Oriel — AI & technology worth noticing, every day.

Physiognomy Revisited: How Vision-Language Models Encode Sensitive Attributes Through Image Transformation

Paper · https://arxiv.org/abs/2610.35910

arXiv:2610.35910v1 Announce Type: new Abstract: Emergent generative AI technologies, while remarkably beneficial across several tasks and domains, can suffer from unforeseen vulnerabilities and difficult-to-anticipate failure modes. Such failure modes can disparately affect vulnerable communities. This paper investigates two well-known vision language models (VLMs) and inspects how these models encode protected identities while generating images. We show that when VLMs are requested to transform human face images to reflect race identities (e.g., Hispanic, Black, White, or Asian) and sexual orientation (e.g., gay or straight), VLMs often comply with such requests. We demonstrate that compliance (and refusal) is non-deterministic, group-sensitive, and generated images are statistically distinguishable, often accentuating deep-seated stereotypes. We further show that representation biases introduced during image generation can re-emerge in textual reasoning and characterization tasks. Finally, we show that this worrisome finding is not limited to racial identity or sexual orientation, image generation can perpetuate bias around criminality, raising concerning possibilities of AI-assisted profiling.

First seen by Oriel on .

AI summary

The paper examines two vision-language models and finds that image transformations targeting sensitive attributes can encode and reinforce stereotypes, with resulting biases also surfacing in textual reasoning.

  • The models often comply with requests to transform faces to reflect racial identities or sexual orientation, but compliance and refusal are non-deterministic and group-sensitive.
  • Generated images are statistically distinguishable and often accentuate stereotypes.
  • Biases introduced during image generation can reappear in textual reasoning and characterization tasks.
  • The authors also find bias around criminality, raising concerns about AI-assisted profiling.

#vision-language models #image generation #algorithmic bias #ai profiling

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
arXiv AI Papersoai:arXiv.org:2610.35910v1https://arxiv.org/abs/2610.35910 Read discussion