Beyond Personalization 2005 A Workshop on the Next Stage of Recommender Systems Research San Diego, January 9, 2005 In conjunction with the 2005 International Conference on Intelligent User Interfaces (IUI 2005) Edited by: Mark van Setten Sean McNee Joseph Konstan http://www.grouplens.org http://www.telin.nl http://www.multimedian.nl Table of Contents About the Workshop 5 Organization 6 Full papers Crossing the Rubicon for An Intelligent Advisor 7 Răzvan Andonie, J. Edward Russo, Rishi Dean Explaining Recommendations: Satisfaction vs. Promotion 13 Mustafa Bilgic, Raymond J. Mooney Identifying Attack Models for Secure Recommendation 19 Robin Burke, Bamshad Mobasher, Roman Zabicki, Runa Bhaumik User-Specific Decision-Theoretic Accuracy Metrics for Collaborative Filtering 26 Giuseppe Carenini Off-Topic Recommendations 31 Elyon DeKoven Item-Triggered Recommendation for Identifying Potential Customers of Cold Sellers in 37 Supermarkets Han-Shen Huang, Koung-Lung Lin, Jane Yung-jen Hsu, Chun-Nan Hsu The Good, Bad and the Indifferent: Explorations in Recommender System Health 43 Benjamin J. Keller, Sun-mi Kim, N. Srinivas Vemuri, Naren Ramakrishnan, Saverio Perugini Impacts of Contextualized Communication of Privacy Practices and Personalization Benefits on 48 Purchase Behavior and Perceived Quality of Recommendation Alfred Kobsa, Max Teltzrow InterestMap: Harvesting Social Network Profiles for Recommendations 54 Hugo Liu, Pattie Maes What Affects Printing Options? - Toward Personalization & Recommendation System for 60 Printing Devices Masashi Nakatomi, Soichiro Iga, Makoto Shinnishi, Tetsuro Nagatsuka, Atsuo Shimada P2P-based PVR Recommendation using Friends, Taste Buddies and Superpeers 66 Johan Pouwelse, Michiel van Slobbe, Jun Wang, Henk Sips DynamicLens: A Dynamic User-Interface for A Meta-Recommendation System 72 J. Ben Schafer Modeling a Dialogue Strategy for Personalized Movie Recommendations 77 Pontus Wärnestål Behavior-based Recommender Systems for Web Content 83 Tingshao Zhu, Russ Greiner, Gerald Häubl, Bob Price, Kevin Jewell Position statements Who do trust? Combining Recommender Systems and Social Networking for Better Advice 89 Philip Bonhard 3 Recommender Systems Research at Yahoo! Research Labs 91 Dennis Decoste, David Gleich, Tejaswi Kasturi, Sathiya Keerthi, Omid Madani, Seung-Taek Park, David M. Pennock, Corey Porter, Sumit Sanghai, Farial Shahnaz, Leonid Zhukov A Multi-agent Smart User Model for Cross-domain Recommender Systems 93 Gustavo González, Beatriz López, Josep Lluís de la Rosa Personalized Product Recommendations and Consumer Purchase Decisions 95 Gerald Häubl, Kyle B. Murray Toward a Personal Recommender System 97 Bradley N. Miller Beyond Idiot Savants: Recommendations and Common Sense 99 Michael J. Pazzani Towards More Personalized Navigation in Mobile Three-dimensional Virtual Environments 101 Teija Vainio Issues of Applying Collaborative Filtering Recommendations in Information Retrieval 103 Xiangmin Zhang 4 About the Workshop This workshop intends to bring recommender systems researchers and practitioners together in order to discuss the current state of recommender systems research, both on existing and emerging research topics, and to determine how research in this area should proceed. We are at a pivotal point in recommender systems research where researchers are both looking inward at what recommender systems are and looking outward at where recommender systems can be applied, and the implications of applying them out 'in the wild.' This creates a unique opportunity to both reassess the current state of research and directions research is taking in the near and long term. Background and Motivation In the early days of recommender systems research, most research focused on recommender algorithms, such as collaborative filtering and case-based reasoning. Since then, research has gone off into various directions. Some researchers continued working on the algorithmic aspects of recommenders, including a move to hybrid and group recommenders; others have been researching the application of recommenders in specific domains; yet others focused on user interface aspects of recommender systems. This has led to the current state in which recommender systems are mature enough to be applied in various adaptive applications and websites. They have been deployed on several large e-commerce websites, such as Amazon.com; they are being integrated into corporate document warehouses; and they are still the center of focus for several research groups around the world. Moreover, these systems are appearing in products and services used by people around the world, such as personalized television programming and Internet- broadcast radio stations, movie recommenders, and even dating services. This workshop aims to answer questions raised both by researchers and practitioners in order to improve both recommender quality and use. Issues discussed at the workshop will have an effect on these systems—and more importantly, the users of these systems—worldwide. Topics and Goals This workshop will focus on the following four main topics: 1. Understanding and trusting recommender systems. Do users understand and trust the recommendations they receive from recommender systems, what kinds of information do recommenders need to provide to users to build trust, and how difficult is it to regain trust in a recommender if it is lost? 2. User interfaces for recommender systems. What are good ways to present recommendations to users, how do you integrate recommenders into the displays of existing information systems, and how can interfaces encourage users to provide ratings in order to 'close the loop' for recommendations, that is, how can you get users to consume the items recommended and then tell the system how good the recommendations are? 3. The future of recommendation algorithms and metrics. How can we generate better individual and group recommendations, develop new metrics and evaluation criteria for recommendations, and achieve cross-domain recommendations? 4. Social consequences and opportunities of recommenders. How do individuals and groups of people respond to recommendations, how can recommendations be integrated with online and real world communities, and in what ways do recommendations affect social organizations? Intended Audience The workshop is intended for both established researchers and practioners in the domain of recommender systems as well as for new researchers and students with interesting ideas on recommender systems and their future. Participants do not have to come from a specific application domain, as long as their research or ideas are on one of the main topics of the workshop. Website All papers and the results of the workshop are also available online at: http://www.grouplens.org/beyond2005 5 Organization Workshop Chairs Mark van Setten Telematica Instituut P.O. Box 589 7500 AN Enschede The Netherlands E-mail: [email protected] Sean M. McNee GroupLens Research Department of Computer Science and Engineering University of Minnesota Minneapolis, MN, 55455 USA E-mail: [email protected] Joseph A. Konstan GroupLens Research Department of Computer Science and Engineering University of Minnesota Minneapolis, MN, 55455 USA E-mail: [email protected] Program Committee Liliana Ardissono - University of Torino (Italy) Jon Herlocker - Oregon State University (USA) Anton Nijholt - University of Twente (The Netherlands) Barry Smyth - University College Dublin and Changing Worlds (Ireland) Loren Terveen - University of Minnesota (USA) Additional Reviewers Betsy van Dijk – University of Twente (The Netherlands) Harry van Vliet – Telematica Instituut (The Netherlands) 6 Crossing the Rubicon for An Intelligent Advisor Razv˘ an Andonie J. Edward Russo Computer Science Department Johnson Graduate School of Management Central Washington University, Ellensburg, USA Cornell University, Ithaca, USA [email protected] [email protected] Rishi Dean Sloan School of Management Massachusetts Institute of Technology, USA [email protected] ABSTRACT years now, we are still at the beginning of using RS on a Recommender systems (RS) are being used by an increasing large scale. In reality, sellers provide an RS to help improve number of e-commerce sites to help consumers find prod- the (long-term) business relationship. This goal gives rise to ucts to purchase. We define here the features that may char- several desiderata that can be difficult to achieve. The RS acterize an ”intelligent” RS, based on behavioral science, should be flexible, scalable, multifunctional, adaptive, and data mining, and computational intelligence concepts. We able to solve complex search and decision problems. present our conclusions from building the WiseUncle Inc. The RS interface with the customer should be based on the RS, named Rubicon, and give its general description. Rather same consumer psychology knowledge and strategies used in than being an advisor for a particular application, Rubicon is marketing. Behind this ”visible” task, a RS can bring valu- a generic RS, a platform for generating application specific able information to marketers, making them improve their advisors. offer and products (customer profiling, marketing segmenta- tion). For instance, RS can help businesses decide to whom Keywords to send a customized offer or promotion. Recommender systems, electronic commerce, user interface, RS use knowledge to guide consumers through the of- user modeling ten overwhelming task of locating suitable products. This knowledge may come from experts