Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

The Effect of Perceived Race and Gender on Police Language Use: Experimental Evidence from VR Simulations

Source

arXiv

arxiv.org/abs/2608.05050v1

Read the source

Source

Plain-Language Brief

Researchers studied how police officers speak to virtual characters depicted as Black adult males in VR simulations. The study found that most officers speak less deferentially to Black male characters, except for White, biracial, and multiracial female officers, especially when the character is known to be a suspect. This difference can lead to conversation breakdowns potentially resulting in violence or danger. The study also explored the use of large language models (LLMs) for analyzing these conversations and making estimates about the effect sizes.

What the Document Says

  • The research explores police officers' speech towards virtual Black male characters in VR simulations.
  • Most officers speak less deferentially to Black male characters, with exceptions for White, biracial, and multiracial female officers when the character is a suspect.
  • Across a conversation, this can lead to notable changes in tone that may contribute to potential violence or danger.
  • The study also examines the capabilities of large language models (LLMs) for analyzing the data and making estimates about effect sizes.
  • The authors recommend mixed effects models with inverse propensity treatment weighted (iptw) approach and LLM for text feature creation for ATE estimation.
  • Further development and refinement are needed for using LLMs to finetune prediction models for ATE estimation.

Key Concepts Mentioned

  • police officers
  • virtual reality simulations
  • perceived race
  • gender
  • deferential speech
  • average treatment effect (ATE)
  • large language models (LLMs)

Claims Made by the Document

  • Most police officers speak less deferentially to Black male characters, except for certain female officers.
  • This difference in tone can lead to potential violence or danger.
  • The study explores the capabilities of LLMs for ATE estimation.

Evidence Used

  • results from VR simulations showing differences in officer speech based on race and gender

Methods / Approach

  • causal inference lens
  • mixed effects models with inverse propensity treatment weighted (iptw) approach
  • using LLM for text feature creation
  • model validation against synthetic data

Limitations Stated by the Authors

  • LLMs require further development and refinement for finetuning prediction models for ATE estimation

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The study is based on virtual simulations, so the findings may not fully represent real-world interactions between police officers and citizens.
  • The study only examined one aspect of police interactions, deferential speech, and did not consider other factors that could impact interactions.

Open Questions

  • What are the long-term effects of these differences in tone on police interactions and outcomes?
  • How can LLMs be further developed and refined to improve their accuracy in ATE estimation?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The study focuses on how deferential police officers speak to virtual characters depicted as Black adult males, not on the reasons behind this behavior.
  • The study uses a scale of 0-10 for deference of tone.

Raw Abstract / Excerpt

Against the backdrop of violence in police interactions with the U.S. public, we explore how deferentially police officers speak to virtual characters depicted as Black adult males in vir- tual reality (VR) simulations. We evaluate the effect of seeing and communicating with these characters through a causal in- ference lens, where the assignment of the Black man character to a police officer and simulation is the treatment variable. Our (marginal) average treatment effect AT E measures the social impact of the character on the deference of officer statements with each turn of the conversation. Soberingly, we find that most officers speak less deferentially to Black man characters, except for White, biracial, and multiracial female officers, es- pecially in settings where the VR character was known to be a suspect. Across a full conversation of a typical VR scene, these marginal AT Es can result in notable changes in def- erence of tone (two to several points difference on a scale of 0-10), above and beyond that due to the initial effect of per- ceiving a Black male character. Even more disconcerting is that this can contribute to conversation breakdowns that po- tentially result in violence or danger to both the public and the police. We also explored the capabilities of large language models (LLMs) for ATE estimation. From our methods com- parison analysis, including model validation against synthetic data, we provide unique scientific insights on LLM-assisted methodologies for ATE estimation. As such, for ATE esti- mation with multilevel data with text, we recommend mixed effects models with the inverse propensity treatment weighted (iptw) approach, which utilized an LLM for text feature cre- ation. While we also tested LLMs for finetuning prediction models ultimately for ATE estimation, we conclude they are an area for further development and refinement.