9/19/2024 Isabel Gauthier: Domain-General Object Recognition Predicts Human Detection of AI-Generated Faces
CCN Brown Bag
Isabel Gauthier, PhD
David K. Wilson Chair of Psychology
Vice Chair Department of Psychology
Professor of Radiology and Radiological Sciences
Date: Thursday, September 19, 2024
Time: 12:10PM-1:00PM
Location: 316 Wilson Hall
Domain-General Object Recognition Predicts Human Detection of AI-Generated Faces
Faces generated by Artificial Intelligence (AI) are now virtually indistinguishable from real ones. However, people vary in their ability to detect AI-generated faces—a skill that has only recently become relevant. Advances in high-level vision have identified a domain-general ability called O, which predicts perceptual, memory, and categorization performance across various objects. We propose that O is ideally situated between broad cognitive skills and specific abilities, making it a strong predictor of AI-face detection—a stable and measurable skill. Using structural equation modeling, we test whether O predicts AI-face detection, compared to face recognition ability in one study and to general intelligence in a second study. Our results suggest that O may play a role in a wide range of perceptual judgments. Down the road, studies like these could improve how we interact with AI and potentially help optimize training data for AI models.
Questions? Contact Isabel Gauthier.