Cognitive Biases in Recommender Systems

Abstract

Psychology and sociology research has revealed a plethora of cognitive biases that describe systematic deviations from rationality and objectivity in humans’ cognitive processes and judgment or collective prejudices of a society that favor one group’s values, norms, and traditions over others. Examples such as confirmation bias, conformity bias, or cultural homophily are particularly relevant in the context of recommender systems. These biases are reflected in user-generated data (such as ratings or tags) and, as a consequence, are embedded in user models that drive recommendation algorithms. As a result, recommender systems may reproduce or amplify such biases, leading to undesired outcomes such as stereotypical recommendations, reduced diversity, or unfair exposure. At the same time, cognitive biases shape how users perceive, interact with, and provide feedback to recommender systems, thereby influencing feedback loops and long-term system dynamics. In this multidisciplinary tutorial, we bridge insights from psychology and computer science to examine cognitive biases that are particularly relevant in the context of recommender systems. We introduce these biases grounded in empirical findings and discuss how they manifest across the recommender system pipeline. Participants are engaged through interactive exercises that allow them to experience the effects of cognitive biases firsthand. In a hands-on part, participants are able to (1) design psychology-informed user choice models based on the presented cognitive biases, (2) integrate them in a provided feedback loop simulation setup with established recommendation algorithms, and (3) observe how different biases influence the evolution of user profiles, recommendations, and system performance over time.