Detecting social biases using mental state inference.

Area of Science:

• Cognitive Psychology
• Social Psychology
• Computational Social Science

Background:

• Social biases are common but often subtle, making detection challenging.
• Existing research primarily focuses on bias consequences, not detection mechanisms.
• Understanding how individuals detect bias is crucial for addressing its impact.

Purpose of the Study:

• To propose and validate a cognitive account of social bias detection grounded in mental state representations.
• To develop a computational model explaining how people infer biases from observed behavior.
• To investigate the role of inferring prior beliefs versus updated beliefs in bias detection.

Main Methods:

• Developed a formal computational model of social bias detection based on mental state inference.
• Conducted four preregistered experiments (N=876) to test the model's predictions.
• Compared the proposed model against alternative models of belief updating in biased agents.

Main Results:

• The model successfully predicted participants' inferences about an agent's prior beliefs and general social biases.
• The model accurately captured detection of specific racial and gender biases across diverse contexts.
• Participants' judgments were best explained by inferring prior beliefs before evidence-based updates.

Conclusions:

• Social bias detection relies on inferring an agent's mental state, specifically their prior beliefs.
• Recognizing a gap between expected unbiased behavior and observed actions is key to detecting bias.
• This research enhances understanding of the cognitive mechanisms underlying social bias detection.