When Are Deep Fakes Effective? Perceptual Fluency and Subtle Digital Alterations

• In Event: Deepfakes in the 2020 Presidential Election: Detecting New Attempts at Deception

Abstract

Although digitally altered videos (also known as “deepfakes”) have been consistently discussed in both political and popular circles, we still know very little about why such videos are effective. This study introduces the idea of “perceptual fluency” to explain why even small changes to videos can have profound political consequences. Unlike more dramatic alterations, the subtleties discussed in this paper are not only easier for the average user to create, but are also more prevalent, as exemplified by the slowing down of a video by Nancy Pelosi to make it seem like she was slurring her words. This paper asks (1) do viewers actually notice even subtler changes and (2) how does they influence their evaluation of the subject of the video? To answer these questions, I employ an original experiment in which videos are manipulated in increasingly subtler ways. I expect to find that viewers will generally respond negatively to such alterations, meaning “deepfakes” may not have any positive use cases which serves as an important insight to future regulatory efforts.

Author

• Bryce Dietrich, University of Iowa