John Riedl


John Riedl
Associate Professor
University of Minnesota
Department of Computer Science


4-192 EE/CS Building
200 Union Street SE
Minneapolis, MN 55455

Phone: (612) 624-7372
Fax: (612) 625-0572

Recommender Systems Research

Since 1992 I have been working on recommender systems, which I believe have the potential to change the way people interact with all types of information, including newspapers, music, books, and Web sites. I co-founded the GroupLens Research group with Paul Resnick in 1992. In 1995 I invited Joe Konstan to join the group as co-director because his expertise in user interfaces opened tremendous opportunities. In 2002 Joe and I invited Loren Terveen, an expert in diverse types of recommender systems, to join our team. The group has been active in research on all aspects of recommender systems, including algorithms, interfaces, and social effects.

Joe Konstan and I co-authored Word of Mouse, a book on the business applications of recommender systems, and hold several patents on various aspects of recommender systems.

  • Algorithms: We have studied many aspects of recommender algorithms, including the orginal collaborative filtering algorithms mentioned under applications below, a careful study of the parameters that affect collaborative filtering algorithms, and a new item-based recommender algorithm. Nearly all of our user interface and applications research involves developing novel algorithms and testing their effectiveness. We have also studied the effects of agents in algorithms. Agents complement recommender systems by providing a tireless source of rapid and reliable opinions. Recommender systems complement agents by choosing the agents that are the best recommender for each user.
  • User Interfaces: We have studied many aspects of the user interface effects of recommenders, including experiments on user preferences for different types of explanations of the recommendations , experiments on the best ways for recommender systems to get to know new users, and experiments on the effect on the user experience of having the recommender manipulate the predictions maliciously.
  • Applications: our application research typically requires new algorithms, new interfaces, and research on what happens when users use the new algorithms and interfaces in practice.

On to e-commerce research