Deepfakes, manipulated videos synthesized by deep learning, are the newest tools in the misinformation arsenal box. Easily accessible via open-source applications, they offer a cheap and efficient way to create deceptive videos of public figures. How powerful are deepfakes? New research finds that misinformation consumed in a video format is no more effective than misinformation in textual headlines or audio recordings. However, the persuasiveness of deepfakes is equal and comparable to these other media formats like text and audio.
The research used two tests to measure the effectiveness of deepfake messaging among 5,750 respondents. The first test was conducted in a social media feed environment, surrounded by regular social media content.
Respondents saw or heard either a deepfake video, a fake audio, a SNL-skit like exaggeration, or a text headline. Deepfake stimuli featured Senator Elizabeth Warren in several scenarios. In them, Warren:
- Calls Biden a pedophile
- Calls Trump a pedophile
- Revives an old controversy about identifying with indigenous people.
- Creates an unexpected controversy about LGBTQ lifestyle
- Goes back on a political position that eliminating student loan debt for anyone is fair or realistic
In all, just under half (47%) of respondents believed the deepfake video was real. However, the deepfake scored no better or worse compared to the audio or text false messaging.
Further, the research delved into respondent characteristics (e.g., gender, income, political party and more) to see if any are predictors of susceptibility to deepfakes. The results showed no significant differential between deepfakes vs. false text vs. audio misinformation. However, selective acceptance of information based on previous beliefs may influence an individual’s response to deepfakes.
The second test alerted respondents to look for misinformation. Participants were asked to identify if a video was true or fake. In all, 55% of the videos were identified correctly. Interestingly, political orientation did have an impact here. Both Republicans and Democrats underestimated the authenticity of real videos if it went against their party or candidate. They were much more likely to call a real video fake if it made their political leader or party look bad.
Seeing is not necessarily believing these days. Based on these findings, deepfakes do not facilitate the dissemination of misinformation more than false texts or audio content. However, like all misinformation, deepfakes are dangerous to democracy and media trust as a whole. The best way to combat misinformation and deepfakes is through education. Informed digital citizens are the best defense.