University of Minnesota
http://www.umn.edu/
612-625-5000
Research
stacks of books

Bookmark and Share (cc) ginnerobot on flickr

Recommender Systems Past, Present, and Future

Tracing the evolution of a ubiquitous online feature.

Computer systems that recommend products, articles, or other items have become a common and expected feature of the digital world. For more than two decades, research at the University of Minnesota has informed and advanced the science of recommender systems, driving commerce on sites like Amazon and Netflix, and content consumption across the Internet.

The Study

Amazon logoThe early 1990’s saw the introduction of automated recommender systems. As access to the Internet spread, recommender systems became a central feature of online commerce and other sites. Today, web users expect the sites they visit to provide personalized recommendations that help them discover new products and content that match their tastes and interests.

Netflix logoIn a paper that traces the evolution of recommender systems, University of Minnesota professors Joseph Konstan and John Riedl review twenty years of research on the topic and highlight major contributions that have changed our world. Included in the discussion:

  • How the systems expanded from collaborative filtering to encompass diverse approaches
  • Early algorithms and subsequent efforts to improve prediction accuracy
  • Commercial applications of recommender systems and the resulting impact on system design
  • Changing notions of recommendation quality
  • Privacy concerns and social effects of recommenders

The authors conclude by charting a course for future research, predicting that the next generation of recommendation engines will integrate more user-contributed content as well as contextual approaches.

Impact

The University’s decades-long history of work in social computing includes some of the foundational research on recommendations systems. Building on that work, researchers have advanced the field in various ways, developing new methods, improving accuracy and quality, and inspiring new applications. Academic work in recommendations systems has fueled the tremendous expansion of online economic activity, and will continue to shape the way people discover new products and content in the future.


Recommender systems: from algorithms to user experience, Konstan, J. A., and Riedl J. T. , User Modeling and User-Adapted Interaction, Volume 22, p.101-123, (2012)