Collaborative filtering or content-based recommendation? The effects of digital platform recommendation type on consumer's intention

In the context of digital consumption, consumers often face decision-making difficulties when faced with a massive selection of products on e-commerce platforms, and recommendation systems (collaborative filtering vs. content-based) have become a key tool to alleviate information overload. However, there is still a lack of systematic exploration on how different types of recommendations affect consumer purchase intention and their underlying mechanisms. To fill this gap, this study aims to reveal the psychological mechanisms and applicable boundaries of collaborative filtering recommendation and content-based recommendation in influencing purchase intention. Based on social identity theory, this article constructs a dual mediation moderation model to explain how recommendation systems influence consumer decision-making by activating different levels of self-identity. Through experimental research, it was found that collaborative filtering recommendation indirectly promotes purchase intention by enhancing social self-identity, while content-based recommendation indirectly promotes purchase intention by enhancing individual self-identity. Further research reveals that product type (hedonic vs. utilitarian) plays a regulatory role in this process: for hedonic products, their social display value highly aligns with group identity needs, leading to a more significant promotion of social self-identity through collaborative filtering recommendations; for utilitarian products, their practical attributes rely more on individual rational evaluation, which leads to a stronger drive for individual self-identity based on content-based recommendations. This study theoretically reveals the self-identification mechanism of the influence of recommendation systems on consumer decisions, and provides a matching framework based on product characteristics for algorithm selection in practice, thereby optimizing the conversion efficiency of recommendation systems.