Can psychographics mitigate over-specialization in recommender-driven consumer markets? Evidence from recommender systems based simulation experiment

Abstract Product recommendation systems exhibit popularity and over-concentration bias, which cause over-specialization, suppress market diversity, and weaken the long-tail effect. Psychographics segmentation, especially users’ personality and value traits, offer a robust theoretical basis for understanding and distinguishing user preferences. The study proposes a diversity-aware personalized recommendation model that integrates psychographic traits with users’ diversity preferences. The model leverages large-scale user review data, computational psychology, and sentiment analysis techniques. Through simulation experiments, we evaluate its impact on recommendation performance and the alleviation of over-specialization. The findings show that diversity-aware re-ranking substantially reduces market concentration and enhances the diversity of recommendation lists. Incorporating psychographic traits further improves personalization, decreases intra-list similarity, and effectively mitigates the filter bubble effect. Comparative analyses of psychographic models reveal that both the Five-Factor Model (FFM) and the Schwartz Value Survey (SVS) achieve optimal performance only when combined with user ratings and diversity preferences. While FFM more effectively reduces market concentration, SVS better captures individual preference orientations. Overall, this study contributes to the development of human-centered recommendation systems and supports the sustainable evolution of algorithm-driven product markets.