Impact of Recommender Systems on the Dynamics of Users' Choices

The major focus of recommender systems (RSs) research has so far been on improving the precision and quality of the recommendations. However, it is also important to understand whether the recommended items are actually chosen and how they influence users' choice making. Few studies have attempted to analyse the impact of RSs on users' choice making. In this PhD research, we aim at better understanding the impact of RSs on the evolution of the choices made by a collection of users. We propose simulation procedures where users are simulated to make choices over a period of time when they are exposed to alternative RSs. We measure several properties of the users' choices distribution, which capture the RS effect. Our goal is to understand the evolution of the choices of a collection of users as time goes; next choices are influenced by previous choices used by the RS to generate recommendations. Additionally, we propose online experiments to study the effect of RSs on real users' choices. We propose to design web-based platforms where alternative RSs recommend items to the users and study RSs impact by analysing the evolution of the choices.