VR study reveals racial bias in police language, explores LLM use for analysis

A new study utilizing virtual reality simulations has investigated how police officers' language use is affected by the perceived race and gender of individuals they interact with. The research found that most officers spoke less deferentially to virtual characters depicted as Black adult males, with notable exceptions being White, biracial, and multiracial female officers. This difference in tone could potentially lead to conversation breakdowns and dangerous situations for both officers and the public. The study also explored the use of large language models (LLMs) for estimating these effects, recommending mixed-effects models with an inverse propensity treatment weighted approach for text feature creation. AI

IMPACT Highlights potential biases in AI-driven simulations and explores LLMs for bias detection in social science research.

RANK_REASON The cluster contains a research paper published on arXiv detailing experimental evidence from VR simulations.

Read on arXiv cs.AI →

• Black
• large-language models
• U.S.
• virtual reality
• White
• average treatment effect
• Black adult males
• Black male character
• Female Officers and Community Policing: Examining the Connection between Gender Diversity and Organizational Change
• multiracial people

• paper
• safety
• other

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →