Political Deepfakes Are As Credible As Other Fake Media And (Sometimes) Real Media

We demonstrate that fabricated videos of public officials synthesized by deep learning (“deepfakes”) are credible to a large portion of the American public – up to 50% of a representative sample of 5,750 subjects – however no more than equivalent misinformation in extant modalities like text headlines or audio recordings. Moreover, there are no meaningful heterogeneities in these credibility perceptions nor greater affective responses relative to other mediums across subgroups. However, when asked to discern real videos from deepfakes, partisanship explains a large gap in viewers’ detection accuracy, but only for real videos, not deepfakes. Brief informational messages or accuracy primes only sometimes (and somewhat) attenuate deepfakes’ effects. Above all else, broader literacy in politics and digital technology increases discernment between deepfakes and authentic videos of political elites. Our findings come from two experiments testing exposure to a novel collection of deepfakes created in collaboration with tech industry partners. ∗For excellent research assistance, we thank Jordan Duffin Wong. We thank the Wiedenbaum Center at Washington University in St. Louis for generously funding this experiment. For helpful comments, we thank the Political Data Science Lab and the Junior Faculty Reading Group at Washington University in St. Louis; the Imai Research Group; the Enos Research Design Happy Hour; the American Politics Research Workshop at Harvard University; the Harvard Experiments Working Group; and Jacob Brown, Andy Guess, Connor Huff, Yphtach Lelkes, Jacob Montgomery, and Steven Webster for helpful comments. We thank Hany Farid for sharing video clips used in this project. We are especially grateful to Sid Gandhi, Rashi Ranka, and the entire Deepfakeblue team for their collaboration on the production of videos used in this project. All replication data and code is publicly available here. All aspects of the research protocol were approved by the institutional review boards of Harvard University, Washington University in St. Louis, and Pennsylvania State University. †Ph.D. Candidate, Harvard University; URL: soubhikbarari.org, Email: sbarari@g.harvard.edu ‡Assistant Professor, Washington University in St. Louis; URL: christopherlucas.org, Email: christopher.lucas@wustl.edu §Assistant Professor, Pennsylvania State University; URL: kevinmunger.com, Email: kmm7999@psu.edu