Discussion

Study finds people often picture AI as White, with consequences for trust

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#12783

People often assign a racial identity to AI even when it has no face, voice or name, a new study finds. The researchers also link those perceptions to how much people trust AI advice and what work they think AI should do.

Watch Desk analysis

What happened

PsyPost reports on research by Monica Gamez-Djokic and Adam Waytz, published in the Journal of Personality and Social Psychology. Across two US studies, participants on average selected White for 72% of the technologies they were asked to racialise. In an open-response study, 75% assigned a racial identity to the technologies, and 71% of those responses described them as White.

The researchers also tested how people imagined AI without being asked about race. In a task where participants chose which blurry faces looked more like AI, 75% of a separate group later classified the resulting composite as White. Other experiments linked perceiving AI as more White with finding its advice more persuasive. In a separate scenario, identical systems described as trained on data from Black and Latino adults were assigned less intellectually demanding work than systems described as trained on White or unspecified data.

Why it matters

AI does not need a face to arrive with a social identity in its users’ minds. Those assumptions could affect whether people accept its recommendations and the authority or work they are willing to hand it. That is a human response to AI, not evidence that the software itself has a race.

The reported link with persuasiveness was modest, and the findings do not show that racial perceptions cause people to trust AI more. The experiments do, however, make the idea of a socially neutral machine look rather less tidy.

Our read

The useful point is not that every user imagines the same AI. It is that apparently blank technology can still pick up familiar social assumptions. Anyone deploying AI in decisions should pay attention not only to what a system produces, but to how people interpret and defer to it.

What to watch

  • Whether further research finds similar patterns in other settings and populations.
  • How perceived AI identity affects reliance on advice in real-world decisions.
  • Whether organisations account for those perceptions when assigning AI a role.

Discussion spark: If people bring racial assumptions to AI even when it has no obvious identity, should organisations address those perceptions in deployment, or focus only on testing the system’s outputs?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.