Towards new recommender systems

The rise of Information Age is transforming our world and, in particular our society. One of the pivotal aspects, often underestimated is the accessibility of the information. On the internet, the over-abundance of contents make necessary the adoption filtering and Recommender Systems (RS). These systems are typically designed to show the user what he is looking for or, more frequently, what the user “like”, according to a profilation performed according to several techniques. While these solutions can improve the “user satisfaction” in the short-term, still little is known about the long-term and global effects that these omnipresent systems might have caused in the last decade. This effect can happen to be at several levels: social dynamics, opinion dynamics, information flows. Phenomena like social network segregation, opinion polarization or even fake news spreading might be linked to those systems. In this project, we aim in first place at quantifying the impact of RS at various level. In particular, we want to measure how much the over-exposition of users to his/her own favourite items (typical of actual RS) hinders the discovery of novel items and the personal evolution of the users. The project also aims to quantify another volatile concept: the so-called “Comfort Zone”, i.e. the “area”, in the considered item space (movies, songs, news, people), where a given user is still close enough to his/her preference but can anyway discover novelties. The final aim is to build a novel RS able to be creative in its suggestions without losing accuracy in the user preference forecast. In other words, a RS that will accompany the user in the exploration of a wider world but always in steps tailored ad hoc.