prosopagnosia (13. 194)

Scoring the average ourselves and not much better than chance, we found this research project from the British Journal of Psychology and the University of New South Wales, pitting control participants with verified super-recognisers, we found some upbuilding ironies—via MetaFilter—to the self-selecting experiment to identify authentic human faces among AI-generated ones. First there was some gatekeeping to take put with a series of CAPTCHA verifications to prove one is a human by identifying traffic features to covertly train autonomous vehicles, and then, I wanted to repeat the trial to test my own biases, wondering if I was giving women and non-white presenting personae (a rejection of the Mar-a-Lago look to embrace our faults and asymmetries) pass or whether I was actually detecting certain tells. Glancingly as profile pictures or avatars, we wouldn’t have questioned the authenticity of any of these and erring on the side of voting human versus depersonalising is much preferable to being deemed a composite image but were nonetheless disturbingly off-putting judgments to make and revealing the tells only will make the masquerade harder to detect.