Summarize this content to 2000 words in 6 paragraphs
To compare AI outputs to the lives of actual teens, researchers asked workshop participants to write words that came to mind about teenagers, to rate words on how well they describe teens and to complete prompts given to AI models. Similarities between AI systems’ responses and the teens’ were limited. (BigStock Photo)
A University of Washington study on how artificial intelligence systems portray teenagers found that the systems’ reliance on media coverage about teens led to strong negative responses.
Information School doctoral student Robert Wolfe and a UW team looked at two common, open-source AI systems trained in English and one trained in Nepali, in order to compare models trained on data from different cultures.
According to UW News, in the English-language systems, around 30% of the AI system responses about teens referenced societal problems such as violence, drug use and mental illness. The Nepali system was closer to 10%. The researchers also held workshops with groups of teens from the U.S. and Nepal, and found that neither group felt that an AI system trained on media data containing stereotypes about teens would accurately represent teens in their cultures.
“We found that the way teens viewed themselves and the ways the systems often portrayed them were completely uncorrelated,” Wolfe said. “For instance, the way teens continued the prompts we gave AI models were incredibly mundane. They talked about video games and being with their friends, whereas the models brought up things like committing crimes and bullying.”
News stories are seen as “high-quality” training data for AI systems, because they’re often factual, but they frequently focus on negative stories.
“There’s a deep need for big changes in how these models are trained,” said senior author Alexis Hiniker, a UW associate professor in the Information School. “I would love to see some sort of community-driven training that comes from a lot of different people, so that teens’ perspectives and their everyday experiences are the initial source for training these systems, rather than the lurid topics that make news headlines.”
The research was presented Oct. 22 at the AAAI/ACM Conference on AI, Ethics and Society in San Jose, Calif.
Read more at UW News for details on the AI systems that were used in the study.
