The Department's GroupLens Research Group is branching out in several exciting directions. Perhaps best known for their MovieLens film recommender system (www.movielens.org), the group has been developing and applying recommender technology to information overload problems for more than a decade. Simply put, systems like MovieLens use the opinions of a community of users to help evaluate and recommend items for individuals in that group. Some users may match best with a subcommunity that recommends "The Two Towers" while others receive recommendations for "Maid in Manhattan."

The group is exploring new directions along three fronts. First, they are looking beyond traditional centralized recommender systems. As Professor John Riedl points out: "Most of today's recommender systems are owned and operated by marketers trying to sell things. By exploring handheld recommenders and peer-to-peer recommendation systems, we can provide an alternative that allows the consumer to level the playing field and ensure honest, unbiased recommendations." The group is also exploring the applications that are enabled when recommendations are delivered to mobile devices. The group reported on their exploration of four mobile recommender interfaces at the 2003 ACM Conference on Intelligent User Interfaces in January.

Second, the group is exploring the application of recommendation technologies to a new domain: that of research papers. The "TechLens" project is exploring how computer systems can help students and researchers find papers they don't already know about. They presented a paper on their preliminary results at the 2002 ACM Conference on Computer Supported Cooperative Work in November, and ran a live demonstration where conference attendees could get recommendations based on the papers the attendee had already written. Professor Joseph A. Konstan explains: "Part of our excitement with research papers is the fact that authors already indicate which other papers they feel are important by citing them. Our CSCW paper both shows that mining these citations can provide useful recommendations to others, and opens the door to a variety of different recommendation techniques depending on the goals and experience of the person searching."

Finally, the Group is exploring a variety of topics related to human reactions to recommender systems. In April, they will present a groundbreaking paper at the 2003 ACM Conference on Human Factors in Computing Systems (CHI 2003) that shows that users can be manipulated when recommender systems lie, but that at the same time users become less happy with systems that lie and choose to use them less often. As Professor Loren Terveen puts it: "We are excited to explore a variety of human factors issues, from the question of what motivates people to contribute their opinions to such a system, to how people decide to trust its recommendations, to broader questions of how people participate and organize themselves in on-line communities."

In all, the GroupLens Research Group promises to bring forward a variety of exciting results for years to come. For more information, visit http://www.grouplens.org.

-Joseph A. Konstan