By Eivind Torgersen

Media professor believes it is meaningless to pit ‘true’ photographs against ‘fake’ AI images and calls for a more nuanced debate.

“Photography has always been contested,” says media professor Liv Hausken at the University of Oslo.

She believes this perspective is lost in discussions about how we should relate to the fact that artificial intelligence (AI) can produce images and illustrations that look like photographs of real people and events. In the future, it may become even more difficult to know what we should trust.

Several newspapers and media organisations have introduced something close to zero tolerance for AI-generated images. The fear of what artificial intelligence might do, and how it could mislead us, has elevated photography’s status to that of a transmitter of ‘the truth’. A status which, according to Hausken, it has never really deserved.

“I think awareness of all the things you can do with an old single-lens reflex camera is very low. A photograph is not automatically either true or false; it is a rhetorical device,” says Hausken.

In the Photofake project, she and her research colleagues have, among other things, studied the role that still images and video play in today’s disinformation crisis. They place this within a broader history in which classical photography is used not only to document factual circumstances.

From above, from below, or right up the nostrils?

Hausken recalls discussions among press photographers in the 1990s where the question was how far down one could bend to photograph Progress Party (Norwegian: Framskrittspartiet) politicians.

“They were to be photographed up the nostrils to give them a pig-like face. You weren’t supposed to go too far, because Progress Party politicians were also human beings, but some bending down was acceptable.”

“Female politicians were photographed from different angles than their male colleagues. Women were often shot slightly from above in order for them to appear small – men slightly from below so they would look powerful,” says Hausken. And she asks:

“Today’s press photographers know all this. So why is it now unproblematic to talk about photography as if it were a neutral form of documentation?”

The Falling Soldier – did he really fall?

Photographs from wars and international conflicts are perhaps the most problematic.

Several of the iconic images from the twentieth century have later proved to be controversial. There are claims that they were staged, or were composites made from several images.

One example is Robert Capa’s famous photograph ‘The Falling Soldier’ from the Spanish Civil War. Some say they can document that the soldier is merely pretending; others that he was not a soldier at all and that the image was taken far from the front.

“I think we should be particularly attentive to war photography. Even if the photographer, in a certain sense, conveys what is happening, as a writing journalist would do, they may sometimes have taken certain liberties to make the photograph happen, ” says Hausken.

Two images became one

A somewhat more recent example is the award-winning American war photographer Brian Walski, who in 2003 was fired from the Los Angeles Times because he used Photoshop to combine two images into one without disclosing it. The image shows a British soldier in Iraq asking a civilian man holding a child to sit down to avoid the line of fire.

The image was published in newspapers across the United States before irregularities were discovered and the photographer was confronted.

“They are two photographs taken very close together in time. Individually they are not perfectly composed, but if you combine them, it becomes a very good image. ”

A journalist might perhaps have described the situation as we see it in the composite image.

“It would at least have been interesting to know whether the photographer intended to convey what he saw, but which his camera could not capture in a single frame,” says Hausken.

She believes this is one of the reasons why we should have a more nuanced view of photographs and the communication of real events.

“Walski lost his job and, as far as I know, has not been able to work as a photojournalist since. This is a very strong reaction. One could imagine a different response, for example writing an explanation stating that this is an image of how the situation was at that moment. Not that second, but at that moment.”

Images have different functions

Far from war and combat operations, there was nevertheless a major stir in press circles when the Norwegian Broadcasting Corporation NRK used an AI-generated image of power lines to illustrate a brief item about the price of electricity in 2022. When they were alerted to the ‘blunder’, they quickly replaced it with a stock photograph of power lines.

At the time, Hausken told the online newspaper Journalisten that she felt the reaction was somewhat hysterical and panicked. She did not see the point in replacing a meaningless AI-generated image with an equally
meaningless image taken by a photographer with a camera.

She still doesn’t. Three more years of work with the Photofake project have convinced her that the discussion must be more nuanced.

“Function is decisive. Images, including photographs, have different functions.”

Sometimes an image should actually document something. At other times it should, for example, merely illustrate an article about what it is wise to bring to the mountain cabin at Easter.

“If the image is only meant to illustrate what you should take with you for Easter holiday in the mountains, it does not matter whether it is a drawing, a photograph, or an image made by artificial intelligence,” says Hausken.

Energy drain and cheap labour

Although she has a more relaxed attitude than some editors and press photographers, Hausken is not an advocate for AI-generated images. There are plenty of reasons to be sceptical.

Among other things, enormous amounts of electricity are consumed when you ask an AI to generate an image.

“I think the energy problem will only get bigger. We may face a situation where we need to make a choice: should energy be used to create AI-generated images, or should we use it to heat our houses in winter?” says Hausken.

And then there are those reasons that are not visible to us. Poorly paid employees working under terrible conditions to ensure that AI models are functioning as they should.

“There are people who make sure that unwanted images are filtered out of the material used to train AI. They must give detailed descriptions of images depicting violence and abuse so that the model can censor this type of material. It must be such a strain that I find it difficult to cope with the thought of it,” she says.

Regardless of the objections one might have to AI-generated images, they will not disappear. So how should we, as recipients, assess the images and photographs that flow through our feeds?

“There is no simple solution. It comes down to trust in institutions, trust in the sources, and a general scepticism – the same critical approach you would apply to a newspaper article,” says Hausken.