Understanding Explainability in Recommender Systems-User Insights and Perspectives

This paper presents our preliminary findings from a systematic literature review on explainability in recommender systems from a user-centered perspective. Despite extensive literature on explainability in Artificial Intelligence (XAI), this study focuses specifically on how explain ability in recommender systems affects user trust, taking into account insights drawn from design and Human-Computer Interaction (HCI). To this end, we extracted 387 journal and conference papers from ACM, IEEE, Taylor and Francis, Science Direct, and Springer. After applying inclusion and exclusion criteria, 10 relevant articles published between 2018 and 2024 were selected for this analysis. According to the results, users value justifications for recommendations to better understand why certain products or services are suggested. Moreover, scrutability is crucial for enabling users to provide feedback when recommendations do not align with their preferences. Explanations should be informative and easily understandable to enhance transparency and decision-making efficiency. The final component of fostering user trust, satisfaction, and transparency is to provide adaptive explanations, allowing users to control the level of detail based on their mental models and personal characteristics.