What the deep history of deepfakes tells us about image credibility

The ability to manipulate and create images with new technologies presents various challenges for traditional media messages as well as scientific communication. However, as Joshua Habgood-Kut discusses the history of fake images rather than heralds a massive breach of confidence, technological innovation has contributed to the ongoing social problems associated with knowledge production.

We seem to be surrounded by images and videos of events that never happened: Pope Francis in Balenciaga. President Trump Arrested, Pentagon set on fireTom Cruise jumping over Keegan-Michael Key. These images are disturbing, and many commentators and scholars use them to raise difficult questions about the validity of visual media. Many people think we should be worried about fake photos:

“Who can trust the accuracy of photography now? (…) What damage this possibility will cause to private reputation and what confusion the historian of the future will entail. The photographs were kept in the belief that they, like the figures, could not lie, but here is the revelation that they can be made to lie with the most deceptive accuracy.

Such rhetoric gives the impression that we are confronted with something like epistemic apocalypse. Except that this warning is not from 2023, but from 1869. This passage is part of a New York World column (quoted from Andy Toocher’s book). Not exactly a lie) and was written after the trial of William H. Mumler, who was tried (and acquitted) of fraud and theft for selling “spiritual photographs” showing sitters with their dead relatives.

During photography historyOpinions about the reliability of visual media tend to fall into two camps. On the one hand, we find figures such as William Fox Talbot (inventor of the calotype process), who in 1844 proclaimed photography nature pencil. Defenders of photographic validity have emphasized the distinction between photography, drawing, and painting, arguing that although “the artist builds, the photographer reveals“. On the other hand, other photographers have been fascinated by the creative possibilities of photographic media, arguing that there is no essential difference between human participation in photographs and paintings.

Each side has its own aesthetic. “Direct” photography focused on the photography’s fidelity to the scene and disapproved of any manipulation, while pictorialist approaches such as Henry Peach Robinson’s often stretched the connection between photography and any recognizable scene and reveled in the possibilities of editing. and creative development.

The current deepfake panic is actually part of this long history of controversy over the epistemological and aesthetic role of photography. Deepfake warnings raise the same concerns as after Mumler’s trial. Here is Franklin Foer writing in The Atlantic in 2018:

“The problem is not only the spread of lies. The fabricated videos will raise new and understandable suspicions about everything we watch. (…) In other words, manipulated video will eventually destroy faith in our strongest remaining connection to the idea of ​​a shared reality.”

The threat posed by fake photos and videos is that they highlight certain skeptical possibilities. In the above column in the New York World, the main question was whether it is possible to trust the image of a public figure in what they call the “hugs of the festive Fleet”, or the late family member of the “Negro gang”. in the cotton field. Now the issue is whether we can trust the exposure of the abuse of power by a politician or the pictures of warfare that we see.

Public concern about deepfake video and image hint software argues that these fake media represent a new type of problem that we lack both the conceptual and institutional resources to address. Public concern after the Mumler trial reminds us that falsification and unreliability have as long a history as photography.

A deeper lesson in the history of photography concerns the nature of the problem that fake media creates. The current debate presents counterfeit software as an inherently dangerous technology. The concern is that if this technology is not tamed, banned, or controlled, deepfakes will pollute our information environment, destroying media credibility and disrupting our contact with reality.

When our future Cassandra are cautious, they recognize that real fakes are often wobbly, unconvincing, and have tell-tales. It is common knowledge that imaging software creates images of people with too many fingers. But, we are being warned, soon completely convincing fake videos and photos will flood our social networks and newspapers.

The situation with widespread fakes masquerading as real photographs is not hypothetical. Until the 1880s, newspapers in the United States were illustrated with prints that were copied by hand from photographs. With the introduction of the halftone printing method, photomechanical reproduction became possible, but—like Tucher documents— Photographers continued to work by hand, retouching and processing their images to add color and life to them. These methods may not have been in the public domain, but they were defended in trade journals, with photographers claiming that counterfeiting was “almost a universal practice”.

Obviously, every falsified photo in an American newspaper is not good, at least if we want to use the photos as a source of evidence, not entertainment. How did newspapers deal with this problem? Toocher argues that journalists have adapted a technique that was used to combat falsification in written reporting: establishing a social norm against embellishment and turning “falsification” into a term of criticism. Despite the fact that there were opponents of this rule against fakes, New York Graphic printed a series of spectacular “composers‘ in the 1920s – this norm led to the formation of the professional identity of the photojournalist and to the so-called “golden age” of newspaper photography.

Heterosexual photographers argue that the purpose of photography is reliability and argue that human intervention in the photographic process undermines that purpose. On the contrary, photo artists argue that the purpose of photography is to express beauty and argue that human intervention is a necessary part of the creative process of photographing. The crisis in photography in the United States at the end of the nineteenth century shows us that both views are biased. Human participation in the photographic process is necessary, but it should not be motivated by the desire to create beauty. If we want reliable photos and videos that represent the world as it is, then we need reliable social practices related to the production and distribution of photos and videos, from setting lighting levels to image captions.

If we need reliable photos and videos that represent the world as it is, then we need reliable social practices related to the production and distribution of photos and videos.

When we use technology, from a hammer to a digital camera, it can often feel like we’re relying on a mechanical object. This feeling is wrong. When we rely on any technology, we are implicitly relying on a whole set of social practices. When we trust what we see in a photograph, we rely on the photographer to correctly set the light levels, on the camera designers to get light on the light sensor, on the software developer to convert this signal into a digital representation, etc. on the editor to avoid changing any important details, and on the subscriber to mark the image accordingly. Most of the time we don’t really know these people, we have to believe that they are involved in well-regulated image production and distribution practices.

Yes, there are reasons to be concerned about fake photos and fake videos. But these are social, not technological problems. These include societal concerns around best practices for photo and video production and distribution, as well as how best to fit image production into the wider media system. Perhaps watermarks, regulation, and detection software play a role in minimizing the harm from deepfakes, but we shouldn’t discount social measures such as creating social norms against falsification, systematic efforts to reinvent the skills needed to create realistic fakes. and an efficient and well-funded public media system.

This post is based on a published article by the author, Deepfakes and epistemic apocalypsepublished in Synthese. This blog post is part of a project funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant Agreement No. 818633).

The content created on this blog is for informational purposes only. This article represents the views and opinions of the authors, but does not reflect the views and opinions of the Impact of Social Science blog (blog) or the London School of Economics and Political Science. Please see our comment policy if you have any concerns about posting a comment below.

Image Credit: Fig.1: An unidentified elderly woman is seated, with three “spirits” in the background., William H. Mulle via the Getty Museum collection (no copyright), fig.2: Fading, 1858Henry Peach Robinson, via Wikimedia (Public Domain), fig.3: Texas corn carGeorge B. Cornish, via The Met (Public domain).