How media outlets and Internet companies fight deepfakes

The new article “Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation” by Ángel Vizoso, Martín Vaz-Álvarez, and Xosé López-García, all of University of Santiago de Compostela in Media and Communication, looks at how three media outlets and three Internet-based companies deal with a new form of misinformation: deepfakes.

Deepfake is a compound word formed from ‘deep learning’ – a form of artificial intelligence – and ‘fake’. According to previous studies on the matter, they are defined as often highly realistic digital manipulations used on videos with computer code. They are a result of what is known as  Generative Adversarial Networks designed to replace actual human faces or voices.  

For example, a deepfake video could show a known politician issuing statements that they have not made by combining video data with inserted audio that is made to sound like the politician in question. They present a challenge for communication and verification of information.

The media organizations studied were well-known: The Wall Street Journal, The Washington Post, and Reuters. Similarly, the Internet organizations analyzed were ubiquitous: Google, Facebook, and Twitter. 

The study consisted of two stages. First, the authors conducted a Systematic Literature Review that turned a sample of 54 research items: conference presentations and published articles. Second, a case study was conducted of the six organizations mentioned above to see what sort of measures they have taken to combat deepfakes. 

All the media organizations had adopted different approaches against deepfakes. The Wall Street Journal had division of 21 journalists whose task was to combat misinformation, particularly deepfakes. 

The Washington Post used an uniform standard in fighting deepfakes as they use on other fake news, but had added a team of video experts in order to counter deepfakes. Finally, Reuters used a strategy of collaborating with Facebook in order to detect deepfakes, and had a devoted blog on the matter of debunking them.

The Internet-based companies also had differing approaches against deepfakes. Google volunteered their datasets of manipulated and non-manipulated videos for the research community, allowing them to utilize the vast quantity of material collected by Google. 

Facebook, in addition to collaborating with Reuters, financed a specific initiative within their ‘Deepfake Detection Challenge’. Facebook also seeks to delete faked materials from their social network with the help of fact-checking organizations. 

The other social media giant, Twitter, used a set of four rules against deepfakes: identification through notice tweets, warning of manipulated content before sharing it, inclusion of a link to genuine news articles explaining the manipulation, and elimination of potentially safety-endangering materials.

The authors make several comments on the findings. First, that collaboration between organizations is increasing like Reuters and Facebook, and with fact-checking organizations. Second, similar technology is used against deepfakes as is used in making them: journalists are increasingly trained in the use of AI-based methods. Third, there seems to be a growing synergy between communications and academic circles.

The importance of the matter is likely to increase in years to come, the authors note. It is also an understudied matter due to its novelty. The authors acknowledge a certain Western-based bias in the studied materials, a limitation hopefully countered by further studies.

The article “Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation” by Ángel Vizoso, Martín Vaz-Álvarez, and Xosé López-García is in Media and Communication. (Open Access).

Picture: Untitled by Sebastian Mark.

License Unsplash.

Give us feedback