Seeing What is Representative.

The article focuses on documenting and experimentally validating a cognitive bias termed "representative signal distortion" (RSD), which occurs in statistical discrimination settings where decision makers evaluate individuals from contrasting groups. RSD causes people to misinterpret new evidence about an individual as more representative of that individual's group relative to a reference group than it truly is, leading to systematic overestimation for high-mean groups and underestimation for low-mean groups. Through a controlled experiment involving abstract groups and perceptual signals, the authors find strong evidence of RSD, distinct from classical biases like base-rate neglect or confirmation bias, and show that it amplifies discriminatory gaps inefficiently. Importantly, they demonstrate that simple interventions—such as withholding group identity until after evidence evaluation or specializing evaluators to assess only one group—can eliminate RSD, reducing discrimination without sacrificing accuracy. The findings suggest that RSD is a robust perceptual bias relevant to real-world discrimination contexts where group contrasts are salient and information is ambiguous.