Accessible Requires Authentication Published by De Gruyter Oldenbourg 2019

5. Explanations and user control in recommender systems

Dietmar Jannach, Michael Jugovac and Ingrid Nunes

Abstract

Adaptive, personalized recommendations have become a common feature of today’s web and mobile app user interfaces. In most of modern applications, however, the underlying recommender systems are black boxes for the users, and no detailed information is provided about why certain items were selected for recommendation. Users also often have very limited means to influence (e. g., correct) the provided suggestions and to apply information filters. This can potentially lead to a limited acceptance of the recommendation system. In this chapter, we review explanations and feedback mechanisms as a means of building trustworthy recommender and advice giving systems that put their users in control of the personalization process, and we outline existing challenges in the area.

© 2019 Walter de Gruyter GmbH, Berlin/Munich/Boston