How Algorithmic Bias Shapes Consumer Acceptance of AI Recommendations: The Roles of Recommendation Strategy, Self‐Identity Threat, Perceived Warmth and Privacy Concerns

ABSTRACT Algorithmic bias, as a socio‐technical phenomenon, may produce discriminatory outcomes that violate norms of justice and equality, thereby adversely affecting specific groups in society. Although algorithmic fairness has received widespread attention, how different recommendation strategies interact with algorithmic bias to influence consumers' psychological responses in intelligent recommendation scenarios remains underexplored. Grounded in identity process theory, this study focuses on two recommendation strategies (relevance‐based vs. similarity‐based recommendations) to examine the impact of algorithmic bias on consumers' recommendation acceptance intention, and further analyzes the moderating roles of perceived warmth and privacy concerns. Through four experimental studies (total N = 895), we find that algorithmic bias reduces consumers' acceptance intention, and this effect is more pronounced under similarity‐based (vs. relevance‐based) recommendation strategies. Perceived warmth and privacy concerns moderate the mediation pathway of self‐identity threat, such that relevance‐based recommendations (vs. similarity‐based recommendations) enhance recommendation acceptance intention by reducing self‐identity threat. These findings reveal that similarity‐based recommendations may exacerbate user resistance in the context of algorithmic bias. This study deepens the theoretical understanding of the psychological mechanisms underlying algorithmic unfairness in the digital age, extends the application of identity process theory to the domain of human‐computer interaction, and provides actionable empirical implications for enterprises in the ethical design and strategic optimization of intelligent recommendation systems.