User Trust in Online Learning Portals: Field Experiment of Explainable Recommender Systems

Abstract

The rapid proliferation of digital education platforms has fundamentally transformed the landscape of modern learning paradigms. As online learning portals accumulate vast repositories of educational content, users increasingly confront severe information overload, necessitating the deployment of automated recommender systems. However, traditional algorithmic models operate as opaque mechanisms, severely limiting user comprehension and subsequently degrading user trust. This paper presents a comprehensive field experiment designed to evaluate the impact of explainable recommender systems on user trust within a large-scale online learning portal. Over a longitudinal period, participants were randomly assigned to either a baseline recommendation environment or a treatment environment featuring multifaceted algorithmic explanations. The study meticulously captures both self-reported psychometric data and granular behavioral interactions. The empirical evidence demonstrates a substantial, statistically significant improvement in multidimensional user trust when transparent, explainable recommendations are provided. Furthermore, the findings reveal that the provision of explanations not only enhances cognitive and affective trust but also significantly increases behavioral engagement, manifested through higher course enrollment rates and prolonged system interaction. This research contributes foundational empirical data to the intersection of artificial intelligence and human-computer interaction, offering critical insights for the architectural design of future educational technologies. The results unequivocally advocate for the integration of transparent explainability mechanisms as a primary design requirement in educational recommender systems.