Contextualizing Recommendation Explanations with LLMs: A User Study

Abstract

Large language models (LLMs) are increasingly used in recommender systems to generate natural-language explanations for contextualized recommendations. This study examines how different types of LLM-generated explanations shape user perceptions and behavioral intentions in the context of movie recommendations. In a between-subjects online experiment (N=759) and follow-up interviews (N=30), we compare LLM-generated (a) generic explanations and (b) contextualized explanations. We find that contextualized explanations better address users’ cognitive and affective needs and strengthen intention to watch the recommended movies. Interview findings further show that the effectiveness of contextualization depends on how well explanations align with users’ genre and narrative preferences. References to prior movies can improve user perceptions and intent, but excessive detail may make explanations feel redundant. The benefits of contextualization are pronounced among users with higher engagement and greater trust in the system. Together, these findings highlight the potential of LLMs to support more user-centered recommendation experiences and offer practical guidance for designing explanations that improve trust, relevance, and engagement in entertainment platforms.