I-Corps: Diversity-aware News Recommendation

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

This project plans to investigate building a diversity-aware news recommendation service to improve people's news reading experience. Researchers have launched several controlled experiments and field trials to validate theories and gain insights on how to measure alternative notions of diversity, how to model users' diversity preferences, and how to present challenging items to users in a way that is palatable to users. In addition, we have developed algorithms, systems and widgets that classify political articles according to their political opinions, produce a balanced list of articles, and present the diversity-enhanced recommendations to users while they are reading related articles. This work is the first attempt to formalize several different instantiations of the general concept of diversity and to devise algorithms that optimize for these measures. Digital news aggregators rely on ratings and links to select and present subsets of the large quantity of news and opinion items generated each day. Opinion and topic diversity in the output sets can provide several benefits. For individual readers, diverse results may be more interesting and lead to greater learning over time. A particular form of diversity, proportional representation of different group interests, can provide common ground for discussion with other people and also lead to understanding of which are majority and which are minority viewpoints. The service proposed by this project, if successful, will be made available directly to consumers, as a website and browser plug-ins. Researchers believe that they can provide a service that will help the public enhance their news-reading habits. Diversity-aware recommendation techniques are likely to be applicable to other domains beyond news reading where selecting a diverse set of items is valuable, such as search engine results, e-Commerce product recommendations, and audience voting on questions to ask of a conference speaker or public official.