A study by researchers from the University of Agder has revealed that while AI-generated images from tools like Dall-E 3 appear diverse on the surface, they often lack true depth and variety.
In collaboration with Orebro University, the researchers found that AI-generated images of teenagers tend to follow a formula: polished, attractive, and overly positive representations of young people. Despite attempts to include ethnic and gender diversity, the images often depict the same social category of successful, beautiful teenagers.
The research also highlighted concerns about how AI’s depiction of happiness and photorealism could influence visual culture in a negative way and encouraged critical awareness of AI-generated content.
Humans defend AI from exclusion
In a study by Imperial College London, humans demonstrated empathy towards AI bots excluded from a virtual ball game. The experiment involving a game called Cyberball showed that participants often defended AI agents by including them in the game when they sensed unfair exclusion.
Researchers highlighted the study’s implications, showing humans tend to treat AI bots as social beings. Lead author Jianan Zhou remarked, “This offers valuable insights into human—AI interactions, potentially shaping how we design future AI”.
Interestingly, older participants were more likely to feel AI’s exclusion was unfair and take corrective action. The findings suggest that AI design should avoid making bots overly human-like to help users distinguish between real and virtual interactions. The study is published in Human Behavior and Emerging Technologies.
Source link
