“Others Are More Vulnerable to Deepfakes than I am”: A Mixed-Method Approach to Investigate the Third-Person Perception of Deepfakes Among Chinese Adults

Artificial intelligence enables the widespread production of videos that closely resemble authentic ones, known as “deepfakes.” However, research investigating individual perceptions of and cognitive biases toward deepfakes is limited, and most related research has relied on the singular approach of structural equation modeling (SEM). In this study, by integrating theories on fear of missing out (FOMO) and considering the mediating effects of media exposure and cognitive load (CL), we propose a two-stage mixed-methods approach that combines SEM and network analysis to demonstrate the factors potentially contributing to the third-person perception (TPP) of deepfakes. We demonstrate that the TPP of deepfakes has a greater impact on individuals experiencing FOMO, frequent social media users, and people with high CL. Furthermore, our results identified two key factors linked to perceived authenticity: demographic patterns, where women on average, rated deepfakes as more authentic than men, and a skills-based pattern, where individuals with lower recognition ability were more susceptible. These findings emphasize the critical importance of investigating underlying psychological determinants and media consumption patterns to comprehensively understand the mechanisms driving the dissemination of deepfake-enabled misinformation.