Assessment of Medically Relevant Ageism Inherent in Large Language Models

As the global population ages, the importance of reducing or eliminating social agism increases. However, although the medical field is known to be pervasive in terms of ageism against the older people, such assessments have not yet been conducted in terms of equity in large language models (LLMs). In this study, we created a dataset to assess the potential for medically related ageism among LLMs and attempted to visualize it. These results suggest that many LLMs exhibit stereotypical ageism in terms of reluctance to intervene in treatment and self-ageism. These results indicate the importance of developing a framework to avoid or reduce ageism in the social implementation of LLMs in the medical field.