For the GroupLens Collaborative Filtering Recommender Systems, which showed how to automate the process by which a distributed set of users could receive personalized recommendations by sharing ratings, leading to both commercial products and extensive research. Recommender systems have become ubiquitous. Whether shopping at Amazon.com, selecting videos at Netflix, or getting news and information, we have become accustomed to sophisticated personalization software that uses our preference data, together with the preference data of others, to provide personalized recommendations for content or products. Automated collaborative filtering, which launched the field of recommender systems in 1994, was introduced, refined, and commercialized by the team of John Riedl, Paul Resnick, Joseph A. Konstan, Neophytos Iacovou, Peter Bergstrom, Mitesh Suchak, Brad Miller, David Maltz, Jon Herlocker, Lee Gordon, Sean McNee, and Shyong (Tony) K. Lam through the GroupLens family of systems. The first generation GroupLens system (1994) was used to create personalized recommendations of Usenet news articles. The second generation (1995-1996) also targeted Usenet news articles, and led to the commercial launch of Net Perceptions, spawning what is now a vibrant recommender systems industry. After commercialization, the team launched MovieLens in 1997. MovieLens, a web-based research platform still operating today, has been used by more than 100,000 users; its online experiments and published datasets have led to hundreds of published advances in the field of recommender systems. Prior to GroupLens, most personalization systems were based entirely on building profiles of content preferences, and then applying them to new content. While this technique worked when interests were topical (e.g., choosing news about a favorite football club), they did not handle less feature-based preferences or distinguish among topical items by quality or taste. In 1992, Xerox PARC's Tapestry project introduced the term and concept "collaborative filtering", creating a centralized database where annotations could be stored and retrieved, but application of this idea was limited by the need to manually form queries and explore the database. The GroupLens team introduced automation to the process, which proved to be the fundamental breakthrough that enabled wide-ranging research and commercial applications. Through the founding of Net Perceptions, GroupLens had an enormous impact on e-commerce and information portals. Early customers, including Amazon.com, CDnow, and Art.com, demonstrated that the software generalized to a wide variety of domains, and validated the usefulness of collaborative filtering recommendation through explicit and implicit user ratings . Net Perceptions grew to become a $1 billion company, providing recommender systems to leading retail and information companies around the world. Recommender systems grew into an extensive research field, drawing from machine learning, human computer interaction, ecommerce, information retrieval, databases, and other areas of computer science. ACM's annual Recommender Systems Conference brings together leading researchers and practitioners. The field was recently energized by the $1 million Netflix Challenge.
For contributions to recommender systems and to social and collaborative computing.
Sign in to view more
Sign in to view more
My research focus is on collaborative systems that support human interaction through computer systems. My career goal is to understand how to develop and apply computer technology to the problems of human organizations. One of the biggest such problems is getting the right information to the right people. The Internet has democratized the publishing process. Now, anyone who wants can publish anything they want, just by creating a Web site. We humans are hopelessly overmatched by the increasing volumes of information that are published. Collaborative filtering is a technology that enables us to all work together to sift through the millions of documents on any topic to find those that are most appropriate for each of us. Collaborative filtering works by learning which kinds of documents each of us likes, and finding other people who share out interests. Across our entire research program, our goal is to understand how computers can be used to help people process information more efficiently, and work together better. I am currently involved in several research projects to explore these topics.