When Diversity Is Needed... But Not Expected!

Sylvain Castagnos (1) , Armelle Brun (1) , Anne Boyer (1)

1 KIWI - Knowledge Information and Web Intelligence

Recent studies have highlighted the correlation between users' satisfaction and diversity within recommenders, especially the fact that diversity increases users' confidence when choosing an item. Understanding the reasons of this positive impact on recommenders is now becoming crucial. Based on this assumption, we designed a user study that focuses on the utility of this new dimension, as well as its perceived qualities. This study has been conducted on 250 users and it compared 5 recommendation approaches, based on collaborative filtering, content-based filtering and popularity, along with various degrees of diversity. Results show that, when recommendations are made explicit, diversity may reduce users' acceptance rate. However, it helps increasing users' satisfaction. Moreover, this study highlights the need to build users' preference models that are diverse enough, so as to generate good recommendations.

Mots clés: Recommender Systems; Diversity; User Modeling; User Study

Domaines: Intelligence artificielle [cs.AI]; Informatique

Sylvain Castagnos, Armelle Brun, Anne Boyer. When Diversity Is Needed... But Not Expected!. International Conference on Advances in Information Mining and Management, Nov 2013, Lisbon, Portugal. pp.44-50. ⟨hal-00931805⟩