Tackling the RecSys Side Effects via Deep Learning Approaches

Erica Coppolillo

Abstract: Digital platforms, such as social media and e-commerce websites, widely make use of Recommender Systems to provide value to users. However, social consequences of such uses and their long-term effects are still unclear and controversial, spanning from influencing users’ preferences to perpetrating bias embedded in the data. In this work, we present our current research activities on these phenomena, based in particular on a proper use of ML/DL techniques: we overview the state of the art and then highlight our contribution, focusing on the study of opinions drift and unfairness in recommendations.