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.
Original language | English |
---|---|
Pages (from-to) | 1390-1403 |
Number of pages | 14 |
Journal | Psychological Science |
Volume | 34 |
Issue number | 12 |
Early online date | 13 Nov 2023 |
DOIs | |
Publication status | Published - 1 Dec 2023 |
Bibliographical note
AcknowledgmentWe thank Sophie J. Nightingale and Hany Farid for providing open access to their stimuli and data.
Funding
This research is supported by the Australian Government through the Australian Research Council’s Discovery Projects funding scheme (Project No. DP220101026), a TRANSFORM Career Development Fellowship to A. Dawel from the Australian National University College of Health and Medicine, and an Experimental Psychology Society Small Grant to C. A. M. Sutherland. The funders had no role in developing or conducting this research.
Data Availability Statement
Supplemental materialsAdditional supporting information can be found at https://journals.sagepub.com/doi/full/10.1177/09567976231207095
Open Practices
This article has received the badges for Open Data and Open Materials. More information about the Open Practices badges can be found at http://www.psychologicalscience.org/publications/badges.
Keywords
- StyleGAN2
- artificial intelligence
- face perception
- face-space theory
- generative adversarial network
- open data
- open materials
Fingerprint
Dive into the research topics of 'AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones'. Together they form a unique fingerprint.Datasets
-
Data From AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones
Miller, J. E. (Creator), Steward, B. A. (Creator), Witkower, Z. (Creator), Sutherland, C. (Creator), Krumhuber, E. G. (Creator) & Dawel, A. (Creator), University of Aberdeen, 3 Nov 2023
DOI: 10.1177/095679762312070, https://journals.sagepub.com/doi/suppl/10.1177/09567976231207095/suppl_file/sj-pdf-1-pss-10.1177_09567976231207095.pdf
Dataset