Sushma Channamsetty and Michael D. Ekstrand. 2017. “Recommender Response to Diversity and Popularity Bias in User Profiles”. In Proceedings of the 30th International Florida Artificial Intelligence Research Society Conference (FLAIRS 2017).
Recommender system evaluation usually focuses on the over-all effectiveness of the algorithms, either in terms of measurable accuracy or ability to deliver user satisfaction or improve business metrics. When additional factors are considered, such as the diversity or novelty of the recommendations, the focus typically remains on the algorithm’s overall performance. We examine the relationship of the recommender’s output characteristics — accuracy, popularity (as an inverse of novelty), and diversity — to characteristics of the user’s rating profile. The aims of this analysis are twofold: (1) to probe the conditions under which common algorithms produce more or less diverse or popular recommendations, and (2) to deter-mine if these personalized recommender algorithms reflect a user’s preference for diversity or novelty. We trained recommenders on the MovieLens data and looked for correlation between the user profile and the recommender’s output for both diversity and popularity bias using different metrics. We find that the diversity and popularity of books in users’ pro-files has little impact on the recommendations they receive.