Deepfakes claims, deepfake detection and the forensic mindset

Curator's Note

An apparent confession video of a senior politician arrested by Myanmar coup authorities in early 2021 implicated Aung San Suu Kyi in corruption. Noting the visual distortions and puzzling voice quality, online audiences and the media scrutinized it for signs of digital manipulation. Meanwhile the backdrop, context and his face were analyzed for fakery or staging in videos (1,2) of the former Georgian president, Mikheil Saakashvili, announcing he was unexpectedly back in the country on the eve of elections in 2021.

Both videos were accused of being deepfakes. In Myanmar the claims came thick and fast on Facebook and Twitter, reinforced by online detection tools erroneously identifying the video as a deepfake. In Georgia they came from senior levels of the ruling party. Both cases partially align with the idea of the ‘liar’s dividend’. Here, the mere existence of deepfakes empowers people to dismiss real footage as false, and places the onus on the public or the media to prove it is true. Deepfakes are framed in public discussion as a mysterious, existential threat: this combines with lack of understanding of their actual technological capabilities or evolution, and no widespread skills and tools to counter these claims to compound an existing problem, often experienced by human rights activists, of controversial footage dismissed by opponents as faked.1, 2

The Myanmar case also reflects how the perceived potential for deepfakery allows already skeptical publics to reinforce their own existing assumptions. The video in Myanmar was a forced confession not a digitally falsified forgery, but just like in online conversations where Trump voters willed his January 2021 Oval Office ‘concession’ address to be a deepfake, many in Myanmar wanted this statement to be an absolute fake.

What are the implications of these illustrative scenarios? On the one hand, an unwinnable forensic turn in the general public encourages them to be DIY investigators in an increasingly manipulatable audiovisual reality where scrutinizing the pixels with the naked eye will not work. Simultaneously activists and journalists in Myanmar, Georgia and elsewhere more readily able to interrogate the images do not have the tools to challenge these claims of deepfakery that are deployed by those in power, or reached-for by those without power. Underlining the forensic turn in analyzing and mobilizing the evidence of our eyes and ears to ascertain true and false are critical questions of equity and access - who does it make sense to turn to in order discern deepfakes, and how will access to the skills, resources and tools  be globally available where needed most?

  1. Wahl-Jorgensen, Karin, and Matt Carlson. "Conjecturing fearful futures: Journalistic discourses on deepfakes." Journalism Practice 15.6 (2021): 803-820.

  2. Gregory, Sam. "Deepfakes, misinformation and disinformation and authenticity infrastructure responses: Impacts on frontline witnessing, distant witnessing, and civic journalism." Journalism (2021): 14648849211060644.

Add new comment

Log in or register to add a comment.