Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Nov 27;10(11):231214.
doi: 10.1098/rsos.231214. eCollection 2023 Nov.

Deepfake detection with and without content warnings

Andrew Lewis  1 , Patrick Vu  2 , Raymond M Duch  1 , Areeq Chowdhury  3
Affiliations

Deepfake detection with and without content warnings

Andrew Lewis et al. R Soc Open Sci. .

Abstract

The rapid advancement of 'deepfake' video technology-which uses deep learning artificial intelligence algorithms to create fake videos that look real-has given urgency to the question of how policymakers and technology companies should moderate inauthentic content. We conduct an experiment to measure people's alertness to and ability to detect a high-quality deepfake among a set of videos. First, we find that in a natural setting with no content warnings, individuals who are exposed to a deepfake video of neutral content are no more likely to detect anything out of the ordinary (32.9%) compared to a control group who viewed only authentic videos (34.1%). Second, we find that when individuals are given a warning that at least one video in a set of five is a deepfake, only 21.6% of respondents correctly identify the deepfake as the only inauthentic video, while the remainder erroneously select at least one genuine video as a deepfake.

Keywords: deepfake; experiments; manual detection.

PubMed Disclaimer

Conflict of interest statement

We declare we have no competing interests.

Figures

Figure 1.
Figure 1.
Manual detection of deepfakes with no content warnings. Notes: Whiskers show 95% confidence intervals calculated from a regression of the indicator outcome variable on an indicator for being in the treatment group using robust standard errors.
Figure 2.
Figure 2.
Manual detection of deepfakes with a content warning. Notes: Distribution of videos identified as deepfakes by participants who are warned they will see at least one deepfake in a set of five videos. Video four is the deepfake. The first five columns show the distribution of choices among those who chose only one video. The final group consists of those selecting more than one video.
Figure 3.
Figure 3.
Distribution of videos selected as ‘out of the ordinary’. Notes: The distribution of videos identified as out of the ordinary by participants in the control and first experimental treatment conditions. The sample is the subset of participants who indicated having observed something out of the ordinary. These participants were asked to identify which videos were out of the ordinary, with the option to select more than one video. The blue bars represent the control group, who viewed five authentic videos. The red bars represent the treatment group, who viewed four authentic videos and one deepfake (the fourth video).
See this image and copyright information in PMC

References

    1. FBI. 2021. Malicious Actors Almost Certainly Will Leverage Synthetic Content for Cyber and Foreign Influence Operations.
    1. Europol. 2020. Malicious Uses and Abuses of Artificial Intelligence.
    1. Fallis D. 2020. The epistemic threat of deepfakes. Phil. Technol. 34, 623-643. ( 10.1007/s13347-020-00419-2) - DOI - PMC - PubMed
    1. Chesney B, Citron D. 2019. Deep fakes: a looming challenge for privacy, democracy, and national security. Calif. L. Rev. 107, 1753.
    1. Kunda Z. 1990. The case for motivated reasoning. Psychol. Bull. 108, 480. ( 10.1037/0033-2909.108.3.480) - DOI - PubMed

LinkOut - more resources