Invited talk: Relevance Displays for Recommendations and Search Results

Guy Shani

Information Systems Engineering, Ben Gurion University

Abstract:

Recommendation systems can now be found everywhere in commercial applications. Many researchers have stated that in such applications, the method for displaying recommendations is as important as the actual relevance of the recommended items. Thus, it is surprising that while recommendation algorithms have been studied thoroughly, the recommendation system user interface have been much less investigated.

In this work we look at one specific question related to such user interfaces - can users benefit from the display of the relevance of each recommended item. We experiment with a number of different displays, in order to find a simple, intuitive, yet sufficiently powerful display. We then investigate whether the presentation of such displays influences user behavior when interacting with the system. We continue to generalize our findings to the display of search results relevance. Our study clearly shows that users notice and appreciate the display of confidence and that this simple addition to commercial recommendation engines may have a positive effect on user behavior.

With this example in mind, we hope that our study will draw additional attention to this unjustly under-investigated line of research - improved user interfaces for recommendation systems - and we hope to see more papers on these issues in our research community.