AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones

Abstract

Recent evidence shows that AI-generated faces are now indistinguishable from human faces. However, algorithms are trained disproportionately on White faces, and thus White AI faces may appear especially realistic. In Experiment 1 ( N = 124 adults), alongside our reanalysis of previously published data, we showed that White AI faces are judged as human more often than actual human faces—a phenomenon we term AI hyperrealism. Paradoxically, people who made the most errors in this task were the most confident (a Dunning-Kruger effect). In Experiment 2 ( N = 610 adults), we used face-space theory and participant qualitative reports to identify key facial attributes that distinguish AI from human faces but were misinterpreted by participants, leading to AI hyperrealism. However, the attributes permitted high accuracy using machine learning. These findings illustrate how psychological theory can inform understanding of AI outputs and provide direction for debiasing AI algorithms, thereby promoting the ethical use of AI.

Extracting intersectional stereotypes from embeddings: Developing and validating the Flexible Intersectional Stereotype Extraction procedure

“…Indeed, so much of the observed bias in AI more generally seems to be a result of imbalances in the training data. For example, AI-generated faces are judged as more realistic than natural human faces, but only for White faces due to the dominance of White faces in training data ( 44 , 45 ). Moreover, androcentric and ethnocentric dominance can also serve to reify mental stereotypes about which groups should dominate in society.…”

The Era of Artificial Intelligence Deception: Unraveling the Complexities of False Realities and Emerging Threats of Misinformation

“…AI hyperrealism has emerged as a groundbreaking yet disconcerting phenomenon in the evolving digital technology landscape. This technological progression has given rise to artificial intelligence systems that not only produce images of human faces that are indistinguishable from authentic ones but, in a paradoxical twist, these synthetic faces can sometimes be perceived as more "human-like" than the faces of real people [106,107]. This striking and counterintuitive aspect of hyperrealism underscores a significant challenge: the potential for AI-generated images to appeal more to human biases, thereby amplifying their effectiveness in manipulating public opinion.…”

“…The authors of [106] highlight the impressive realism of AI-generated faces. However, they raise concerns about the public's ability to discern these synthetic visages from real human faces-an issue of growing importance as these AI faces could be used to craft deceptive social media profiles.…”

“…This discrepancy in perception can lead to severe repercussions, such as being duped by a fake online persona. In their research, the authors of [106] sought to pinpoint the visual characteristics that set AI faces apart from humans. This is a critical step in understanding why people might overlook AI faces despite their ubiquity and lifelike appearance.…”

“…The authors of [106] comprehensively explored two crucial aspects of AI hyperrealism. First, it emphasizes the importance of integrating psychological theories to deepen our understanding and develop strategies to counteract the effects of AI hyperrealism.…”