People Project Racial Identity onto Artificial Intelligence Systems
Research reveals humans subconsciously associate AI with Whiteness, impacting trust and task delegation.
1Subconscious Racial Projection
New research published in the Journal of Personality and Social Psychology reveals that people frequently perceive artificial intelligence as White. Even when software lacks physical features like faces or voices, human users project racial identities onto these systems. This psychological process suggests that technology is not viewed as a neutral entity.
2The Role of Anthropomorphism
The study, led by Monica Gamez-Djokic and Adam Waytz, builds on the concept of anthropomorphism. Humans naturally attribute personalities and social traits to non-human objects. Once a computer program is treated as a social partner, the brain automatically applies mental shortcuts used to categorize human beings, including race, age, and gender.
3Experimental Evidence of Bias
In studies involving hundreds of U.S. adults, participants consistently assigned a White identity to various technologies. When asked to choose from a list or provide open-ended responses, approximately 71% to 75% of participants identified AI as White. This pattern remained consistent regardless of the participant's own racial or ethnic background.
4Visualizing the AI Face
Researchers used a reverse-correlation task to reconstruct the mental image of AI. Participants selected blurry images that looked more like artificial intelligence, which were then averaged into a composite face. A separate group identified this composite as White 75% of the time, confirming an implicit, unspoken mental association with Whiteness.
5Global Cultural Perspectives
The study expanded to India and Japan to test if this was an American phenomenon. In Japan, 82% of participants classified AI as White. In India, while the broad category of AI was frequently labeled as White, specific device ratings were more varied, suggesting that the term AI triggers stronger stereotypes.
6Impact on Perceived Persuasiveness
Perceiving AI as White is linked to how persuasive its advice seems to users. In scenarios involving political fact-checking and consumer recommendations, participants who viewed AI as White rated its output as more credible. This effect persisted even when controlling for political ideology and general attitudes toward new technology tools.
7Training Data and Status
The research suggests that Whiteness is often used as a default identity associated with high status and technical expertise. People form impressions based on the broader ecosystem, including who they believe programs the software and whose knowledge is fed into training data, leading to the assumption that AI reflects White sources.
8Experimental Manipulation of Data
Study 5 tested the causal link by manipulating the described racial makeup of training data. Participants were told an AI was trained on data from either White adults or Black and Latino adults. The results showed that participants assigned less intellectually demanding work to the AI trained on Black and Latino data.
9Task Delegation Disparities
When an AI was described as trained on Black and Latino data, participants assigned it tasks with an average intelligence score of 4.27. In contrast, an identical AI trained on White data received tasks with an average score of 4.51. This indicates that racial perceptions directly influence how humans delegate complex responsibilities.
10The Default Identity Assumption
Notably, the AI trained on White data and the control AI with unspecified training data were treated almost identically. This suggests that when no racial information is provided, human users default to assuming that an AI system reflects predominantly White knowledge, reinforcing existing social hierarchies within the digital landscape of modern workplaces.
11Implications for Future Adoption
As organizations increasingly rely on automated decision-support systems, these findings highlight potential risks. A 2026 survey indicated that higher trust in AI is associated with a greater intention to rely on it. If users subconsciously devalue AI based on perceived racial training data, it could lead to biased decision-making in professional environments.
12Conclusion on Social Bias
The study concludes that AI does not escape social biases simply because it is a technological tool. Humans bring their own assumptions to interactions with software, often without awareness. Understanding these psychological mechanisms is essential as society integrates more automated tools into daily life, hiring processes, and critical civic decision-making roles.
Original article: PsyPost — People Project Racial Identity onto Artificial Intelligence Systems