How do disparities reproduce themselves? "Ground truth'' inference from utility-maximizing agent's sampling behavior
- Yuan Meng, Psychology, University of California, Berkeley, Berkeley, California, United States
- Fei Xu, Psychology, UC Berkeley, Berkeley, California, United States
AbstractIn an ideal world, evidence of disparities motivates people to fight them; in reality, disparities often reproduce themselves. Upon seeing the police stopping some groups at higher rates, people may believe members of these groups are more prone to crime and therefore seek more punitive measures against them. In this paper, we argue that even without stereotypic associations linking some groups with crime, people may still reproduce observed disparities via rational inference: Assuming an agent is knowledgeable about a target trait's "hit rates" in different groups and acts to maximize the expected utility of checking, you may infer that groups checked more often have higher hit rates. In Experiment 1, this "Naïve Utility Calculus" captured how people inferred the hit rate in a population based on how often an agent sampled from it ("check rate"). In Experiment 2, when hit rates in the samples were revealed, people predominantly relied on this new information more heavily than the agent's check rates. Our work both provided a novel explanation for why people reproduce disparities and a potential intervention to combat such a tendency.