Accepted for/Published in: Journal of Medical Internet Research
Date Submitted: Dec 10, 2024
Date Accepted: Jun 11, 2025
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Artificial Intelligence-Generated Images Recapitulates Ageism in the Digital Age
ABSTRACT
Background:
Positive images of aging in media are linked to better health outcomes in older adults, including increased life expectancy. The advent of generative systems allow almost anyone to generate high-quality images in real time. Thus, they often reflect and amplify societal age-related biases, a phenomenon known as digital ageism. There is a gap in research on how digital ageism in AI-generated images changes over time as these AI systems continue to evolve.
Objective:
This study examined how instances of digital ageism in AI-generated representations of older adults change over time. Examining how digital ageism evolves in AI-image generation systems can provide insight into the interplay between technology advancements, societal attitudes toward aging, and the well-being of older adults interacting with the dynamic digital landscape.
Methods:
This qualitative and quantitative project compared 164 images generated by Open AI’s DALL-E 2 image generator at two time points, one year apart. Text prompts, identical at both points of inquiry, included terms from the geriatric lexicon (e.g., frail older adult, dementia, etc.). Eleven individuals evaluated the images generated at each point to identify perceived sex, race, socioeconomic status (SES), and emotional expression. The findings were compiled to determine the frequency with which each category was represented. These values were compared between two years and evaluation categories using type III two-way Analysis of Variance (ANOVA).
Results:
164 images were analyzed by eleven independent investigators. For race distribution, the number of images with White older adults was 5-fold higher than other races both years. The mean number of Asian racialized representations increased from 20 to 31 (P=.004). The mean values of other racialized representations also increased from 6 to 14 (P=.007). Representations of socioeconomic status showed mean values for middle class significantly higher than other classes in 2022 and 2023 with no changes in socioeconomic status from one year to the next. Prompts were largely neutral for expression terms while image analyses for expressions s did not show significant difference between positive, neutral or negative emotions from 2022 to 2023. Prompts used for image generation had more male-oriented terms thus as expected, male representation was higher in images as compared to females with no difference in sex representation between the two time points.
Conclusions:
Despite the increased recognition of the importance of social emphasis on positive views on aging, AI text-to-image generators persistently generated negative imagery with White racialized representation higher than other races in the images generated at two time points, one year apart and no statistical difference between positive, neutral and negative emotions even with emotion-neutral prompts. A limitation of this study is that it focuses only on AI text-to-image generation and no other AI-generated content that may express digital ageism.
Citation
