The Psychology of Algorithmic Bias

Algorithmic bias—the systematic discrimination of disadvantaged social groups by AI systems—is widely documented, yet its psychological foundations remain poorly understood. We argue that algorithmic bias originates primarily from human prejudice and social cognition, which shape AI training data and use of AI systems. We introduce the psychology-centered Human–AI Loop model that explains how biases can enter an AI system during its creation and consumption, specifying the psychological mechanisms operating at each entry point. These mechanisms and their interactions create feedback loops through which AI systems amplify, perpetuate, and obscure existing prejudices and social inequalities. We then outline psychology-informed strategies for disrupting these cycles and discuss implications for theories of prejudice and discrimination in the age of AI.