A study found that Generative AI such as ChatGPT can reflect certain stereotypes when describing older adults. It confirmed a tendency to portray older adults as kind and warm while depicting their agency and expertise as relatively low.
A research team led by Choi Mun-jung, a professor at the KAIST Graduate School of Science and Technology Policy, said on the 28th that it quantitatively analyzed age-related bias in sentences generated by OpenAI's Generative AI model GPT-4o. The findings were published in a special issue of the international journal in gerontology, The Gerontologist, in the February edition.
Generative AI learns from large-scale data such as internet documents to generate answers. Concerns have been raised that it can also learn social prejudices contained in the data during this process. While previous AI bias research has mainly focused on gender and race, this study centered on ageism, which views or discriminates against a particular group negatively because of age.
The team asked GPT-4o to describe the characteristics of each age group from the teens to the 90s and collected 900 sentences. It then analyzed the sentences using a social psychology theory that views perceptions of people or groups through two dimensions: "warmth" and "competence."
The analysis found that people 60 and older scored high on "warmth," which refers to qualities such as kindness, trustworthiness, and consideration. In contrast, "competence," meaning ability, expertise, and efficiency, tended to be expressed as lower than that of younger generations.
It also found repeated use of similar expressions for those in their 70s and older. The researchers noted that while the AI depicted older adults as wise and benevolent figures, it showed their willingness to express opinions actively and act proactively relatively less.
The team viewed that such expressions, when exposed repeatedly, could reinforce social prejudice against older adults. It warned that if a perception takes hold in digital services that older adults are passive users, it could lead to "digital ageism," which blocks older adults' use of and participation in technology.
Choi Mun-jung said, "AI bias is not just a technological issue but a societal one," and added, "To build inclusive AI, diverse generations must take part in the development process."
References
The Gerontologist (2026), DOI: https://doi.org/10.1093/geront/gnaf291