Researchers tested kids on telling AI content from real human content. They scored at or below a coin flip. Worse than guessing.
A new study in Technology in Society tested 37 children, ages 6-10, after introducing them to a "SmartBot" that could write text, invent photos, and create art.
The Findings
Children performed at or below chance in discriminating human-created from AI-generated content across text, images, art, and faces -- and their overall accuracy was significantly below that of adults in a comparison group.
Kids' individual ability to spot AI content was significantly and negatively associated with parent reports of at-home technology use -- more screen time correlated with worse detection, not better.
One silver lining: 95% of kids understood the AI agent itself wasn't alive. They just couldn't tell its work apart from a human's.
The Takeaway
The researchers' takeaway: more time on digital devices is linked to susceptibility to artificial content and a real risk of mistaking AI output for something a person made, underscoring the need for early AI literacy education.