People project a specific racial identity onto artificial intelligence, study finds

New research published in the Journal of Personality and Social Psychology reveals that people tend to view artificial intelligence systems as being White, even when those systems lack any visible racial indicators such as a face, voice, name, or body. This racial association has been linked to a higher level of trust in AI recommendations, with systems trained on data from Black and Latino adults assigned less intellectually demanding tasks than identical systems that were trained on data from White or unspecified sources.

The findings suggest that human users project familiar social hierarchies onto emerging technologies, which may influence the level of authority granted to automated tools.

Researchers from Purdue University and Northwestern University conducted a series of studies to explore this phenomenon. They suggest that anthropomorphism, the tendency to attribute human-like qualities to non-human objects, plays a significant role in this racialization of AI. In Western cultures, Whiteness often serves as an unstated default identity and is associated with high status, technical expertise, and intelligence, thereby transferring these stereotypes onto smart technologies.

Additionally, people form impressions about AI based on the broader ecosystem surrounding it, including the programmers, users, and training data, which may contribute to racial categorizations. These perceptions could impact how individuals evaluate AI advice in various settings, such as workplaces and daily life. The study emphasizes the importance of understanding what drives confidence in automated advice, as organizations increasingly adopt AI tools.