimination

A Supervised Machine Learning Link Prediction Approach

341

problem: linking and not-linking [9, 22, 21]. In machine learning community we talk usually about positive class (the linking class here), and the negative class (notlinking). Learning a classification model requires having a training data that describe examples of both classes. For sake of clarity, we explain hereafter how to generate training data for link prediction in the case of a simple unimodal graph. Then we show how to adapt this approach to bipartite graphs. The training data is constructed as follows. Let Glearn be a graph that summarize in some way the temporal sequence of networks G =< G1,...,Gt >. A simple way, often applied, to compute Glearn is to make the union of all snapshots in the sequence G. An example will be generated for each couple of nodes (x,y) that are not linked in Glearn but both belonging to the same connected component. The class labeling is obtained by checking whether the couple of nodes is indeed connected in Gt+1, in that case it will be a positive example in the supervised learning task - or not - in that case it will be a negative example. For each such couple of nodes, we compute a set of topological attributes that characterize their roles in the network as well as their similarity. Three types of topological metrics can be computed namely: Neighborhood based metrics, Combinations of node's topological metrics and Distance based metrics. Before giving a brief account of these measures, we introduce some basic notations that will be using later. We denote by ГG(x) the set of direct neighbors of a node x belonging to a graph G. The set of neighbors of a node x is denoted Г(x) when there is no ambiguity concerning the considered graph. |E| refers to the cardinality of set E. Graphs used in our study are non oriented or undirected and the degree of a node x in a graph G is equal to || ГG(x) ||. Commonly known single node's topological measures are the degree, the PageRank [6], the hub and the authority metrics computed by the HITS algorithm [14], the relative position of the node in the connected component: a central or a peripheral node, etc. The product of one of these measures for any two nodes can be used as a score for a pair of nodes. One such well known link score measure is preferential attachment [2] that takes the product of the degree of two nodes to find the score for a link between them. Preferential Attachment (denoted PA(x,y) ): ||. Г(x) × Г(y) ||. [10]. Under the category of neighborhood based metrics three most frequently used measures are the following: 1. Common Neighbours (denoted CN (x, y)): Γ (x) ∩ Γ (y) , (x)∩Γ (y) 2. Jaccard’s coeﬃcient (denoted JC(x, y)): Γ Γ (x)∪Γ (y) , 3. a measure proposed in [1] (denoted AA(x, y)): 1 Σz∈Γ (x)∩Γ (y) log(|Γ (z)| .

The simplest distance based metric is the shortest path between x and y denoted by SP(x,y). Another such attribute, also quite frequently used in affiliation analysis between nodes of a graph, is the measure proposed in [13]: it consists of computing a weighted sum of all paths between x and y. Link prediction in a bipartite graph : For predicting links in the bipartite graphs, we first take the union of the concerned bipartite graphs for the period of training, labelling and testing. For example, if G = < G1, ..... ,Gt > is a set of temporal graphs.

342

M. Pujari and R. Kanawati

As stated before, the learning graph is given by Glearn = i∈[1,tlearn ] Gi where tlearn is the time up to which we want to consider the learning period. Examples are generated for each pair of nodes in this graph that are labelled using a graph for the time considered for labelling. We also select a period for testing for which we also generate a union of the temporal graphs. The resulting graph is used to validate the learned training model. To generate the examples we consider only the node pairs that belong to the same connected component in the learning graph Glearn but do not have a link between them during the learning period. We select a set of topological metrics to be used as attributes characterizing the node pairs. These data is then used to train a model using supervised machine learning which in the end, is used to predict links in the test graph.

4 Experimentation The first experiment is done on data extracted from website CiteULike. We applied the Jaccard's index for computing both users and resource similarities. We use a modified version of link prediction approach proposed in [3] and Kemeny optimal method [8] for list aggregation. The inputs for the system are a user (target user) and a resource (target resource). The dataset covers a time period from year 2004 to year 2010. The total numbers of data entries are 10504915. After pre-processing we get a tripartite graph with 71464 users, 2402913 resources and 489682 tags. We use only meaningful tags, discarding the system generated ones. We found that there are 397252 resources without a tag which counts for 16.53% of total resources. We use data from period of 2005-2007 for training the link prediction model. We use a boosted decision tree classifier (the Adaboost using J48 classifier in the Weka platform. To make a comparison with a classical approach, we experimented with a basic method of collaborative filtering for the same target users. The input graph is the union of the temporal Resource-Tag graphs for top k similar users. We make a

Fig. 4. Preliminary results

A Supervised Machine Learning Link Prediction Approach

343

prediction of tags for resources used by the target user in validation period of 2009. This prediction is made on the basis of target user's history and the choice of similar users. Using this approach, for the same 24 users, the average precision is found to be approximately 0.282, average recall is 0.301 and average F-measure is 0.291. Figure 4 shows a comparison between the two approaches.

5 Conclusion In this paper we propose a new approach of recommendation of tags in folksonomy based on a link prediction approach in the bipartite graphs. The approach includes decomposition of a tripartite graph representing a folksonomy, into three bipartite graphs. The system consists of three areas of specialization: selecting users who are most similar to the target user, applying the link prediction approach on the graphs associating resources and tags used by these users to compute lists of tags for recommendation and then aggregation of these lists to have a final ranked list of tags to be proposed to the target user for recommendation. First experiments on real data prove the validity of the approach. This is an on-going project and in future we intend to experiment our approach on many other folksonomies. We also intend to use some other methods of similarity measures for users and resources. This approach can also be applied for recommendation of resources, although at present it has only been implemented for recommendation of tags.

References 1. Adamic, L.A., Buyukkokten, O., Adar, E.: A social network caught in the web. First Monday 8(6) (2003) 2. Barabasi, A.-L., Jeong, H., Neda, Z., Ravasz, E., Schubert, A., Vicsek, T.: Evolution of the social network of scientific collaboration. Physica A 311(3-4), 590–614 (2002) 3. Benchettara, N., Kanawati, R., Rouveirol, C.: Calcul de recommandation par prédiction de liens dans un graphe biparti. In: Actes de L’Ateiler Sur L’Apprentissage et Graphes Pour Les Systèmes Complexes (Plate-Forme AFIA 2009), Hammemt, Tunisie (May 2009) 4. Benchettara, N., Kanawati, R., Rouveirol, C.: Apprentissage supervisé pour la prédiction de nouveaux liens dans des réseaux sociaux bipartie. In: Actes da la 17iéme Rencontre de la Société Francophone de Classification (SFC’2010), St. Denis, La réunion, pp. 15–18 (June 2010) 5. Benchettara, N., Kanawati, R., Rouveirol, C.: Supervised machine learning applied to link prediction in bipartite social networks. International Conference on Advances in Social Network Analysis and Mining, ASONAM (2010) 6. Brin, S., Page, L.: The anatomy of a large scale hypertextual web search. In: Proceedings of Seventh International Conference on the World Wide Web (1998) 7. Cooke, R.J.: Link prediction and link detection in sequences of large social networks using temporal and local metrics. Master thesis, University of cape Town (November 2006) 8. Dwork, C., Kumar, R., Naor, M., Sivakumar, D.: Rank aggregation methods for the web. In: WWW, pp. 613–622 (2001)

344

M. Pujari and R. Kanawati

9. Hasan, M.A., Chaoji, V., Salem, S., Zaki, M.: Link prediction using supervised learning. In: SIAM Workshop on Link Analysis, Counterterrorism and Security with SIAM Data Mining Conference, Bethesda, MD (2006) 10. Huang, Z., Li, X., Chen, H.: Link prediction approach to collaborative filtering. In: Marlino, M., Sumner, T., F.M.S. III (eds.) JCDL, pp. 141–142. ACM, New York (2005) 11. Jäschke, R., Marinho, L.B., Hotho, A., Schmidt-Thieme, L., Stumme, G.: Tag recommendations in folksonomies. In: Hinneburg, A. (ed.) LWA, pp. 13–20. MartinLuther-University Halle-Wittenberg (2007) 12. Jäschke, R., Marinho, L.B., Hotho, A., Schmidt-Thieme, L., Stumme, G.: Tag recommendations in social bookmarking systems. AI Commun. 21(4), 231–247 (2008) 13. Katz, L.: A new status index derived from socimetric analysis. Psychmetrika 18(1), 39–43 (1953) 14. Kleinberg, J.: Authoritative sources in hyperlinked environments. Journal of the ACM 4, 604–632 (1999) 15. Lahiri, M., Berger-Wolf, T.Y.: Structure prediction in temporal networks using frequent subgraphs. In: CIDM, pp. 35–42. IEEE, Los Alamitos (2007) 16. Liben-Nowell, D.: An Algorithmic Approach to Social networks. PhD thesis, M.I.T., ( June2005) 17. Limpens, F., Gandon, F., Buffa, M.: Rapprocher les ontologies et les folksonomies pour la gestion des connaissances partagées: un état de l’art. In: Prize, Y. (ed.) Actes d’IC, pp. 123– 134. Institut National Polytechnique de Lorraine (2008) 18. Lipczak, M.: Tag recommendation for folksonomies oriented towards individual users. In: ECML PKDD Discovery Challenge 2009, CEUR Workshop Proceedings, vol. 497, pp. 189–199 (2009) 19. Lu, Y.-T., Yu, S.-I., Chang, T.-C., jen Hsu, J.Y.: A content-based method to enhance tag recommendation. In: Boutilier, C. (ed.) IJCAI, pp. 2064–2069 (2009) 20. Mrosek, J., Bussmann, S., Albers, H., Posdziech, K., Hengefeld, B., Opperman, N., Robert, S., Spira, G.: Content-and graph-based tag recommendation: Two variations. In: ECML PKDD Discovery Challenge 2009, CEUR Workshop Proceedings, vol. 497, pp. 189–199 (2009) 21. Murata, T., Moriyasu, S.: Link prediction based on structural properties of on-line social networks. New Generation Computing 26, 245–257 (2008) 22. Pavlov, M., Ichise, R.: Finding experts by link prediction in co-authorship networks. In: Zhdanova, A.V., Nixon, L.J.B., Mochol, M., Breslin, J.G. (eds.) FEWS, CEUR Workshop Proceedings, vol. 290, pp. 42–55 (2007), http://CEUR-WS.org 23. Popescul, A., Ungar, L.H.: Structural Logistic Regression for Link Analysis. In: Dzeroski, S., Raedt, L.D., Wrobel, S. (eds.) 2nd Workshop on Multi-Relational Data Mining, MRDM (August 2003) 24. Taskar, B., Wong, M.F., Abbeel, P., Koller, D.: Link prediction in relational data. In: Thrun, S., Saul, L.K., Schölkopf, B. (eds.) NIPS. MIT Press, Cambridge (2003) 25. Wartena, C., Brussee, R., Wibbels, M.: Using tag co-occurrence for recommendation. In: ISDA, pp. 273–278. IEEE Computer Society Press, Los Alamitos (2009) 26. Zhang, Z.-K., Zhou, T., Zhang, Y.-C.: Personalized recommendation via integrated diffusion on user-item-tag tripartite graphs. CoRR, abs/0904.1989 (2009)

Beyond the Social Search: Personalizing the Semantic Search in Social Networks Júlio Cesar dos Reis1,2, Rodrigo Bonacin2,3, and M. Cecília C. Baranauskas1 1 Departament of Information System at Institute of Computing University of Campinas (UNICAMP) - 13083-970, Campinas, SP, Brazil 2 Center for Information Technology Renato Archer (CTI) - Rodovia Dom Pedro I, km 143,6, 13069-901, Campinas, SP, Brazil 3 Faculty of Campo Limpo Paulista (FACCAMP) - 13231-230, Rua Guatemala, nº 167 Jardim América, Campo Limpo Paulista, SP, Brazil {julio.reis,rodrigo.bonacin}@cti.gov.br, [email protected]

Abstract. Nowadays search mechanisms are decisive to people successfully access relevant and valuable information. In the current Web environment, people connected to each other through Social Network Services (SNSs) bring new challenges for search approaches. Recent literature in the SNS field shows expressive evolution but the approaches for searching still utilize mechanisms that just consider the syntactical processing of the information. Based on sociocultural aspects of a SNS, we propose a novel approach that enables to provide personalized semantic search. This approach makes use of ontologies to represent the shared meanings, and consequently to present better organized, ranked and adapted search results for each SNS user. The paper presents the way the proposed mechanism was implemented and points out examples that illustrate the possibilities raised by the approach. Keywords: Semantic Search, Organizational Semiotics.

Web ontology, Search Personalization,

1 Introduction SNSs play a vital role in the Web environment nowadays. Since their introduction, Social Networks such as MySpace, Facebook, and Orkut have attracted millions of users, many of whom have integrated these sites into their daily practices [1]. In SNSs people interact making connections and constituting communities; they produce information, communicate, collaborate, and share their objectives, interests, activities and practices. SNSs emerged as a propitious virtual space to access and share information, representing an opportunity to share experiences and knowledge [2]. The access to the digital content generated in these applications depends on resources such as search mechanisms. In order to deal with the limitations of traditional search mechanisms, literature has proposed a set of approaches and techniques over the years, such as: semantic search, personalized and adaptive search, personalized information filtering, collaborative information seeking, as well social A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 345–354, 2011. © Springer-Verlag Berlin Heidelberg 2011

346

J.C. dos Reis, R. Bonacin, and M.C.C. Baranauskas

search. Some works have proposed to enrich the search mechanisms using users’ preferences and characteristics, while others proposed to create ontology-based search engines. However, usually such solutions are not integrated, loosing information about how to make a bridge between user’s preferences and characteristics to the semantic models used. Moreover, most of the search solutions for SNSs are directed to the search of users or experts grounded on friendship and reputation relationships. Certainly these solutions for social search are important; nevertheless they can be improved by providing a personalized semantic search for the content produced in the SNS. Besides friendship relationships, the shared contents and their meaning should be considered by the search mechanism. While SNSs key technological features are fairly consistent, the cultures that emerge around them are varied [1]. Actually, people in SNSs share much more than just friendship relationships; they also share culture, language and meaning. People behave according to commitments, linguistic constraints, culture and other social aspects. Therefore, the main challenge in this context is to have methods that can deal with these aspects, being able to model the shared meanings from the SNS, and to explore it to improve personalized search results. The current search solutions for SNS do not take into account the intrinsic semantic aspects that come from the network’s activity itself (i.e. they do not consider the meanings of words used by people in the SNS). Usually semantic search solutions are based on computational representation models disconnected of the reality (i.e. they do not include the social actors that interpret the concepts). Thus, a search mechanism capable of “interpreting” the shared meanings used by people in the SNS is necessary. For that, we need methods and approaches that allow to capture and to represent social and cultural aspects from the SNS such as their colloquial language. This paper brings a semantic search solution for SNS able to deliver personalized semantic search results not just focused on people, but also on the content created by them in the SNS. Based on a semantic model created from the SNS contents, the results are adapted according to the user that performs the search. Since search results personalization are very dependent on who performs the search, the proposed solution considers that the meaning of the words depends on who produces the information and also who searches for it. Consequently, the solution is based on a semantic model that links the concepts meanings from who produces and who searches the information. In order to reach it, the proposal uses a Semiotic-based approach [4] in the modeling process resulting on Web ontologies (WOs) that enable search results more adequate for users of that context. This represents a novel approach for ontology-based search in SNS, grounded on semantic modeling aspects and techniques that can reach more adequate and personalized search results. The solution is implemented and illustrated in the VilanaRede1 SNS. The paper is organized as follows: Section 2 reviews proposals for Web search strategies and situate them into SNS context. Section 3 describes the methods used in a Semiotic-based approach to personalize semantic search results in SNS; Section 4 presents the design and implementation issues of the search mechanism developed at VilanaRede, illustrating the search strategies with examples as well the software architecture used to explore the semantic aspects of the semantic model. Finally, we 1

www.vilanarede.org.br

Beyond the Social Search: Personalizing the Semantic Search in Social Networks

347

conclude the paper with a brief discussion, including a summary and directions for further work in Section 5.

2 Reviewing Search Approaches and Strategies Situated in SNS Literature has presented a number of approaches and solutions to improve search mechanisms (both in the Web and also in SNS) based on various strategies, such as: recommendation, trust, as well as semantic and personalization aspects. In this section we briefly explain and review each of these. 2.1 Social Search Usually the studies of SNS search focus on search for SNS users. The work conducted by Yu & Singh [5] proposes a solution to find among participants of a SNS, specialists in some domains. Their proposal makes use of multi-agent system techniques. Based on users’ information regarding their social network, artificial agents find people based on reputation. Vieira et al. [6] argue that the friendship relationship is an important evidence to find people in a SNS, and the search results could be modeled in function of a friendship graph. Algorithms in graphs try to find the shortest path to friends, aiming at ordering the users’ profile during the search. More recently, other approaches have proposed to use aspects of recommendation as well as trust to find items in SNS. For instance, Gürsel & Sen [7] propose a search system to find items based on social network concept. This system uses a framework based on artificial agents that mine the group of relationships of a user on the network. The objective is to facilitate the search for items of interest through the use of social connections. As a consequence, it is possible to rank the search results highlighting the items of preference recently posted by friends of the SNS, which will be of particular interest to a user. Other approaches stress the use of cognitive authority from the SNS for information retrieval in folksonomy-based systems [8], in order to reach information quality. 2.2 Semantic Search The goal of optimizing search results has motivated research in the semantic search field. The need to retrieve information semantically enriched gave rise to an increasing interest in research on ontologies. According to Wei et al. [8], semantic search extends the scope of traditional information retrieval (IR) paradigms from mere document or people retrieval to entity and knowledge retrieval, improving the conventional IR methods by looking at a different perspective, i.e. the meaning of words. In order to provide search mechanisms with semantic features, it is necessary to interpret a knowledge model about the domain. The knowledge can be represented by means of ontologies. A wide adopted definition for ontology in Computer Science is done by Gruber [10]. The ontologies are usually described by computational ontology languages such as Web Ontology Language (OWL) [14]. Works such as Heflin & Hendler [11], Guha et al. [12] among others have introduced the idea of using

348

J.C. dos Reis, R. Bonacin, and M.C.C. Baranauskas

ontologies by semantic search mechanisms. Considering the SNS context, Choudhari et al. [13] describe an architecture to develop semantic search for SNS. However, instead of using ontologies they used an approach based on clustering of terms semantically related to perform the semantic search. 2.3 Personalized and Adaptive Search There are also a number of studies for personalizing Web search that develop strategies of adaption of the search results according to user’s characteristics. Many works claim that the users’ interests, preferences and need are an essential element in identifying the user’s context for the search. Various approaches employ user modeling components based on the users’ interactions, such as: previous queries or Web navigation activities history (e.g. visited pages). In this context, some works analyze implicit information about the user’s interests [e.g. 22, 21, 22, 24]. However, each one tackles a single facet of personalization, either biasing search results to general user-interests [22] or respecting the current search session’s context [21], or using both strategies [e.g., 24]. Some personalization techniques not only consider a single user, but also take into account the actions of a surrounding group of users [e.g., 22] through a collaborative filtering approach. Another alternative to address personalization is the use of ontologies. The proposals include the (semi) automatic generation of semantically enriched ontology-based user profiles. Ontology-based personalized search approaches [e.g. 25] involve building models of user context as ontological profiles, by assigning implicitly derived interest scores to existing concepts in the domain ontology. Moreover, recent literature has presented adaptive ontologies process for domain evolution [e.g., 25] in adaptive ontology-based Web search.

3 A Semiotic-Based Approach for Search Personalization in SNS Semiotics [15], the ancient doctrine of signs, leads us to a precise understanding of information as various properties of signs. Signs are simple entities easy to deal within the inter-subjective domain. Anything standing for another thing or used to signify something is an example of sign: words, traffic lights, diagrams, a wave of hand or a facial expression. From a Semiotic point of view, people communicate in SNS by sharing signs, and such signs are socially constructed and shared over interaction. In this sense, we have to understand the signs and how they influence the communication process. By studying the use of signs in the organized behaviour, Organisational Semiotics (OS) [3] theory and methods support us in understanding and modeling social aspects in SNS. More adequate search mechanisms for SNS should take into account the user’s language aspects and their applied meanings aiming to deliver search results that make sense to them [16]. For that, meanings must be represented as they are used in the SNS context; thus, computational models (e.g. ontologies) capable of representing the semantic of the language that come out from the SNS are necessary. The proposed method, developed to construct more representative WOs that come out from SNS content, is based on the Semantic Analysis Method (SAM) [8]. SAM delineates the

Beyond the Social Search: Personalizing the Semantic Search in Social Networks

349

area of concern of a social context and identifies the basic patterns of behaviour (affordances) of their agents [3]. It produces a stable Ontology of the context. This Ontology describes the semantic aspects of the signs shared in a SNS. A WO that represents the semantics of the SNS application should contain the agents that interpret the socially shared signs. This approach incorporates and takes to WOs concerns, representations arising from a Semiotic perspective [4]. In this process theoretical and methodological concepts of SAM are used in conjunction with technologies from the Semantic Web (SW) initiative to describe computationally tractable ontologies using OWL. The proposed method is based on a subjectivist paradigm [3], which understands reality as a social construction based on the behaviour of agents participating on it, instead of an objectivist paradigm which presupposes that there exists a world independent of the observer. SAM uses an agent-in-action ontology, determining the underlying semantics of a social context and the relationship between the human agents and their patterns of behaviour [3]. SAM supports the analysis, specification and representation of a social system assisting users or problem-owners in eliciting and representing their meanings in a formal and precise semantic model - the Ontology Chart (OC). OC is a graphical representation of a conceptual model that describes a view of responsible agents in the focal domain, including their pattern of behavior named affordances and the ontological dependences between them [3]. The most important concepts in SAM are Affordance and Agent. Affordance is a concept introduced by Gibson [17] that can be used to express the invariant repertories of behaviour of an organism made available by some combined structure of the organism and its environment. In SAM the concept introduced by Gibson was extended by Stamper [18] to include invariants of behavior in the social world; affordances are social constructs in a certain social context. The Agent is a special kind of affordance, which can be defined as something that performs responsible behaviour. An agent can be an individual person, a cultural group, a language community, a society, etc. (an employee, a department, an organization, etc.). SAM provides a different and independent view of the social context under study addressing issues that are not represented in any traditional WO. Moreover, SAM allows us to discover the agents and their respective pattern of behaviour (affordances) within a SNS. In the process to generate OWL ontologies from the SNS content, the OCs inform the SNS semantic search mechanism. Figure 1 illustrates the method to create OWL ontology from SNS content through SAM. The transformation from OC to OWL is carried out by an assisted process [19]. A case study applying the entire process based on real contents of a SNS is described in [20].

Fig. 1. Informing the Search Mechanism with WO built from SNS content through SAM [adapted from 16]

350

J.C. dos Reis, R. Bonacin, and M.C.C. Baranauskas

4 Providing Personalized Semantic Search Results in SNS The agents modeled during the proposed method (that also composes the OWL ontology) are adequately associated to each SNS user. The key point for the proposed personalization strategy is to use information about the social agents as a context for the search. Thus, the possible meanings of a word are connected to agents in both situations, during the search time and during the processing of the search index of the contents generated by the SNS users. Thus, the disambiguation of possible polysemic words is led by the agents that interpret the terms. For example, considering a query as “Java”, some users may be interested in documents dealing with the programming language, while others may want documents related to “coffee” or touristic information (Indonesian island). Therefore, since a SNS user may be connected to an agent “programmer”, the meaning applied will be “programming language”, as well whether the user is connected to an agent “tourist”, the meaning applied would be other. The relationship between the users with the possible social agents can be reached following various strategies. First, the users can point out some agent that most fits their profile; in addition techniques can be applied trying to find the possible agents for a user based on their activity on the SNS (e.g. the contents posted). Moreover, the social network can also be used to point out such agents. Based on the agents, semantic rules described by Semantic Web Rule Language2 (SWRL) are created to infer the correct meanings. Such rules make a matching between a possible term and an agent inferring as meaning a class of the OWL ontology. E.g. if the agent is a ‘programmer’ and the term is ‘java’ then the meaning (class) is ‘programming language’. These rules are used in both cases to classify search keywords during the search time, as well during the processing of the search index formed based on the SNS contents (e.g. announcements). Each of these is processed using the rules that classify each term of the content forming an inverted index search. In this index a term would have different classifications (classes) that point out to various contents in the SNS, since the same term can have different meanings treated by the rules based on the agents. For example, the combination “Java” (programming language) may point out to some content in the index, and another combination “Java” (tourism) may point out to other content. This classification is used to organize the search results and to personalize it. Rather than providing the search results as a single ranked list, we propose a structured organization for the search results based on boxes (see Figure 2). In the User Interface (UI) for the search results, the semantic results are organized into two main boxes, embedding other boxes. The first box has the title ‘Personalized’ and the second has the title ‘Other’; both have the same intern organization as presented in Figure 2. The internal boxes organize the semantic results according to the expansion of the queries in the ontology-based search. The words found during it are queried in the index to find semantic search results. The most relevant are placed right above the internal boxes. These results strictly match the keywords considering the semantic and syntactic criteria. The internal boxes may contain results with ‘separated keyword’, once the search engine could not find results containing exactly the combination of all keywords; there is also a box that keeps the synonymous (the synonymous terms are also modeled by the ontology); and a box with ‘more specific 2

http://www.w3.org/Submission/SWRL/

Beyond the Social Search: Personalizing the Semantic Search in Social Networks

351

results’ and ‘more generic results’ that contain results of generic-specific relation (is-a) from the OWL ontology; and also a box of ‘related results’ got from relationship between concepts in the ontology, for instance, ontological dependence relationships.

Fig. 2. Organization for the Personalized Semantic Search Results

For the “Java” example, supposing a “programmer agent” performing the search, all the results concerning “Java” (as programming language) as well as the synonymous and generic-specific results in the same context must be placed into the ‘personalized’ box and into its internal boxes. All other possible search results, with other meanings or without any recognized meaning are placed into the ‘Other’ box. Since both boxes and internal boxes results are hidden by default, the users can explore the results in a small initial list, and expand it as they like. Additionally to this approach to search by the meanings, other search results items may also be ranked into the ‘Personalized’ box using an agents approach. Since the agents are connected to all SNS users, in this approach, the items that were posted by users that contain some agent equal to the agent of the user that performs the search are also considered as personalized results. The proposed search mechanism was implemented at VilanaRede SNS. This SNS was developed as a product of e-Cidadania's3 Project, with the objective of being 3

www.nied.unicamp.br/ecidadania

352

J.C. dos Reis, R. Bonacin, and M.C.C. Baranauskas

accessible for all, including those users less familiar with technology or with low literacy levels. The OWL ontology that informs this search mechanism was built based on the VilanaRede contents. The VilanaRede system uses the Content Management System Drupal 54 to manage users’ accounts, content types and other general functionalities; the proposed semantic search service is hosted in a JBoss Application Server5 as an external and independent Web Service. The implemented Web Service manages all the necessary information and provides all services to handle the search request, such as: the index processing, the adding of agents and user-agent relationship as well as the ontology management. The search service includes the search algorithm that provides the search results by lists. The organization of these lists is performed at the Drupal side. Figure 3 presents an example of the VilanaRede UI after the search request processing.

Fig. 3. VilanaRede User’s Interface showing Personalized Semantic Search Results

5 Final Considerations Most of the information available in SNSs are in a format that can not be directly interpreted by machines. Also, most of the search mechanisms have not been designed to interpret or to represent the meanings of the terms. They have been fundamentally constructed based on the approach “one fits all” and on keywords comparison with lexical-syntactic processing, i.e. do not considering semantic 4

Drupal. http://drupal.org

Beyond the Social Search: Personalizing the Semantic Search in Social Networks

353

knowledge. Consequently, the mechanisms do not have the capacity to differentiate the relevant search results from the semantics point of view. In this context, ontologies may be a fundamental cornerstone for designing useful semantic search services. The personalization approaches for search described in literature are based on topics of preferences and on users’ needs. These topics are discovered and modeled mainly by browse history and past user’s search queries analysis. Even the proposals that consider semantic knowledge to create the user profile use a number of calculations and scores that do not model the users’ language. In personalization approaches using the users’ social network, the search results are mainly ranked with items viewed or highlighted by people that compose their personal network. We have observed that it is necessary a novel way to personalize and to organize the search results by taking into account the SNS semantic aspects. In our approach the users’ social characteristics are considered in the adaptation of the semantic search. Results are given according to responsible social roles, i.e. the agents identified during the semantic modeling process by SAM (e.g. social identity of the users as their profession). The agents determine the meanings of the words used in the index and during the search. Therefore, our approach provides means to discover as well to distinguish the meanings used by people at the SNS through the agents represented into OCs. Such agents are used by semantic rules described in SWRL in the ontology to get adequate meaning for the words. Based on that, we presented a semantic search mechanism implemented at VilanaRede SNS. This mechanism is able to personalize and better organize the semantic search results according to the user that performs the search, presenting first the most relevant and meaningful search results. The mechanism considers both lexical and conceptual aspects (e.g. generalization, specialization and related results). Further works involve conducting a case study with real users using the mechanism at VilanaRede in order to discuss experimental results.

References 1. Boyd, D.M.: Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication, 210–230 (2008) 2. Jamali, M., Abolhassani, H.: Different aspects of social network analysis. In: Proc. of 2006 IEEE/WIC/ACM International Conference on Web Intelligence, pp. 66–72. IEEE Computer Society Press, Washington, DC (2006) 3. Liu, K.: Semiotics in information systems engineering. Cambridge University Press, Cambridge (2000) 4. Reis, J.C., Bonacin, R., Baranauskas, M.C.C.: A Semiotic-based Approach to the design of Web Ontologies. In: ICISO 2010, Reading, UK, pp. 60–67 (2010) 5. Yu, B., Singh, M.P.: Searching social networks. In: Proc. of the 2th International Joint Conference on Autonomous Agents and Multiagent Systems, pp. 65–72. ACM Press, New York (2003) 6. Vieira, M.V., Fonseca, B.M., Damazio, R., Golgher, P.B., Reis, D.C., Neto, B.R.: Efficient Search Ranking in Social Networks. In: Proc. of the 16th ACM Conference on Information and Knowledge Management, Lisbon, Portugal, pp. 563–572 (2007)

354

J.C. dos Reis, R. Bonacin, and M.C.C. Baranauskas

7. Gürsel, A., Sen, S.: Improving search in social networks by agent based mining. In: Proc. of the 21st International Jont Conference on Artifical Intelligence Table of Contents, Pasadena, California, USA, pp. 2034–2039 (2009) 8. Pereira, R., Silva, S.R.P.: The Use of Cognitive Authority for Information Retrieval in Folksonomy-Based Systems. In: ICWE 2008, Yorktown Heights, NJ, pp. 325–331 (2008) 9. Wei, W., Barnaghi, P.M., Bargiela, A.: Search with Meanings: An Overview of Semantic Search Systems. Int. Journal of Communications of SIWN 3, 76–82 (2008) 10. Gruber, T.R.: A translation approach to portable ontologies. Knowledge Acquisition 5(2), 199–220 (1993) 11. Heflin, J., Hendler, J.: Searching the web with SHOE. Artificial Intelligence for Web Search. In: AAAI Workshop, WS-00-01, pp. 35–40. AAAI Press, Menlo Park (2000) 12. Guha, R., McCool, R., Miller, E.: Semantic Search. In: Proceedings of the 12th International Conference on World Wide Web, Budapest, Hungary, pp. 700–709 (2003) 13. Choudhari, A., Jain, M., Sinharoy, A., Zhang, M.: SmartSeek: A Semantic Search in Social Networks (2008), http://www.cc.gatech.edu/projects/disl/courses/8803/2008/ project/project_deliverables/group22/final_report.pdf (accessed, January 2011) 14. Web Ontology Language (OWL). W3C Recommendation (February 4, 2004), http://www.w3.org/TR/owl-features (accessed, January 2011) 15. Peirce, C.S.: Collected Papers. Harvard University Press, Cambridge, Mass (1931-1958) 16. dos Reis, J.C., Bonacin, R., Baranauskas, M.C.C.: Prospecting an Inclusive Search Mechanism for Social Network Services. In: Filipe, J., Cordeiro, J. (eds.) ICEIS 2010. Lecture Notes in Business Information Processing, vol. 73, pp. 555–570. Springer, Heidelberg (2011) 17. Gibson, J.J.: The Theory of Affordances. In: Shaw, R., Bransford, J. (eds.) Perceiving, Acting, and Knowing (1977) 18. Stamper, R.K.: Social Norms in requirements analysis – an outline of MEASUR. In: Jirotka, M., Goguen, J., Bickerton, M. (eds.) Requirements Engineering, Technical and Social Aspects. Academic Press, New York (1993) 19. Reis, J.C., Bonacin, R., Baranauskas, M.C.C.: From Ontology Charts to Web Ontologies: Heuristics and Transformation Rules (IC-11-02), IC/UNICAMP (2011) 20. Reis, J.C., Baranauskas, M.C.C., Bonacin, R.: Modeling Meanings from Inclusive Social Network Services (IC-11-03), IC/UNICAMP (2011) 21. Shen, X., Tan, B., Zhai, C.: Implicit user modeling for personalized search. In: CIKM (2005) 22. Teevan, J., Dumais, S.T., Horvitz, E.: Personalizing search via automated analysis of interests and activities. In: ACM SIGIR, Salvador, Brazil, pp. 449–456 (2005) 23. Sugiyama, K., Hatano, K., Yoshikawa, M.: Adaptive web search based on user profile constructed without any effort from users. In: WWW, New York, USA, pp. 675–684 (2004) 24. Elbassuoni, S., Luxenburger, J., Weikum, G.: Adaptive Personalization of Web Search. In: 1st WS on Web Information Seeking and Interaction, Amsterdam, The Netherlands (2007) 25. Sieg, A., Mobasher, B., Burke, R.: Web Search Personalization with Ontological User Profile. In: CIKM, Lisboa, Portugal, pp. 525–534 (2007) 26. Guelfi, N., Pruski, C., Reynaud, C.: Experimental Assessment of the TARGET Adaptive Ontology-based Web Search Framework. In: 10th International Conference on New Technologies of Distributed Systems, Tozeur, Tunisia, pp. 297–302 (2010)

Factors Influencing Online Social Interactions Ravi Shankar IDC, IIT Bombay, Mumbai 400076, India [email protected]

Abstract. Popular discourse provides glimpses of the relationship between social motives and user experience. In this paper, factors influencing social interactions with seven popular social networking sites are studied in the Indian context. Relatively a novel method of focused unstructured interview technique with adaptation from triadic sorting is used to unearth the motivation and behavior of the user. It indicates subtle connections and motivations and started revealing some of the interesting dimensions - self orientation, social presence, mode of relationship and participation, temporal and socio-cultural context. This study successfully gathers information and proposes an approach to organize the user motivation into a dimensional scale for online social Interactions. Keywords: Social user experience, Social interaction, Social networks, Triadic sorting.

1 Introduction Although the internet and web applications have grown considerably, people are still looking for more effective methods of communicating, collaborating and interacting with each other better. As the architecture of participation is merging into the rapid and agile collaboration, information sharing, new ways of harnessing collective intelligence is emerging. All future consumer sites and business applications are looking at adding a layer of social interactions above the regular functionality. Users used to social communication and collaboration concepts are expecting changes in the way they interact with each other over business transactions, as businesses are innovating social concepts. Our aim is to understand the average Indian user’s motivations and behaviour in the popular social networking sites and corporate social application, so we can propose an approach towards a dimensional scale to gather requirements of various social applications. This study is part of a larger research project and is limited to understanding of the factors influencing the user’s current social behavior, perception and motivation in the social applications that are popular in this geography. In terms of the scope the study covered largely the urban, educated population in India. The fundamental questions and scope was around • What are users interested; whom are they connecting to, why are they connecting • What are users doing, behaving, responding in different context A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 355–362, 2011. © Springer-Verlag Berlin Heidelberg 2011

356

R. Shankar

• Why are they doing so; what is the motivation behind different behavior • What is the social incentive behind sustained presence and users interactions

2 Background Related research has investigated on socio-cultural issues of social interactions; social media and communication; social networking and commerce; online communities and usability metrics; motives for social behaviour. No direct research on dimensions for Social Interactions or metrics for social user experience is available. A review of the literature shows a number of social interaction factors (Table 1). Table 1. Factors affecting social interaction and reference to previous research Factors affecting social interactions Social activities Social presence Sense of belonging Object centeredness Mode of relationship Facilitation Lack of immediacy / connectivity Sense of time / space and flexibility Lack of verbal and visual interaction Private-public conversation Sense of awareness, trust and interruption

Reference [1], [2] [5], [11] [6], [1] [6], [7] [7], [12] [7], [ 2] [3], [7] [9] [12], [2] [9] [10], [12]

Several studies focus on Blogs and wikis, which represent new repositories of information generated by employees in the corporate world. Though these tools help connect to colleagues, the value of these tools for the average employee is more information centric than social or business centric. Studies on various social networking sites are generally focused on four main aspects: privacy issues, self-presentation, network analysis, and social capital benefits. It is to be noted that, these studies validate the need for new tools and enhanced metrics for measuring usability and sociability in the enterprise applications. Our aim is to understand the average Indian user’s motivations and behaviour in the popular social networking sites and corporate social application, so we can propose an approach towards a dimensional scale to gather requirements of various social applications. The object of investigation in this study is to understand the dimensions of social interactions of an average urban Indian user.

3 Methodology The study was carried out on 26 users across six most popular social networking sites in India, Face book, Twitter, LinkedIn, Orkut, YouTube, Flickr, in addition to a corporate intranet or a social application that they use in their environment. The participants were selected through convenience sampling method, drawn from the professional contacts of the author. One of the requirements was that all subjects have sustained social presence with more than four of the selected popular sites. But specifically participants with higher presence were preferred over inactive users.

Factors Influencing Online Social Interactions

357

Before scheduling the interview their interest in social networking and interactions were gauged to select the participants. Our initial unstructured discussions with early four participants, was focused in creating the overall methodology for the interview, with questions such as… “Why is this important to you?” “What does it mean to you?” “Does it facilitate your response?” “Does it support sharing?” “Does it encourage you joining…?” Following this, our template for focussed unstructured interview (Fig. 3) with a short pre-structured questionnaire to cover demographic details, social presence, perceived frequency of common interactions, along with motive for starting, frequency, activities, Face to Face vs. offline in the context of the selected popular social networking sites was introduced. The unstructured interview and analysis of the recent one month activity at a higher level showed the frequency of visit across the selected sites and frequency of common interactions were studied.

Fig. 1. Frequency of visit across the selected sites

Fig. 2. Frequency of common interactions

358

R. Shankar

Fig. 3. Focused unstructured interview template

A new method of unstructured interviews combined with questioning based on triadic sorting as discussed by Kelly [13] was used to unearth the dimensions of interest and promote involvement and engagement. We start with broad area of interest and behaviour and narrow down appropriately and presented three different concepts or ideas to participants (based on their initial responses on their perception and frequency of use) and ask them to identify how two of them are related and differ from the third. Participants tried to describe the attributes in their own terms, which gave insight about their perceptions, how they sort different parameters. “How do you differentiate relationship in the context of Friend - Family – Coworker?” “Can you compare Searching for people - Searching for media - Searching for someone’s comment / response?” Using the frequency data and as the discussions unfold the triad questions were formed. Respondents were presented with a meaningful context within which to make the distinctions. Sometimes participants couldn’t explain why a concept was different. We found it was helpful to keep track and revisit those issues later once they’d gotten used to the method. With initial slowness Participants started revealing different aspects that were interesting to compare the concepts. Starting with casual discussion and jumping to questions based on triadic sorting without specifically announcing the approach helped the user settle down. Participants were drilled iteratively following this

Factors Influencing Online Social Interactions

359

process to identifying additional attributes. Typically respondents came with about 2 to 3 different distinctions for a given category. After the data collection the author used most relevant or common aspects as inspiration for further classification and trends. Table 2. Sample triadic sorting analysis - Relationship with friend - family - co-worker Triadic Sorting Friend vs. others

Attribute Openness

Co-worker vs. others

Sense of place

Co-worker vs. others

Time orientation

Family vs. others

Identity

Family vs. others

Sense of Trust

Comment “Share freely any video, message…It’s restricted with family and co-worker…” “Primarily Interact in Orkut, co-workers more in social site…” “With friends and family I stay connected forever…” “Mostly true identity…vs. messages for attention…” “Family and Co-workers are trustworthy vs. new friend request…”

4 Analysis and Discussion Data collected from this study, were coded, classified and analyzed in the context of interesting attributes. Social relationship include people who use it, their attitudes, roles, and relationships; the Objects include the physical structure, norms and routines of the place the object is used; their interests and the social and cultural contexts of the people who create and consume the object. Though the study did not aim at concluding results, the data trends and classification gave an approach for social interactions dimensions that can be further explored and validated. 4.1 Factor 1: Self-Orientation vs. Others-Oriented The data trends point out that users vary between self-orientation and others oriented in different contexts. Most users in our study cared about their own interests compared to others social ratings. This not only includes their self-reflection and understanding of their own activity, but also anticipate how others perceive them. For them their self image and identify was vital to make an impression or attract attention. They were more conscious about projecting their personality style and credibility. They tend to spend more time on their profiles and feels pride. Data trends also show some users are most interested in others and their impressions of others. They consider how they appear to others, and what others seem to think of them. This interest includes others in general, specific individuals or the community.

360

R. Shankar Table 3. Analysis of various comments across self-orientation vs. others-oriented scale

Interest

Behaviour

Intent

Self-Orientation

“I think my profile closely reflects my personality and interests…” “Enjoy collecting ratings, request feedback, post comments” - with reciprocate… “It’s a sense of achievement, when my slides were featured… compare against peers…” “Friends like the way I state my status messages and update my profile pictures” “I frequently respond to message updates and links from my friends” “I notice status change of my friends and update my status frequently…“ “I am primarily here to see what’s going on in their life…“

Interested in self-image

Others oriented

Self-Motivated

Seek recognition

Sensitive to pride

For acknowledgment Impressions on others Users are interested in others

4.2 Factor 2: Relationship vs. Object Centered Some users show interest in relationship while others show interest in the object. Typical first reaction to a Social context depends on the relationship or object centeredness. The data indicate not only about their understanding of social relationship but also about trust, belongingness and relationship type and degree of separation. Some users are interested in who created instead of what is posted. For them source of information is important that the information itself. This orientation tends to influence a pattern of interactions and responses. Table 4. Analysis of various comments across Relationship vs. object centered scale

Interest

Behaviour

Intent

Relationship centered

“I comment for my close friend’s contribution, irrespective of…” “Nice to follow my friends and see who they are connected…” “I decided to read a book as my friend was excited…” “I have only sent links to my post that I want my friend to see…” “I receive invites to so many sites …which I haven’t seen at all…” “I generally search and land up instead of browsing from…” “I think others do not pay attention to common messages, so what’s the fun; I use it as a collection of good links” “Primarily interested in the video…”

Interested in who posted…

Object centered

Sense of belongingness Social discovery Interests is relational Attention to object Searching object Focus on the goal

Interested in what is posted

Factors Influencing Online Social Interactions

361

4.3 Factor 3: Spatial Presence vs. Low Presence Some users are co-present or meet face to face and socially interact online frequently, while there are contacts who haven’t met. Visibility of being available online compared to non-visibility influence social Interaction. Level of participation and relation type along with awareness about who others are, what they do, and their intention; play a larger role in this aspect. 4.4 Factor 4: Compete vs. Collaborative Engagement Is the engagement mode a collaboration or competition? Some users in our study tended to care more about participating and playing an active role compare to others. The data trends gave insights about mode of engagement, asynchronous communication, aspect of participation in Private or private space. It also indicated preferred Interactions of competing vs. collaborative modes. Sharing, publishing, Tagging, feedback, comments are preferred compared to collection, favorites, subscribe, review, rating, promote for a collaborative mode. 4.5 Factor 5: Short Term vs. Long Term Orientation Trends of long term and short term orientation are not only noticed in the context of Friends and relationship but also in the object’s interest itself. Users have a long term orientation to self-created objects in comparison to collected and published. Periodic interest is also expected even to close relationship that are perceived as long-term acquaintance. 4.6 Factor 6: Openness vs. Closedness Does the Interaction primarily orientation towards context of environment and social settings? Interaction behaviour differed across context of environment, including sense of place, sense of time, sense of awareness and trust, private-public conversation and other social settings. Aspects of interruption vs. motivation enhance the interaction behaviour.

5 Conclusion The study confirmed some of the earlier findings of user activity on social network as personal and social routine and pastime. Users use social media to relate to themselves and others, to show and present themselves. When commenting on any Object, they perceive it as their comment being a reflection of their personality. Reviews become a means to attract or draw attention to who they are. Testimonials written say as much about their author and his or her relationship to the person for whom the testimonial is written. Users are more attentive to what others are doing than to what you would like them do. Users are interested in relationships, interaction, and communication for acknowledgment and acceptance; users seek recognition and validation, often on the

362

R. Shankar

basis of their like-ability and desirability. User motivation is necessary for user engagement, sustained attention and active participation. The study successfully gathered information and organized the behaviour and motivation. The analysis and discussion indicates potentially beneficial insights towards studying dimensions of social interactions of a typical Indian user. Secondly the study also validates the factors studied and helped arrive at an approach for creating a multi-dimensional scale for Social Interactions in general. Though our study was not conclusive yet, our findings provide several insights for further research.

References 1. Boyd, D.M., Ellison, N.B.: Social network sites: Definition, history, and scholarship. Journal of CMC 13, 1 (2007) 2. Maloney-Krichmar, D., Preece, J.: A multilevel analysis of sociability, usability, and community dynamics in an online health community. ACM Transactions on ComputerHuman Interaction (TOCHI) 12(2), 201–232 (2005) 3. DiMicco, J.M., Millen, D.R., Geyer, W., Dugan, C., Brownholtz, B., Muller, M.: Motivations for Social Networking at Work (2008) 4. Boudourides, M.A.: Social and Psychological Effects in Computer-Mediated Communication (1995) http://www.math.upatras.gr/~mboudour/articles/csi.html 5. Gunawardena, C.N.: Social presence theory and implications for interactions and collaborative learning in computer conferences. International Journal of Educational Telecommunications 1(2/3), 147–166 (1995) 6. Kollock, P., Smith, M.: Communities in Cyberspace (1999) http://www.sscnet.ucla.edu/soc/faculty/kollock/papers/communities_00.htm 7. Kreijns, K., Kirschner, P.A., Jochems, W.: Identifying the pitfalls for social interaction in computer supported collaborative learning environments: a review of the research. Computers in Human Behaviour 19(3), 335–353 (2003) 8. Lee, A., Danis, C., Millar, T., Jung, Y.: Fostering Social Interaction in Online Spaces. Human-Computer Interaction (2001) 9. Wellman, B.: The network community: An introduction to Networks in the Global Village. In: Wellman, B. (ed.) Networks in the Global Village, pp. 1–48. Westview Press, Boulder (1999) 10. McDonald, D.W., Beschastnikh, I., Kriplean, T., Borning, A., Zachry, M.: System Design for Social Translucence in Socially Mediating Technologies. In: Socially Mediating Technologies Workshop, CHI 2009 (2009) 11. An Approach to Designing Systems that Support Social Processes. ACM Transactions on Computer-Human Interaction 7(1), 59–83 (March 2000) 12. Ackerman, M.S., Starr, B.: Social Activity Indicators: Interface Components for CSCW Systems (UIST 1995) 13. Kelly, G.A.: The Psychology of Personal Constructs. Norton, New York (1955)

Effects of a Synchronized Scoring Interface on Musical Quality Yuji Takai, Masao Ohira, and Ken-ichi Matsumoto Graduate School of Information Science, Nara Institute of Science and Technology 8916-5, Takayama, Ikoma, Nara, Japan +81-743-72-5312 {yuji-t,masao,matumoto}@is.naist.jp

Abstract. Collaborative music composition among casual users has the potential of creating advanced music that cannot be composed by a single user since the users can complement a shortage of musical knowledge each other. Although some studies have proposed music composition interfaces which synchronously visualize the composition data in real time, their effects on the quality of music are not still clear. As a result of our experiment, we found that the EMD value (0 is the best score) in using the interface Marble we proposed was lower than that of asynchronous music composition systems and Marble has increased the total number of utterances among users. Keywords: Synchronized Scoring Interface, EMD, music quality, ComputerSupported Cooperative Work.

1 Introduction In recent years, computer music software such as Apple’s GarageBand has been widely used from novice users to professional musicians, to efficiently compose music and/or effectively perform music since it can complement with the lack of real musical instruments with computational ones. One of the advantages to compose with computer music software is that it can give an ability of collaborative music composition to casual users who have knowledge of a limited number of musical instruments. Such the users can compose high-level music by combine each other’s musical knowledge. In the past study, we have developed a system called Marble which supports collaborative music composition by the casual users [1]. Marble has a synchronized scoring interface to help users collaboratively compose music online. We have confirmed that the system could make collaborative music composition among users easier and faster than the traditional environment1 where users have to share score data (e.g., MIDI), communicate with each other about musical ideas, request each other’s modified/edited score data over again and again, via e-mail or IRC (Internet Relay Chat). 1

We use a composition system with an asynchronous interface.

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 363–372, 2011. © Springer-Verlag Berlin Heidelberg 2011

364

Y. Takai, M. Ohira, and K.-i. Matsumoto

The easy and fast way of music composition is important in collaborative music composition among casual users. However, we believe the more important thing is the quality of composed music: “how are users satisfied with the quality of music composed through the collaborative music composition with Marble (synchronized scoring interface)?” This paper forces on the above question and discusses the effects of synchronized scoring interface on music quality and collaborative composition process. We quantitatively evaluate the quality of music composed through the collaborative music composition with Marble. We also conduct qualitative evaluation on user’s satisfaction with composed music.

2 Online Collaborative Music Composition 2.1 Types Table1 shows that the online collaborative music composition can be classified by members and their DTM composition skills. In Table 1, each type of the online collaborative music composition is briefly describes. Due to the space limitation, we cannot discuss each type in detail, but our target of this study is on the online collaboration among acquaintance, casual users. Table 1. Classification of Online Collaborative Music Composition Member

DTM Composition Skill

Casual Acquaintance

Advance

Casual Stranger Advance

Collaborative Method and Feature of Composed Music A melody composer mainly composes based on melody by him/herself as the people in charge of each musical instrument part append the musical note in the accompaniment part. Variety of music is composed to the course of member; however, maturity is not always high. In spite of take professional advice from friends and modify, individual user basically composes all the musical instrument parts. High-level music is composed including the genre (e.g., Orchestras) is needed professional skill. There are various forms; Decided a theme and a role (composition or writing the lyrics of a song), arranged to composition individually composed by high-level users. Multimedia composition is derivatively composed by high-level composer through mutual citation. There are high quality compositions with CG.

2.2 Process In general, collaborative music composition among casual users proceeds as follows:

Effects of a Synchronized Scoring Interface on Musical Quality

365

Step 1. A melody composer composes a melody part of music based on her knowledge of a particular music instrument (e.g., piano). She cannot compose accompaniment parts by herself because of her limited knowledge on music instruments even if she has her images of the accompaniments. Step 2. She asks other users who are knowledgeable about other instruments to make the accompaniment parts (e.g., bass guitar and drum). Step 3. After taking the accompaniments of music from the other users, She tries to fine-tune to bring the accompaniments to her images. Step 4. If there is a big gap between the composed accompaniments and her images, she asks the others to modify them again. Her images is often unclear and vague because her limited knowledge (e.g., how to render) about accompaniments [2]. So, the other users anticipate her images based on a composed melody part and conversations with her. At the same time, they have to compensate the lack of information to compose accompaniment parts, based on their knowledge and experience. In general, there is a big gap of knowledge between users because expertise in music instruments is different from each other. Therefore, it is too difficult for a melody composer to exactly convey her images to others. All the members involved in the collaborative music composition need to spend much time and efforts to complete satisfying compositions because they must repeatedly send and take the music data to construct mutual understandings among them.

3 Marble Marble is a system which can modify synchronized musical data mutually among members. Marble consists of two parts; the user interface to compose music among multiple user simultaneously, and the data control part to synchronize musical data as users can modify music in real-time. Figure 1 shows the interface of Marble. Track provides a function to alter tune and volume of musical notes. Score Edit Space is to modify musical notes. Musical score editor adopted the Piano roll interface which is widely used for DTM software. It can be used to externalize user’s musical images and to share them among members. Users can easily find who is composing which parts because users can compose music in real-time while watching others’ composition process. Moreover, users can intervene in other users’ composition process on musical score editor. 3.1 Proceed of Marble The collaborative music composition with Marble proceeds as follows; Step 1. Users respectively compose music parts based on their musical images on each musical editor. Step 2. Users compare composed music with her images while viewing each composition process listening to music as needed. Step 3. Users intervene in other user process of composition to directly adjust others’ music parts to get close to their images Users repeat Step1~3 until they agree with the quality of composed music.

366

Y. Takai, M. Ohira, and K.-i. Matsumoto

Fig. 1. Collaborative Music Composition with Marble interface

3.2 Problem of Marble about Evaluation In the previous study, we have focused on the time to be required to compose music with Marble. From the experiment, we have found that Marble could help degrease the time to compose music in a process of the online collaborative music composition. We believe the more important thing is whether users can satisfied with the quality of music composition through the collaborative composition. In this paper, we quantitatively evaluate the quality of music composed with EMD (Earth Mover’s Distance).

4 EMD (Earth Mover's Distance) 4.1 Outline of EMD EMD is a method to evaluate the similarity between two signatures. A signature is a set of features and their corresponding weights. Computing the EMD is based on a solution to the transportation problem for which efficient algorithms are available [3]. Given two and Q q ,w ,…, q ,w , where signatures P p ,w ,…, p ,w p and q are the points in some Euclidean space, the musical feature in our study, and

Effects of a Synchronized Scoring Interface on Musical Quality

367

w , w are the weights of the points. We want to find a flow C, with C the flow between p and q , that minimizes the overall cost C |p

q|

subject to the following constraints; C

0,

1

i

m,

1

j

(1)

n

C

w ,

1

i

m

(2)

C

w ,

1

j

n

(3)

C

min w , w

(4)

Constraint (1) allows shipping of supplies from a supplier to a consumer and not vice versa. Constraint (2) forces the consumers to fill up all of their capacities and constraint (3) limits the supply that a supplier can send as a total amount. Constraint (4) force to move the maximum amount of the flow possible. The EMD can be defined as: EMD p, q

∑

∑

∑

q

∑

C p ∑ C

∑

C p

q

min w , w Then, we will illustrate the application of EMD to this study in the following. 4.2 Application of This Study In our study, the features are the musical features (Note Position, Pitch, and Transition of Pitch) and the weights are length of a musical note. “Note Position” signifies where the musical note from start point is, a beat is a musical note having the time value of an eighth of a whole note. "Pitch" signifies the difference of pitch from front note. "Transition of Pitch" signifies the transition of pitch from the pitch of note on start place. We define that the chromatic scale is 1 on pitch and transition of pitch.

368

Y. Takai, M. Ohira, and K.-i. Matsumoto

Fig. 2. Example of the calculation of EMD (Simplified values)

5 Experimentation We conducted an experiment for comparing the quality of music composed using Marble and existing computer music software without a synchronized scoring interface. The object of the experiment is to evaluate the complete of music composed with a synchronized scoring interface and find the factor from process of composition. Then, we examined whether it is different of complete music composed by the environmental condition about synchronization with a set piece of music we prepared. Further, we analyzed the activation of argument from utterance data in experiment. Materials. The experimental materials are to compose a accompaniment part which a melody has composed, by a couple of users. The set piece of music which we prepared consists of two parts (a melody and an accompaniment) and 4 5 bar The music is original to affect the materials. We attempted to compose the set piece of music without the difficulty of its, we consider to user the same number in each condition with environment not to depend the result of experiment by the difficulty.

～

Table 2. Definition of Subjects DTM Skill Novice User

Medium User

Feature User hasn’t experience of musical performance and using DTM. User can know the usage of the piano roll, but cannot compose by herself because user does not have musical knowledge. User has experience of musical performance or using DTM. But user does not have the skill to compose the high-level music because he does not only have a specific musical skill.

Effects of a Synchronized Scoring Interface on Musical Quality

369

Environment. We compared Marble (Synchronous System) with a traditional system (Asynchronous System). We utilized Score Editor of Marble as a traditional system to remove other factors as much as possible. Marble has the synchronized function, but we switched off the function for the traditional system. Subjects used the Skype video conference to communicate with others in remote sites. Subjects. 10 couples of 20 subjects in total joined the experiment. Each couple of subjects well knows each other and consists of a novice and a medium-skill user of DTM in accordance with the definition of subjects (see Table 2). A novice subject played the role of a melody composer and the medium subject’ role was an accompaniment composer. Procedure. After we fully explained each role, the tasks, and how to operate Marble, the subjects performed a trial task to accustom to the operation of Marble. Then the subjects started to experimental tasks as follows; Step 1. A melody composer memorizes an accompaniment image after he listens to music which is given by experimenters. Step 2. Step 3. Step 4. Step 5.

Both subjects take music data only consisting of a main melody. Both subjects communicate each other using the Skype video conference. An accompaniment composer listens to a melody part, start to experiment. The melody composer tells her accompaniment image to the accompaniment composer. The accompaniment composer composes and modifies the musical note. Step 6. One task finished when subjects reported that composition was over, or it took the maximum 15 minutes from the beginning of the task.

6 Results We evaluated the quality of music composed using Marble and the traditional system and analyzed the number of subjects’ utterances through the collaborative music composition. 6.1 Music Quality Table 3 shows the results of the evaluation based on the EMD value (0 is the best score). It shows that the average and median of EMD values (means music quality) are lower in Marble than the traditional system. Table 3 also shows that the standard deviation of the EMD value of traditional system is larger than Marble. There was also a significant difference of EMD values between Marble and the traditional system (p < 0.05). It indicates that Marble can successfully help users compose music as users can get close to their musical images, compared to the traditional music composition system.

370

Y. Takai, M. Ohira, and K.-i. Matsumoto Table 3. Difference of EMD values between Marble and the traditional system Marble 1.40 1.16 1.45

Average Median Standard Deviation

Traditional System 2.06 2.22 2.28

6.2 Utterances We found that the music quality is higher in Marble. However, it is necessary to clarify what kind of situation in the collaborative composition process has an effect on the EMD value. So, we analyzed discussions among subjects in the collaborative composition from using video data. We counted the number of utterances when speakers changed or remained silent more than constant distance (around 3 seconds). In the following, we describe the three sorts of utterances: (1) All utterance, (2) Convey message, (3) Proposal message. (1) “All utterance” is the utterance of all that a subject utters including a beck and an instructions word, regardless of the length of an utterance. (2) “Convey message” is the utterance conveying the musical image from a melody composer to other users; humming and “Higher.” “I think it is correct.” etc. (3) “Proposal message” is the utterance proposing to confirm the musical images and urge a main composer to modify from a melody composer to other users; “is it OK?” and “Is it correct although this musical note is disharmony?” etc. Table 4. The number of utterance in collaborative composition of music with each system

Number all utterance Convey message Proposal message Number per minute all utterance Convey message Proposal message

Marble 71.65 31.9 14.85 Marble 5.81 2.63 1.23

Traditional System 63.15 26.6 7.65 Traditional System 4.48 1.89 0.54

p-value 0.05 0.04 0.001 p-value 0.001 0.001 0.001

The result of table 4 is separated into two parts: the upper part is representing the average number of utterance in all time of experiment and the lower part is representing the average number of utterance per minute because collaborative composition time is different. Also, we eliminated the time to send and take the music data with Asynchronous system. It was found that Marble has made a significant difference (5%) for the number of all sort of utterance. Also We found that Marble has increased the total number of utterances among users. In particular, the number of utterances to suggest music ideas to partners has been

Effects of a Synchronized Scoring Interface on Musical Quality

371

doubled. Therefore, it seems that the subjects performed a constructive discussion in collaborative composition with Marble. 6.3 Discussion In the collaborative composition with Marble, the degree of difficulty of tasks less influenced the quality of composed music. In general, casual users can convey the rhythm of music to others correctly, but they often face the difficulty in discussing the pitch of music. In many cases of our experiment, subjects tried to complete music while they discuss the pitch again and again. Compared with the traditional system, Marble seemed to ease the difficulty because the subjects with Marble could see other subject’s operations on musical notes and then understand the intention of the operations. In this paper, we used relatively short music as tasks. In case of composing more real, longer music, the differences of interfaces for music composition would have a great impact on the collaboration among users, resulting in the music quality. We also observed that each subject performed a task alone immediately after beginning the task. Especially in case of the traditional system, the subjects tend to individually compose each music part without communicating with other subjects since it was difficult to know what others were doing. In fact, the number of the “proposal utterance” was larger in using Marble. The awareness of other’s operations would trigger to discuss the operations and then increase the number of proposals for changing positions of notes on other’s musical notes. As a result, the subjects with Marble seemed to be able to compose the higher quality music.

7 Related Work In related work of asynchronous system, Latta suggested Netjam[4], which one of first system supported collaborative composition through internet, supported to improve the give and take of MIDI file with e-mail and Jorda suggested FMOL[5], which the system supported the public users join collective composition on the internet by real-time based on e-mail. In related work of synchronous system, there is hardly a case study, as for many research only proposed the new interface[6][7]. These researches are differently circumstanced to this study, which we evaluate the musical quality and analyze the collaborative composition process, while the system to stimulate the musical idea and facilitate on the process of collaborative composition.

8 Conclusion In this paper, we analyzed the effects of a synchronized music scoring interface of on the music quality. We conducted an experiment to measure the quality of music composed using a synchronized interface (Marble) and asynchronous interface (traditional system). As a result of the experiment, we found that the users with Marble can correctly convey musical images to each other and then can compose

372

Y. Takai, M. Ohira, and K.-i. Matsumoto

better quality music, compared with a traditional system. We also found that the number of the “proposal utterances” in using Marble was twice larger than the traditional system, resulting in creating higher quality music. In the future, we would like to develop a system for supporting the collaborative music composition among strangers.

References 1. Ohira, M., Kimura, M., Mastumoto, K.-i.: Marble: A System for Supporting Online Collaborative Music Composition among Casual Users. Human Interface Journal 12(3), 219–229 (2010) (in Japanese) 2. Sloboda, J.A.: The Musical Mind: The Cognitive Psychology of Music. Oxford University Press, Oxford (1985) 3. Rubner, Y., Tomasi, C., Guibas, L.J.: The earth mover’s distance, multidimensional scaling, and colorbased image retrieval. In: Proc. of the ARPA Image Understanding Workshop, pp. 661–668 (1999) 4. Latta, C.: A New Musical Medium: NetJam. Computer Music Journal 15 (1991) 5. Jorda, S.: Faust Music On Line (FMOL): An approach to Real-time Collective Composition on the Internet. Leonardo Music Journal 9 (1999) 6. Barbosa, A., Kaltenbrunner, M.: Public sound objects: A shared musical space on the web. In: Proc. Intl. Conf. on Web Delivering of Music (2002) 7. SoundWIRE., http://ccrma.stanford.edu/groups/soundwire/

Measurement of Tagging Behavior Differences Li-Chen Tsai1, Sheue-Ling Hwang1, and Kuo-Hao Tang2 1

Department of Industrial Engineering and Engineering Management National Tsing Hua University, Hsinchu, Taiwan 2 Department of Industrial Engineering and Systems Management Feng Chia University, Taichung, Taiwan

Abstract. This paper presents the result of a practical study comparing how domain expertise readers tag documents with different descriptions. We conducted an experiment comparing tagging behavior of experts and novices groups and intended to discover which group would generate more reliable and more representative sets of tags when they were asked to provide tags for document bookmarks in a Mozilla Firefox browser. In this analysis, we defined two measures, similarity and relevance, to describe the differences between the two groups. Result from this experiment showed that tags chosen by experts yielded better similarity and relevance in both analyses. Tags chosen by the expert group had higher commonality in pairwise similarity analysis; further, the relevance analysis showed that tags chosen by experts reflected better understanding of the content. Tagging behavior has become highly popular on the web; implications on the design of future social information system are discussed. Keywords: keyword-based tagging behavior, expert, novice, similarity, relevance.

1 Introduction Social tagging is one of the most popular online communications and collaboration phenomena, evolving the Web from a static platform into an actively shared information space. Although many authors have discussed the tagging of “informal information” such as personal web pages and blogs, tags for “hard content” in professional domains have emerged as an important trend. For example, some databases of academic articles, such as CiteULike and Connotea, allow users to create their own tags. Articles with such “hard content” are not often understood by a large number of people, thus tags assigned by non-experts might be of little use. Campbell (2006) compared the semantic web to social tagging systems within a phenomenological framework adapted from Husserl (Husserl, 1929) and found that an intersubjective consensus of tagging systems emerged from important entities and from relationships based on loose structure and high flexibility. Campbell’s results were echoed in a study by Halpin et al.(2007) that pointed out that a meaningful ontology was extracted by the tag frequency distribution of the short head rather than A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 373–377, 2011. © Springer-Verlag Berlin Heidelberg 2011

374

L.-C. Tsai, S.-L. Hwang, and K.-H. Tang

that of the long tail. The most frequently used tags in commonalities always happened at the “short head” and were combined with stability and information value. It is believed that a set of shared “intersubjectivity” concepts, or, “short head,” is more achievable through the collaboration of a panel of domain experts, particularly for the tags describing “hard content” since tags are mostly likely extrinsic to taggers. Pak suggested that in order to generate item-specific tags for articles and web pages that are difficult to describe, deep processing of domain knowledge and good verbal ability are generally required (Pak et al., 2007). For this reason, “expert opinions” would be a solution to the problem of tag inconsistency, particularly for academic or scientific documents, where the content must be known in sufficient depth. In such circumstances, tags generated by experts might have a greater probability of accurately representing the relevant knowledge. In this paper, we investigate whether experts would generate a more reliable and representative set of tags for academic and scientific documents than novices would generate.

2 Method First, we collected domain-related web pages from the Internet and imported into a directory of HTML files as a document bank for our later experiments. We recruited twenty subjects participated in this experiment and divided into two groups, one of novices and one of experts. The members of the expert group were materials science experts, but they were not tagging experts. The expert group consisted of ten professionals from a university department of Materials Science and Engineering. Two of the experts were Ph.D. candidates, and the others had Ph.D. degrees and were undertaking post-doctoral work. These two groups were asked to provide tags for document bookmarks in a Mozilla Firefox browser. During the experiment, each user was first familiarized with Firefox bookmark tags in a 10-minute introductory session; immediately afterward, the user was required to tag the 10 pretest articles. Participants were explicitly asked to tag articles using noun phrases and to assign exactly 5 tags for each article. After the pretest training session, each participant took a 10-minute break, and then tagged the rest of 25 articles. Next, we defined two measures, similarity and relevance; these measures transformed qualitative data (noun phrase tags made by the users) to quantitative data. Consistency is the extent to which agreement exists on the terms used to index the same document. Tag similarity measures the level of consistency between tags. Tag relevance measures whether tags reflect the subject matter or concepts of an article. In our experiment, tag relevance was assigned by a domain-specific expert with extensive nanomaterial experience.

3 Results Descriptive statistics showed that among all 1250 (5 tags * 25 articles * 10 participants) tags generated by each of the groups, the experts generated 612 different tags and the novices generated 711 different tags. Fig. 1 illustrates tag frequency distributions for all participants which are similar to those of Halpin et al.(2007). Fig. 2 and Fig. 3 show the

Measurement of Tagging Behavior Differences

375

“short head” and “long taiil” respectively for both expert and novice groups. Thhese two figures show that expeert group has a more prominent “short head” while novvice group generates a longer taail. The tags of the expert group followed the power law: the top 10 tags accounted for f 39.2% of all tag assignments and created a prominnent “short head.” The correspo onding number for the novice group was 23.3% (Topp 20 tags for expert was 57.1%, and for novice was 36.2%). The novice-assigned tags had a longer tail than expert-asssigned tags had.

Fig. 1. Taag frequency distributions for all participants

80

Taag Frequency Distribution: short head Frequency

60 Expert

40

Novice

20 0 1

2

3 4 5 6 7 8 Top 10 tags (different tags)

9

10

Fig. 2. The frequency of top chosen c tags by experts and novices, showing expert group hhas a more prominent “short head”

Three levels of sim milarity and relevance (high, medium, and low) w were distinguished in 25 articles by K-means methods. Fig. 4 illustrates the similarity and a expert groups. In the novice group, 92% of the articcles relevance for both novice and fell into the medium and low relevance categories and 89% of the articles fell into the H in expert group, 88% of the articles fell into the low similarity categories. However, high and medium relevance categories and 96% of the articles fell into the low and medium similarity categoriies. This suggests that expert tag choices were of higgher quality and consensus. Thee similarity of experts was higher than that of novices for most articles.

376

L.-C. Tsai, S.-L. Hwang, and K.-H. Tang

400

Tag Frequency Distribution: long tail

350

358

Number of tags

300

Expert

297

Novice

250 200

205 206

150

147

100

108

50 0 counts equals 1

counts equals 2 Tag counts

counts more than 3

Fig. 3. The number of tags chosen once, twice, or more than three times by two groups of participants among the 25 documents ϲϰй

ϲϬй ϱϬй

ϰϬй

ϮϬй

ϭϮй

ϯϲй

Ϯϴй

ϯϬй

ϰϴй

ϰϴй

ϰϬй

ϱϲй

3HUFHQWDJHVRIDUWLFOHV

WĞƌĐĞŶƚĂŐĞƐŽĨĂƌƚŝĐůĞƐ

ϳϬй

ϭϮй

Ϯϴй

Ϯϰй

ϭϮй

ϭϬй Ϭй

,ŝŐŚ

DĞĚŝƵŵ >Žǁ ZĞůĞǀĂŶĐĞ ǆƉĞƌƚ EŽǀŝĐĞ

+LJK

0HGLXP ^ŝŵŝůĂƌŝƚǇ

ǆƉĞƌƚ

/RZ EŽǀŝĐĞ

Fig. 4. Similarity and relevance for expert and novice groups

To determine whether and how participants’ expertise levels influenced different tags, we used analysis of variance (ANOVA) to analyze the collected data. In the ANOVA analysis of pairwise similarity, the participant expertise level was the independent variable and similarity was the dependent variable. The results indicated that the participants’ expertise levels had a significant effect (F(1, 89)= 40.029, p<0.01). The mean of experts was 0.2665 (SD=0.065), higher than that of novices (Mean=0.195, SD=0.041). The results suggest that experts tended to agree on the tags that ought to be assigned to particular documents, and novices did not agree as much as experts. In the ANOVA analysis of relevance, participant expertise level was the independent variable and relevance was the dependent variable. A one-way ANOVA analysis of tag relevance revealed a significant effect, F(1, 19)= 39.548, p<0.001. The mean of experts was 1.4352 (SD=0.0789), higher than that of novices (Mean=1.2288, SD=0.0674). This result resembled the similarity result, in that the tags assigned by experts were more relevant. The ANOVA relationships of similarity and relevance confirmed our expectations.

Measurement of Tagging Behavior Differences

4

377

Conclusions

The purpose of this study is to examine whether experts can provide a more consistent and representative set of tags for academic and scientific documents than novices can generate. Our findings show that the tags assigned by the expert group differed from those assigned by the novice group, especially in pairwise similarity. Tags chosen by experts had more consistency, and reflected domain knowledge and competent understanding. Moreover, pairwise similarity among experts was higher than among novices, which may have been due to variations in domain knowledge, textual comprehension, or literacy. Experts could perceive a key point and ignore any interference from unrelated information, while novices could conjecture by textual literacy and common-sense judgment, which resulted in superficial understanding. When novices faced a reading barrier, they tended to ignore it or skip over it, and to generate a tag by straightforward judgment. The readers’ behaviors were also reflected in tag relevance, in that novices’ tags mostly had medium and low relevance, and often had medium and low pairwise similarity. Tagging is a “bottom-up” method of organization that reflects users’ real knowledge and cognition, and our findings suggest that domain-specific experts can assist the functioning of our information society by generating well-chosen tags.

References 1. Campbell, D.G.: A phenomenological framework for the relationship between the Semantic Web and user-centered tagging systems. In: Furner, J., Tennis, J.T. (eds.) Proceedings of the 17th ASIS&T Classification Research Workshop, Austin, Advances in Classification Research, vol. 17, pp. 1–17 (2006) 2. Halpin, H., Robu, V., Shepherd, H.: The Complex Dynamics of Collaborative Tagging. In: WWW 2007, Banff, Alberta, Canada (2007), http://www2007.org/papers/paper635.pdf (accessed February 6, 2010) 3. Husserl, E.: “Phenomenology”, Encyclopedia Britannica. In: Kearney, R., Rainwater, M. (eds.) The Continental Philosophy Reader, pp. 15–22. Routledge, London (1996) (reprinted) 4. Pak, R., Pautz, S., Iden, R.: Information organization and retrieval: An assessment of taxonomical and tagging systems. Cognitive Technology 12(1), 31–44 (2007)

Community Search: A Collaborative Searching Web Application with a User Ranking System Papon Yongpisanpop, Masao Ohira, and Ken-ichi Matsumoto Graduate School of Information Science, Nara Institute of Science and Technology 8916-5, Takayama, Ikoma, Nara, Japan +81-743-72-5312 {papon-y,masao,matumoto}@is.naist.jp

Abstract. People are using search engine in daily life. But most of the tools that we have today treat information-seeking tasks as a transient activity. In this research paper we introduce a web application system that provides collaborative function and experts finding system. We develop a system that will help user to organize search result and to do the collaboration with others. With the new iterative algorithm, users will also gain more percentage of needed result and the system will be able to suggest the experts related to the search keyword. Keywords: Search Engine, web mining, web search interface, collaborative search.

1 Introduction Although search engines have become a part of our daily lives and allow people to access the vast quantities of information available online, tools that we have these days typically treat information-seeking tasks as a transient activity, and consider each of a user’s search and browse action as unrelated events rather than as part of a larger task. For example, if a user is searching for a car to buy in the future, he/she will perform several low-level interactions with a web browser and one or more search engines (e.g., typing the URL of a search engine site, querying for “Nissan Skyline,” clicking a result, hitting the back button, submitting a new query and so on). In other words, tools such as web browsers and search engine sites are designed to support single-user scenarios. They provide the contents of the web page from the query of a user, but they don’t help the user to decide which results are good and useful. In the past few years, Microsoft researchers started to study these problems. They separated the topic into two main problems and created tools for solving the problems: Searchbar [1] and SearchTogether [2]. Searchbar is a plug-in for the Microsoft Explorer browser to keep track of and organize a search query, so a user can continue the same search at a later time without having to start over. SearchTogether is software that runs on the Windows operating system and provides a means for collaborative search in which users can help each other when searching together for specific topics. However, These two tools still do not cover the entire problem in supporting people’s information seeking activities. The problem with the A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 378–386, 2011. © Springer-Verlag Berlin Heidelberg 2011

Community Search: A Collaborative Searching Web Application

379

first tool is the user is still doing the individual search. The program just creates an interface and organizes the contents to avoid the same task in the future. The problem with the second tool is you need to have software running on the Windows operating system, which means it is not really convenient for users of other operating systems. Another problem is sometimes a search engine will return the results based on the score that the algorithm calculate it from how many time that this page has been clicked which mean that there is no guarantee for the users that the result will match for his/her purpose. Also when user can’t find any good result, sometimes user needs to ask some one who has an expertise about the field. In this paper, we propose the concept Community Search, which is a community based web technology that acts itself as a Meta search portal and allows users to manage search result contents. It provides the collaborative searching function for users to share search results with others and to get the contents that has been analyzed by real users in the community. It also suggests a user the existences of experts who are related to user’s search keywords, so that user can have reliable information the experts recommend. Community Search is implemented using our new iterative algorithm. This algorithm is derived from HITS (Hyperlink-Induced Topic Search) algorithm [3] that is the most famous algorithm for search engine. The purpose of this algorithm is to help users get a better search result by iteratively running it. The algorithm is used to collect keywords and links which connect the web pages to users and web pages to web pages. Then it weights and gives the score for each web page and also ranking user’s relevance to a topic through link analysis on web logs We need to make our algorithm to have enough accuracy so that can drive the community search to return good results and be able to suggest the experts related to the keyword that a user is searching for. The rest of this paper is organized as follows. In section 2, we present related work on organize search query, collaborative searching and expert finding. In section 3, we describe the concept of Community Search. In section 4, we present the proposed approach of Community Search’s collaborative searching function and our new iterative algorithm that can calculate score for users and web pages. Then, we show how Community Search classifies experts. In section 5, we will give an example on usage scenario. In section 6, we will discuss about the usage of Community Search. And finally the conclusion and future work will be the last part in section 7.

2 Related Work A number of web pages ranking [8], automatic expert finding and collaborative search system have reported it so far. For example in automatic expert finding system, HelpNet [5] and Expert/Expert-Locator (EEL) [6][7]. [5] allow user filled information to provide the expertise profile and [6][7] use a representative collection of user’s technical document to build the expertise index. There are a lot of previous system that use textual analysis as the basis for expert finding but [9] uses graph analysis to find interlinked cliques for interest group and [10] that uses users’ visit path record of experts. So our study will focus on [9][10]’s technique and adapt it into our iterative algorithm. As we mentioned earlier about 2 of Microsoft research that we have been studied them, which are SearchBar [1] and SearchTogether [2]. Searchbar is a user interface for web search, including browsers and search engine sites, which is a system for proactively and persistently storing query histories, browsing histories, and

380

P. Yongpisanpop, M. Ohira, and K.-i. Matsumoto

users’ notes and ratings in an interrelated fashion. For the SearchTogeter is software designed to enable either synchronous or asynchronous remote collaborate when searching the web. Both of these systems are designed to help user to make searching more productive in the interface level. But what we need in our research is to develop a system that can be able to help user to do the collaboration, organize search result contents and also classify the experts out of the community. Then we have found a research called “Ranking User’s Relevance to a Topic through Link Analysis on Web logs” [4], which is close to what we need to make Community Search smart enough to return a better search result and be able to classify experts. Basically this research derived HITS algorithm [3] to a link analysis algorithm, which evaluates the web user’s level of expertise on a given topic by his/her web browsing record.

3 Community Search Community Search is a platform independent, web-based technology that allows users to manage search result contents and also provides the collaborative searching function for users to share and obtain contents from others. For the advantages, Community Search provides users with a means for gathering people who are interested in the same topic and/or share the context of work/hobby/everyday life, it will change the old style of searching and make searching more productive and useful. The structure of Community Search is separated into two parts. Front-end which provides collaborative searching and search result organize functions for the users. Back-end part, which is using our new iterative web pages and users, ranking system that will be able to calculate score for web pages and users for the better search result. Also the new iterative algorithm will be able to classify experts that related to the keyword that users use for search out of the community, we first describe the general model and the link analysis algorithm and then we describe the procedure of how the algorithm giving score for web pages and users and also classify the experts from users habit. When user uses Community Search to do the search, our system will keep track of what action user does. For Example, user might query, “Ruby on Rails”. And then click on a few links to look at the content of the pages. User might also bookmark the page that he/she thinks that is match with his/her purpose. Community Search will use that information to calculate the score and ranking the web pages and users. Also will be able to classify the expert among the users. Next, we will explain our iterative algorithm that derived it from HITS algorithm [3].

4 Implementation We separated the implementation of the Community Search into 2 parts. First is the front-end part, we designed the interface based on 2 main functions, which are search result organize function and collaborative searching function. In this part we need to study a user behavior that using the system from the previous researches, which are SearchBar [1] and SearchTogether [2]. And then we will design the interface that fits for Community Search using Ruby on Rails1. Second is the back-end system is 1

Ruby on Rails is an open-source web framework that’s optimized for programmer sustainable productivity.

Community Search: A Collaborative Searching Web Application

381

implemented using our new iterative ranking algorithm. This algorithm is derived from HITS (Hyperlink Induced Topic Search) algorithm [3] that is the most famous algorithm for search engine. The purpose of this algorithm is to help users get a better search result by iterative running it. The algorithm is used to collect keywords and links which connect the web pages to users and web pages to web pages. Then it weights and gives the score for each web page and also ranking user’s relevance to a topic through link analysis on web logs. The data in Community Search can be model as a directed graph, as shown in figure [1]. In this model there are two kind of entities represented by nodes, which are users and the web pages. There are 3 kinds of directed edges between the nodes. The dark lines with white arrow head are represented the links between web pages. The dash lines are when the users bookmark the web page. And the dark lines with small arrow head are represented the links from users to web pages (when user clicks to visit the page). In this graph, the edges between web pages are hyperlinks links and the edges from a user to a page indicates that the user visited or bookmarked the page. The directed edges can be assigned numeric weights to reflect the different importance. To define the expert, the visit and bookmark edge from those users may be assigned with higher weights.

Fig. 1. The directed graph model showing the interaction between users and web pages in Community Search

4.1 The Iterative Algorithm As we mentioned, this iterative algorithm is derived from Kleinberg’s HITS algorithm [3], There are three assumptions in calculating the importance of users and web pages. The first assumption is that the more high quality web pages the user has visited, the user will gain more experience. The second assumption is the more frequently a web pages is cited by other pages and visited by experienced users, the higher quality the web page is. And the last assumption is the importance/quality of the users and pages may reinforce each other in an iterative way.

382

P. Yongpisanpop, M. Ohira, and K.-i. Matsumoto

a( p) = β ∑ h(q) + (1 − β) ∑ u(r) q→ p

r→ p

h( p) = β ∑ a(q) + (1 − β) ∑ u(r) p→q

r→ p

⎡ ⎛ ⎛ ⎞ ⎞⎤ u(r) = ⎢α⎜⎜ ∑ a( p) + ∑ h(q)⎟⎟ + (1 − α )⎜⎜ ∑ a(y) + ∑ h(z)⎟⎟⎥ ⎢⎣ ⎝ r → p ⎝r→ y ⎠ ⎠⎥⎦ r→ q r→ z Fig. 2. The iterative ranking algorithm that runs on the background of the Community Search to ranking user and giving score for web pages

In this algorithm, to calculate the authority of weight of a web page p, the sum of hub values of all pages q pointing to p and the sum of weights of all users r visited p are combined to form the final authority weight of p. The hub weight is similarly calculated. The weight of user r is calculated by summing up authority and hub weights of all pages he/she has visited or bookmarked. Then the weights of web pages and users reinforce each other in an iterative way according to our second assumption. In the algorithm, β is the parameter to adjust if the system needs to weight more score on when the user clicks on the page. And α is the parameter to adjust if bookmark is more importance to give more weight score to the user. 4.2 How to Classify Experts Community Search can identify experts for a given query topic and rank users’ level of expertise by collecting the keywords that user used to search and then categorize web pages that users interact with (click and bookmark). Then the iterative ranking algorithm will calculate score for users and web pages. The result from the algorithm will be able to tell which topic users have been experienced on. To do this, we need to create a program to collect keywords and be able to categorize them into their own categories that we already prepared beforehand in the system. This program divided into two parts. The first part as show in the figure [3] is to create a model that will be able to predict which category the keywords should be. The model that we will use to predict which category the keywords should belong to is needed to be train first. The step of training the model has shown in the figure [3]. First we have to prepare all the possible categories that we can have in the system. For example, if you are using the Community Search in the Software Company, most of your category should involve with software and programming languages. After we define the possible categories that we should have. Then we will build up the model based on the hypothesis that if the documents have a lot of the same keywords of the category; it means that this document should be in the category. For example, the documents in “Java Programming” category should have a lot of “java” keywords inside the document. So in figure [3], We send the list of our category to search in the Google. Then we check the web pages that returned from the Google to see what

Community Search: A Collaborative Searching Web Application

383

Fig. 3. Showing the step of how to build a classification model

keywords that the web pages contain and we use the Bayesian network2 to build the classification model. The second part as shown in the figure [4] is to predict which category the keyword should be by using the trained model from the figure [3]. After we got the classification model from figure [3], Then when user insert the keyword to search, we will send that keyword to Google again and give the return result from the Google to the model that has already been trained. Finally The System will now be able to tell which keyword should goes to which category

Fig. 4. Show the steps of the classify process. User inserts the query, which is “Ruby on Rails”. Then the system use Google to search the result from the keyword. At last using to trained classification model to categorize the category of the keyword, which is “Web Programming”.

2

Bayesian network is a probabilistic graphical model that represents a set of random variables and their conditional dependencies via a directed acyclic graph.

384

P. Yongpisanpop, M. Ohira, and K.-i. Matsumoto

5 Example Usage Scenario Community Search will work best if we deploy it into a specific organization. Because of The topics that use to classify experts can be limited due to which type of the organization. Also the collaborative function will be really useful in the situation when you have to search as a team. For example, we have a car dealer company. And these is a situation that a customer call and ask for a new type of a car, which our company doesn’t have any information yet.

Fig. 5. Briefly illustrates the idea and flow of user behavior in Community Search

[Step 1] User search for a Nissan Skyline, which is the car that customer wants to know the information. Then obtain search result contents about the car from search engine. [Step 2] User analyzes the contents of the results and may find useful information. [Step 3] User can organize the contents by using Search Organize Function in the Community Search. And also user can be able to share this information out to the Community Search. Other users who have the same interests into this information can use it. [Step 4] Next time, when a customer calls and asks for “Nissan Skyline” again to the difference user. The new user can search and use the contents that have been shared by the user before. Or when the user cannot really find the information that he/she needs. Community Search will be able to suggest experts who related with the topic.

Community Search: A Collaborative Searching Web Application

385

6 Discussion In this section, we discuss about the usage of Community Search. As we said in the example usage scenario part that Community Search will work best if deploy in some sort of organization. First reason, in the classify experts system part; we need to limit the keywords in the system because we need to prepare the category of the keywords that can be possible for user to search beforehand. For example, if you deploy Community Search in a car dealer company. The limitation of keywords would be something only about cars (truck, four-wheel drive, sedan, etc.) So the system doesn’t need to collect the unnecessary keywords to classify the expert. In the car dealer company, the system will have the experts that involves only with cars. Second reason, in the organization, people are always assigned to do the task as a team. In this case, the collaborative function will work best in the situation that they need to do the research together. The system needs sometime to collect web logs data to analyze and calculate score for web pages and users before it can classify the experts and re-ranking web pages. When users use Community Search, the system will gain more data to analyze and increase the percentage of accuracy.

7 Conclusion and Future Work We presented Community Search, a collaborative searching web application with a user ranking system. The goal of this study is to make a system that can help users to search more productively. In our study, we found that the interface design of Community Search will be able to help supporting the group of organization to do the collaborative searching and organize search result. Also we found the percentage of gaining needed search result contents increases by using the community search’s iterative 2algorithm. For the future work, right now we completed the program of iterative ranking algorithm that will be running in the background of the Community Search. Next step, we need to make an interface, which is the front-end part of the system that helps users to organize search result contents and do the collaborative searching. In this part we need a really well design for the interface and we will use Ruby On Rails, which is a web application framework to build up our interface system.

References 1. Morris, D., Morris, M.R., Venolia, G.: SearchBar: A Search-Centric Web History for Task Resumption and Information Re-finding. In: Morris, D., Morris, M.R., Venolia, G.,S. (eds.) Proceedings of Proceeding of the Twenty-Sixth Annual SIGCHI Conference on Human Factors in Computing Systems (CHI 2008), pp. 1207–1216 (2008) 2. Morris, M.R., Horvitz, E.: SearchTogether: An Interface for Collaborative Web Search. In: Proceedings of the 20th Annual ACM Symposium on User Interface Software and Technology (UIST 2007), pp. 3–12 (2007) 3. Kleinberg, J.: Authoritative sources in a hyperlinked environment. Journal of the ACM (JACM) 46(5), 604–632 (1999)

386

P. Yongpisanpop, M. Ohira, and K.-i. Matsumoto

4. Jidong, W., Zheng, C., Li, T., Wei-Ying, M., Liu, W.: Ranking User’s Relevance to a Topic through Link Analysis on Web Logs. In: Proceedings of the 4th International Workshop on Web Information and Data Management 5. Maron, M.E., Curry, S., Thompson, P.: An Inductive Search System: Theory, Design and Implementation. IEEE Transaction on Systems, Man and Cybernetics, SMC-16(1), 21–28 (1986) 6. Steeter, L.A., Lochbaum, K.E.: An Expert/Expert Locating System based on Automatic Representation of Semantic Structure. In: Proceedings of the Fourth IEEE Conference on Artificial Intelligence Applications, pp. 345–349. Computer Society of the IEEE, San Diego (1988) 7. Steeter, L.A., Lochbaum, K.E.: Who knows: A System Based on Automatic Representation of Semantic Structure. In: RIAO 1988, Cambridge, MA, pp. 345–349 (1988) 8. Brin, S., Page, L.: The anatomy of a large-scale hypertextual Web Search engine. In: 7th WWW Conference, Brisbane, Australia, Computer Networks and ISDN Systems, vol. 30(1-7) (1998) 9. Schwartz, M.F., Wood, D.M.: Discovering Shared Interests Using Graph Analysis. Communications of the ACM 36(8), 78–89 (1993) 10. Cohen, A.L., Maglio, P.P., Barrett, R.: The Expertise Browser: How to Leverage Distributed Organizational Knowledge. Presented at Workshop on Collaborative Information Seeking at CSCW 1998, Seattle, WA (1988)

Author Index

Ahmad, Rahayu 3, 81 Alducin-Quintero, Gerardo Aliprandi, Carlo 90 Baek, Sung-Wook 12 Bailey, Brian P. 219 Baranauskas, M. Cec´ılia C. Ba` u, Roberto 172 Bonacin, Rodrigo 345 Borning, Alan 100 Braman, James 20 Brunswicker, Sabine 203

153

345

Damianos, Laurie E. 162, 188 Derthick, Katie 100 dos Reis, J´ ulio Cesar 345 Drozdetski, Stan 162 Dudley, Alfreda 20 229

Ferron, Michela 172 F¨ uller, Johann 203 Gamberini, Luciano 172 Gao, Qin 110 Garc´ıa, Jos´e Javier 229 Giguere, Veronica 309 Gim´enez, Rafael 229 Goh, Dion Hoe-Lian 120 Guerra-Zubiaga, David A. Han, Tack-Don 12 Han, Young Suk 319 Harrison, Viorica 182 Ho, Yi-Lun 277 Holm, Jukka 326

188 373

Iizuka, Kayo 40 Iizuka, Yasuki 40

Carstens, Deborah S. 309 Chen, Heng 241 Choi, Jong Kyu 319 Colt, Kidd 20 Connolly, Christopher M. 30 Contero, Manuel 153 Cuomo, Donna L. 162

Enrich, Regina

Holtzblatt, Lester J. Hwang, Sheue-Ling

Ji, Yong Gu 319 Johnson, Michael D.

Kadenbach, Daniel 50 Kanawati, Rushed 336 Kelliher, Aisling 241 Kleiner, Carsten 50 Koblentz, Manami 197 Koch, Giordano 203 Kriplean, Travis 100 Lee, Chei Sian 120 Lee, Nien-Chieh 213 Lehtiniemi, Arto 326 Lewkowicz, Myriam 267 Lin, Yi-Sheng 140 Liu, Che-Hung 140 Lutters, Wayne G. 3, 81 Marchetti, Andrea 90 Mart´ın-Guti´errez, Jorge 153 Martino, Francesco 172 Masaki, Hitoshi 251 Matsumoto, Ken-ichi 251, 363, 378 Matsuo, Tomoya 297 McDonald, David W. 100 Meiselwitz, Gabriele 30, 60, 69 Minutoli, Salvatore 90 Ohira, Masao

153

153

251, 363, 378

P´erez, Mar´ıa 229 Pujari, Manisha 336 Pusey, Portia 60 Qu, Qu

110

Rahimifar, Mohsen 261 Ravi, Shankar 355 Reed, Kevin 69 Ronzano, Francesco 90

388

Author Index

Salim, Siti Salwah 261 Seo, Jong-Hoon 12 Sharmin, Moushumi 219 Spagnolli, Anna 172 Sreenivasan, Nirupama Dharmavaram 120 Takai, Yuji 363 Tan, Wee-Kek 130 Tang, Kuo-Hao 373 Teo, Hock-Hai 130 Tesconi, Maurizio 90 Tixier, Matthieu 267 Torrent-Moreno, Marc 229 Tran, Hi 213 Tsai, Li-Chen 373 Tsai, Tsai-Hsuan 277

Tsao, Patrick 100 Tseng, Kevin C. 277 Vincenti, Giovanni 20 Voiskounsky, Alexander

287

Wang, Yuanqiong 20 Wei, June 182, 197, 213 Wu, I.-Chin 140 Yin, Albert 213 Yongpisanpop, Papon 378 Yoshida, Kyoko 40 Yoshino, Takashi 297 Zachry, Mark Zhang, Xuhui

100 110