Generative AI systems like ChatGPT carry age-related stereotypes, a study from the Korea Advanced Institute of Science and Technology (KAIST) has found. The research team quantitatively analyzed responses to uncover hidden prejudices embedded in the technology. These biases may shape how users perceive different age groups.
The work sheds light on how AI can reflect societal prejudices, potentially reinforcing negative stereotypes at scale. As generative models become more integrated into daily life, such biases could influence everything from hiring decisions to healthcare advice. The researchers emphasize the need for more inclusive development practices.
Specific metrics from the study were not provided in the available reporting, but the team confirmed the presence of systematic age-based patterns in the AI's language. The analysis focused on comparing how the system responded to prompts involving different age demographics.
The findings suggest that current training data and model design may inadvertently encode ageism. Developers are urged to audit for such biases and adjust datasets to reflect a more balanced view of aging. Without intervention, these tools could perpetuate harmful stereotypes.
Experts outside the study have called for broader transparency in AI development. The KAIST team plans to extend their methodology to examine other forms of bias in future work.