THIS CONTENT IS BROUGHT TO YOU BY THE University of Agder - read more

AI-generated image of four teenagers, two boys and two girls.
“The young people in the images are strikingly similar. They all wear jeans and Converse shoes, are beautiful, with lovely features and voluminous hair, and none have pimples," researcher says.

AI-generated images of teenagers: 
"They're made so beautiful that it becomes intrusive"

Researchers have found that AI-generated images have four common features. Some of these give cause for concern, researchers believe. 

Published

If you've been using artificial intelligence (AI) for a while, you may have noticed that both the texts and images it creates can be a bit generic and polished. Some would call them soulless.

Or, as Professor Gunhild Kvåle at the University of Agder puts it:

“ChatGPT has a voice that gives me a peculiar itch.”

Together with colleague Gustav Westberg at Örebro University in Sweden, she has investigated how the AI tool Dall-E 3 creates images of teenagers. Dall-E is an image generator that is built into ChatGPT.

One of the goals was to find out where this itch was coming from. What creates this artificial quality in AI images?

They found that the images have four common features. Some of them give cause for concern, the researchers believe.

Surface-level diversity

"This is a specific social category of young, successful, beautiful teenagers,” says Gunhild Kvåle.

“It's striking how, on the one hand, Dall-E pays attention to ethnic and gender diversity, while on the other hand, the images are not very diverse,” says Kvåle.

To obtain a sample of images that was representative and could be compared, the researchers provided general instructions like ‘create images of teenagers.’ They also asked AI to generate its own text prompts to create the images.

“You can see that the teenagers in the images are supposed to represent different ethnicities. It was also striking how strongly diversity was emphasised when ChatGPT itself wrote the prompts,” she says. 

This suggests that the companies behind the technology have taken on board the criticism of the lack of diversity in previous versions.

But:

“The young people in the images are strikingly similar. They all wear jeans and Converse shoes, are beautiful, with lovely features and voluminous hair, and none have pimples. This is a specific social category of young, successful, beautiful teenagers,” Kvåle says.

"They're made so beautiful that it becomes intrusive," she adds.

Happiness sells

The other common feature was that the images are very positive. The young people are depicted as studying together, attending concerts, roasting marshmallows, or  – absurdly – collaborating on a local community garden.

“Everyone's happy in the images, no one's sad. They engage in activities valued by society. But none of the images show them at work or sleeping. This is the leisure time of young people in the upper middle class,” says Kvåle.

There are also some norms written into this positive portrayal. All the individuals are thin. Not even direct prompts could change this.

Kvåle explains that the technology sets clear boundaries for what kinds of images are possible to create. This is not entirely positive, although the intentions are good. 

"We can see it in the context of the culture we live in, where the texts and images we share are not only meant to objectively inform, but also to promote ourselves,” she says.

Almost real, but not quite

Kvåle notes that the lighting and how things are placed in the foreground or background mimic photography. This was the third common feature the researchers identified.

The settings depicted in the images vary from parks and youth clubs to concert stages and messy teenage bedrooms.

“The images create an impression of authenticity, but through the context, these young people are also positioned socially. They're never depicted at work or in urban settings associated with social issues,” she says.

The limitations of imagination 

The fourth discovery made by the researchers is how AI can display imaginative situations. One such example is young people skating inside a snow globe at the North Pole.

However, this is not the norm in Dall-E. It is something you have to specifically ask for.

“Photorealistic images are obviously preferred by Dall-E. Sometimes, they turn into graphic illustrations, but photorealism is clearly the standard,” explains Kvåle.

Critical awareness 

Image generators such as Dall-E 3 have become highly proficient and accessible to all. However, we are not drowning in AI-generated images. There is surprisingly little of it, in fact, according to Kvåle.

“It's said that everything will be changed by artificial intelligence, but that's clearly not true. Newsrooms, communication departments, and institutions have industry standards. Not everything changes overnight just because it’s possible,” she says.

At the same time, the researchers were surprised by how similar the images were.

“Services like Dall-E can have a strong influence on our visual culture. We owe it to one another to have a critical awareness of these images, as they don't represent the way we want our society to look. And that makes the world a bit more boring,” concludes Kvåle.

Reference:

Westberg, G. & Kvåle, G. The generic uniqueness of AI imagery: A critical approach to Dall-E as semiotic technologySage Journals, 2024. DOI: 10.1177/09579265241274788

Powered by Labrador CMS