Affiliation:
1. Georgia Institute of Technology Atlanta, Georgia, USA
Abstract
This study was a follow up to previous work (Tidler & Catrambone, 2021) which showed that one’s ability to correctly identify emotional states of others based on their facial expressions is strongly, positively correlated with their ability to detect fabricated videos known as “deepfake” videos. Because there has historically been a theoretical association between autism spectrum disorder (ASD) and the aforementioned affect detection ability, this study was designed to address the question of whether those with ASD are less able to detect deepfake videos compared to neurotypical counterparts. Thirty participants who self-reported having been formally diagnosed with ASD were compared with twenty-six participants who self-reported not having been diagnosed with ASD in terms of their performance on a deepfake-video detection task. Although no significant performance differences were observed, the ASD group was significantly more confident in their judgments of the videos’ authenticity. The results of this study, while preliminary, potentially help to identify groups that may be particularly susceptible to deception via deepfake video so that steps can be taken to mitigate potentially dangerous outcomes of this susceptibility.