Psychological inoculation against deepfakes

Despite growing interest in countering the threat of deepfake misinformation, most interventions aimed at enhancing human detection have thus far proven unsuccessful. In the current study we explore the potential of a popular psychological intervention: cognitive inoculation—which to our knowledge—has not been tested in the context of deepfakes. In a pre-registered between-subjects experiment (N = 297), we evaluate whether an inoculation-based video intervention improves people’s deepfake detection skills compared to a control intervention. The inoculation intervention forewarned participants that they may be manipulated by deepfake content and prebunked it by presenting and deconstructing deepfake misinformation in weakened form. Following the intervention, participants were asked to classify a series of 10 videos as either real or deepfaked. We find that mean detection accuracy of deepfake recognition was significantly higher in the inoculation group compared to the control group (d = 0.625, 95% CI [0.61, 1.32]) indicating that the inoculation intervention substantially improved deepfake discernment. Further investigation of the mechanisms behind the inoculation process finds that both motivation to resist deepfakes and self-reported ability to recognize them increased following the intervention, with the latter mediating the impact on deepfake discernment. Overall, our findings highlight the potential of inoculation interventions to build resistance to deepfake misinformation.