GENERATIVE AI
Study of AI-generated Images on Teenagers: Beautiful, Happy, and Slim
By studying images generated on youth with the generative image tool Dall-E3, a positive image is shown - but no variation in their appearance could be created.
Gustav Westberg, a communication researcher in the Swedish language at Örebro University, together with Professor Gunhild Kvåle at the University of Agder, has examined the results of the images generated on youth in the Dall-E3 tool.
By using general prompts such as "create images of teenagers" and by giving specific instructions to Dall-E to come up with prompts to generate images, the study has examined the result of how an AI visualizes youth.
Both cases yielded the same result: the youth are beautiful, with thick hair and perfect skin. Everyone wears jeans and the same type of shoes, they look happy and radiate success - while the images are ethnically diversified. The researchers believe this is because OpenAI, which owns Dall-E and ChatGPT, has taken the criticism of lack of diversity in earlier versions of the model into account.
A second common feature is that the images only show happy people. The teenagers are doing well, and their bodies, clothes, and gadgets signal wealth and success. They all appear to belong to an upper middle class.
- We received images of youths studying in libraries, having picnics, playing instruments and video games together, engaging in sports, cooking, and doing art, activities that are highly valued in our culture. There are also norms embedded in the technology. Everyone is slim, and not even specific prompts that requested more varied bodies could change this, says Gustav Westberg.
A third common feature is that the images in various ways give the impression of depicting genuine scenes from real life through the use of photographic expressions like depth of field, light contrast, and under- and overexposure.
The authors also discuss a fourth distinctive feature of how AI images can depict scenarios taken from the imagination. An example is teenagers skateboarding inside a snow globe at the North Pole - even though photorealistic images are the standard with Dall-E3.
- Generative AI must be regarded as revolutionary as photography and television. Mass communication is by nature repetitive, and we should critically question what the consequences are when our visual communication becomes increasingly smoothed out, despite the seemingly endless possibilities that generative AI offers, says Gustav Westberg.