Using a Personality-Aware Recommendation System for Comparing Inventory Performances

This study delves into personality-aware recommendation systems, comparing two specific personality inventories to determine which one offers better input for more accurate recommendations. To that end, we utilize two personality inventories, NEO-FFI and TIPI-J, derived from the Big Five trait model. In a series of experiments, 17 participants completed these inventories and ranked 10 movies. Based on a leave-one-out scheme, we developed a straightforward personality-aware recommendation system to suggest movies and subsequently tested its effectiveness. Our recommendation system operates on the assumption that each personality trait exerts a distinct influence on movie tastes represented in terms of weights. By generating 64 recommendation configurations, we optimized these weights and measured the disparity between the resultant movie rankings and the actual ones. Additionally, we explored two alternative recommendation approaches and investigated which configuration outperforms each alternative the most. Interestingly, the scheme deploying NEO-FFI outperformed the alternatives more frequently than TIPI-J. Finally, we evaluated the competence of the personality-aware recommendation system by comparing the outcomes of the best configuration with the gold standard (i.e. most aligned ranking in the training set). The results revealed that the recommender exhibits a smaller disparity from the ground truth than the gold standard, confirming its competence.