Leveraging Social Information To Improve Recommendation Novelty

Abstract

In today’s digital world, users are often confronted with an abundance of information. Whether the user is looking to compare online prices for products, searching for new movies to watch or music to listen, the available information at hand exceeds the amount of information which the user wants to consider before making a choice. For this, various recommendation systems have been developed. Primarily, the recommendations made by these recommendations systems have been evaluated based on their accuracy. More recent research has begun evaluating the subjective perception of the recommendations and has developed additional attributes such as diversity and novelty to evaluate recommendation systems. However, the impact of such attributes on actual user satisfaction has been explored less. Recent research has seen an increase in evaluating other aspects of recommendation systems such as recommendation interfaces. Nonetheless, very limited work has been done in terms of recommendation system interfaces to improve the perceived quality of the recommendations and overall user satisfaction.

This master thesis introduces and evaluates a novel interface, MovieTweeters. It is a movie recommendation system which incorporates social information with a traditional recommendation algorithm to generate recommendations for users. Few previous studies have investigated the influence of using social information in interactive interfaces to improve the novelty of recommendations. To address this gap, we investigate whether social information can be incorporated effectively into an interactive interface to improve recommendation novelty and user satisfaction. Our initial results suggest that such an interactive interface does indeed helps users discover more novel items. Also, we observed users who perceived that they discovered more novel and diverse items reported increased levels of user satisfaction. Surprisingly, we observed that even though we successfully were able to increase the system diversity of the recommendations, it had a negative correlation with users perception of novelty and diversity of the items highlighting the importance of improved user-centered approaches.