Scholarlink

Objectives: <p>This study evaluates whether AI-generated police officer profiles can serve as valid, reproducible visual stimuli for experimental criminology. We test whether systematically varied images differing in race, gender, and body composition convey intended traits and elicit theoretically coherent impressions of competence, agency, and warmth.</p> Methods: <p>Using a <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(4 \times 2 \times 3\)</EquationSource> </InlineEquation> factorial design, we generated 24 AI-based officer profile cells (72 individual profiles) and collected 2,565 evaluations from a nationally representative U.S. sample (n = 513). Participants identified each officer’s race, gender, and body type, detected one-at-a-time manipulations, and provided impression ratings across key social perception dimensions.</p> Results: <p>Respondents accurately recognized all intended attributes and reliably detected manipulations across otherwise identical images. Impression ratings reproduced established patterns, including gender-linked warmth and extended body-type effects on competence and agency. Muscular police officers experience perceptual benefits while heavy officers experience a range of penalties.</p> Conclusions: <p>AI-generated officer images provide a transparent, scalable, and valid framework for visual experiments in policing, criminology, and broader social science.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AI-generated human stimuli for experimental social science

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Chandler G. Robinson,
  • Ian T. Adams,
  • Matthew W. Logan,
  • J. Pete Blair

摘要

Objectives:

This study evaluates whether AI-generated police officer profiles can serve as valid, reproducible visual stimuli for experimental criminology. We test whether systematically varied images differing in race, gender, and body composition convey intended traits and elicit theoretically coherent impressions of competence, agency, and warmth.

Methods:

Using a \(4 \times 2 \times 3\) factorial design, we generated 24 AI-based officer profile cells (72 individual profiles) and collected 2,565 evaluations from a nationally representative U.S. sample (n = 513). Participants identified each officer’s race, gender, and body type, detected one-at-a-time manipulations, and provided impression ratings across key social perception dimensions.

Results:

Respondents accurately recognized all intended attributes and reliably detected manipulations across otherwise identical images. Impression ratings reproduced established patterns, including gender-linked warmth and extended body-type effects on competence and agency. Muscular police officers experience perceptual benefits while heavy officers experience a range of penalties.

Conclusions:

AI-generated officer images provide a transparent, scalable, and valid framework for visual experiments in policing, criminology, and broader social science.