人們過度自信辨識 AI 臉部

• The AI Face Test measures individual differences in detecting synthetic faces by analyzing domain-general object recognition ability, quantified as shared variance between perceptual and memory judgments of both novel and familiar objects[1] • Object recognition ability is a stable trait that remains consistent across retesting[1][2] • Modern face-generation systems no longer produce obvious flaws; realistic outputs show convincing faces that are difficult to judge using traditional visual cues[3] • The research employed latent variable modeling to test whether detection ability can be predicted by domain-general visual perception capabilities[1]

As face-generation technology continues to improve, the gap between plausible and real faces may widen, making recognition of human perceptual limitations increasingly important[3]. The discovery of potential 'super-AI-face-detectors'—individuals with exceptional ability to spot synthetic faces—suggests future applications in digital authentication and misinformation detection[3]. The finding that object recognition rather than expertise predicts detection ability has broad implications for training programs and defensive strategies against AI-generated imagery in news, social media, and security contexts.