The promise of the information age is the delivery of more information than has ever been available before. Unfortunately, human ability to manage and process information remains mostly unchanged. The result of this mismatch is information overload. This project extends one approach to alleviating information overload--collaborative filtering--to provide a new way to assist individuals in personal information selection. Collaborative filtering uses the opinions of member of a community about information objects to recommend a subset of those objects to each individual in the community. The technique works well for small communities in which each user has evaluated many of the information objects. This project uses sophisticated statistical and computation techniques to extend collaborative filtering to communities with millions of users who may have evaluated only a fraction of a percent of the available items. As part of this work, the researchers are developing measures of confidence to help users determine how much faith to place in a particular recommendation. This research project will allow an already booming personalization industry to move from niche markets into mainstream communities and markets; at the same time, it will provide the measures needed for other researchers to evaluate new filtering algorithms and techniques. www.cs.umn.edu/Research/GroupLens
