The More Attractive the More Real? Cognitive Processes Involved in Deepfake Detection in Young and Older Adults

Humans struggle to distinguish between real and deepfake face stimuli. Given the important forensic implications, cognitive science research has started focusing on mechanisms (e.g., attractiveness) impacting detection accuracy. Because older adults are particularly vulnerable to online fraud and misinformation, this study investigates whether subjectively perceived facial attractiveness is associated with deepfake detection in a sample of older (N = 37) and younger (N = 35) adults. They categorized 120 faces depicting young Caucasian adults (50% real, 50% generated by a Diffusion Model) as real or deepfake, followed by attractiveness ratings. Results confirm lower detection accuracy in older adults. Perceived attractiveness was associated with higher detection accuracy for real faces but lower accuracy for deepfakes, especially in older adults. Response times revealed that participants were slower to categorize real and increasingly attractive faces. These findings offer key insights into cognitive processes during deepfake face detection, paving the way for future age-tailored research.