Improving Recommender Systems Using Knowledge Obsolescence as a Predictor of Trust

Abstract

In the current context of the Social Web, trust has emerged as a concept and mechanism to differentiate users of this Social Web and the content they generate. Much effort has been devoted to study trust predictors with the aim to provide some operational use of the concept. We propose in this work a new predictor for trust: knowledge obsolescence. We provide a characterization of the concept and a description of the relation between trust and knowledge obsolescence. We applied the concept to a generic recommender system. For this purpose, we have developed a software simulator that allow us to test trust and knowledge obsolescence networks in the recommender systems context. Interesting results were obtained. We found that recommender systems success is augmented. Moreover, we found an improvement in some cases for the coverage of potential recommendable items. We did not find statistical significant benefit on the quality of recommendations.