AI-Generated Faces Are Perceived as More Trustworthy Than Real People, Study Finds

People are no better than chance at distinguishing between real human faces and those generated by artificial intelligence, and surprisingly, they tend to find AI-generated faces more trustworthy, according to a study from Lancaster University. This finding raises concerns about the potential for AI-powered identity fraud and disinformation campaigns.

Researchers found that participants could only identify AI-generated faces correctly about 58.4% of the time, a success rate only slightly better than random guessing. The study, published in the Journal of Vision, involved 169 participants assessing a mix of real and AI-generated faces. While early AI-generated images often contained noticeable flaws, such as extra fingers or misaligned features, the latest generation of AI models produces images that are increasingly difficult for humans to detect.

"The fact that people often perceive AI-generated faces as trustworthy makes them particularly powerful tools for online scams and disinformation," said lead author Alexis McGuire, a PhD student at Lancaster University. "For example, a text-based scam may become more convincing if it is accompanied by a face that people instinctively trust."

In a follow-up test, participants were asked to rate the trustworthiness of the faces on a scale of one to seven. Real human faces scored an average of 4.04, while AI-generated faces from Generative Adversarial Network (GAN) models scored 4.36. Faces created by newer 'diffusion models' received the highest trust rating at 4.7, even though participants found them less realistic than GAN-generated faces.

The researchers suggest that AI-generated faces may appear more trustworthy because they often represent an 'average' human face, clustering towards a common representation that humans find familiar. "AIs tend to create 'polished, idealised faces that appear especially attractive', which people tend to instinctively find appealing," McGuire explained. "They have features that people naturally associate with trust, such as being more attractive. Research has long shown that people often perceive attractive individuals as more trustworthy."

This tendency for AI-generated faces to be perceived as more trustworthy, despite sometimes being less realistic, highlights a potential psychological paradox. The findings suggest that the perceived trustworthiness of AI faces might stem from different cognitive mechanisms than the judgment of realism, possibly linked to an inherent bias towards attractiveness and idealized features. The ease with which AI can now generate these convincing and appealing faces presents a significant challenge for online security and the potential for manipulation.