Toward personalized AI-powered recommender systems to support users’ daily music choice experiences

Abstract

This paper explores a human-centered approach to developing personalized recommender systems by incorporating users’ individual characteristics. Despite the influence of interactive recommender systems on a wide range of everyday decisions, there has been limited research in the field of HCI on how to tailor these interactive computing systems to diverse user decision-making experiences. Existing literature primarily focuses on optimizing recommendation algorithms and lacks insights into developing human-centered recommender systems. To address this knowledge gap, this paper investigates (1) how users’ decision-making tendencies affect their interactions with AI-powered recommender systems, and (2) how these systems can be designed to align with varying user preferences. We developed a prototype of a conversational music recommender system with four different interaction modes by considering users’ different decision-making styles and information processing preferences. An in-lab experiment with 62 participants tested the effects of users’ decision-making styles and the prototype’s interaction modes on their music choice experience. The findings demonstrate that personalizing recommender systems based on users’ different decision-making styles and preferred information processing approaches can enhance user experiences. The paper concludes by discussing implications for future research on improving personalized recommender systems.