The importance of being dissimilar in recommendation

In recommendation scenarios, similarity measures play a fundamental role in memory-based nearest neighbors approaches. In fact, they recommend items to a user based on the similarity of either items or users in a neighborhood. In this paper, we argue that similarity between users or items, although it keeps leading importance in computing recommendations, should be paired with a value of dissimilarity (computed not just as the complement of the similarity one). We formally modeled and injected this notion in some of the most used similarity measures and evaluated our approach in a recommendation scenario showing its effectiveness with respect to accuracy and diversity results on three different datasets.