Politics and porn: how news media characterizes problems presented by deepfakes

ABSTRACT

“Deepfake” is a form of machine learning that creates fake videos by superimposing the face of one person on to the body of another in a new video. The technology has been used to create non-consensual fake pornography and sexual imagery, but there is concern that it will soon be used for politically nefarious ends. This study seeks to understand how the news media has characterized the problem(s) presented by deepfakes. We used discourse analysis to examine news articles about deepfakes, finding that news media discuss the problems of deepfakes in four ways: as (too) easily produced and distributed; as creating false beliefs; as undermining the political process; and as non-consensual sexual content. We provide an overview of how news media position each problem followed by a discussion about the varying degrees of emphasis given to each problem and the implications this has for the public’s perception and construction of deepfakes.

KEYWORDS:

• Deepfakes
• image abuse
• artificial intelligence
• cybermisogyny
• disinformation

Acknowledgement

We want to thank Marta Kopp for her help on earlier versions of this research.

Disclosure statement

This is to acknowledge that there is no financial interest or benefit associated with this research.

Notes on contributors

Chandell Gosse is a PhD Candidate in Media Studies in the Faculty of Information and Media Studies at Western University. Her research takes an interdisciplinary approach and sits most broadly at the intersection of feminism, digital culture, and anti-violence work. Find her on Twitter @ChandellEnid

Dr. Jacquelyn Burkell is an Associate Professor in the Faculty of Information and Media Studies at the Western University. Her research focuses on the social impact of technology, with a specific emphasis on privacy. In her work, she links empirical research with legal and policy outcomes.

Data availability statement

The data that support the findings of this study are openly available in fig share at https://doi.org/10.6084/m9.figshare.12098307.

Additional information

Funding

This work was supported by the Social Sciences and Humanities Research Council of Canada (SSHRC) under [grant number 895-2015-1002].