Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Art, aging, and artificial intelligence: Exploring age scripts in the universe of ai-generated art

Allen, Laura ; Xu, Wenqian LU orcid ; Nishikitani, Mariko ; Atul Patil, Vaishnavi ; Hule, Sunil and Bradley, Dana (2024) In Innovation in Aging 8(S1). p.925-926
Abstract
This study investigates age-related biases in AI-generated art through a mixed method investigation of images of older and younger individuals. Employing 19 text prompts associated with social exclusion for older adults, we collected 465 images from an AI-art generator, MidJourney, and conducted quantitative and qualitative analyses to understand differences between age group depictions. We used Amazon Rekognition software to detect the numerical image properties. Results that were compared statistically (STATA and Python) revealed significant differences in attributes: images with older people exhibited lower overall brightness (M = 46.54) and sharpness (M = 58.58) compared to younger individuals (brightness: M = 60.50; sharpness: M =... (More)
This study investigates age-related biases in AI-generated art through a mixed method investigation of images of older and younger individuals. Employing 19 text prompts associated with social exclusion for older adults, we collected 465 images from an AI-art generator, MidJourney, and conducted quantitative and qualitative analyses to understand differences between age group depictions. We used Amazon Rekognition software to detect the numerical image properties. Results that were compared statistically (STATA and Python) revealed significant differences in attributes: images with older people exhibited lower overall brightness (M = 46.54) and sharpness (M = 58.58) compared to younger individuals (brightness: M = 60.50; sharpness: M = 72.55). Moreover, older people were more likely to be smiling (16%) and with eyeglasses (56%) compared to younger individuals (6% and 12%), all with p < 0.001. Next, we conducted a qualitative semiotics analysis with a randomly selected sample of 76 images. Older people were often depicted with a strong emphasis on nostalgia, vulnerability, and aloneness, reflected in vintage-style clothing, worn objects, and worried expressions. In contrast, younger individuals were often portrayed with a focus on modernity, energy, and independence, characterized by vibrant clothing, digital technology, and confident expressions. Together, these findings demonstrate the perpetuation of negative and reductionist age scripts in AI-generated art, such as older adults’ incompetence, loneliness, fear, and outdatedness. Implications of this research extend to societal representations of aging, emphasizing the importance of challenging age-related biases in AI-generated content. Future research should explore refinement of generative models which foster more equitable outcomes. (Less)
Please use this url to cite or link to this publication:
author
; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Innovation in Aging
volume
8
issue
S1
pages
925 - 926
publisher
Oxford University Press
ISSN
2399-5300
DOI
10.1093/geroni/igae098.2986
language
English
LU publication?
yes
id
3fd508fc-7432-432c-b56d-8b40417fb815
date added to LUP
2025-01-03 18:31:01
date last changed
2025-04-04 15:23:35
@misc{3fd508fc-7432-432c-b56d-8b40417fb815,
  abstract     = {{This study investigates age-related biases in AI-generated art through a mixed method investigation of images of older and younger individuals. Employing 19 text prompts associated with social exclusion for older adults, we collected 465 images from an AI-art generator, MidJourney, and conducted quantitative and qualitative analyses to understand differences between age group depictions. We used Amazon Rekognition software to detect the numerical image properties. Results that were compared statistically (STATA and Python) revealed significant differences in attributes: images with older people exhibited lower overall brightness (M = 46.54) and sharpness (M = 58.58) compared to younger individuals (brightness: M = 60.50; sharpness: M = 72.55). Moreover, older people were more likely to be smiling (16%) and with eyeglasses (56%) compared to younger individuals (6% and 12%), all with p &lt; 0.001. Next, we conducted a qualitative semiotics analysis with a randomly selected sample of 76 images. Older people were often depicted with a strong emphasis on nostalgia, vulnerability, and aloneness, reflected in vintage-style clothing, worn objects, and worried expressions. In contrast, younger individuals were often portrayed with a focus on modernity, energy, and independence, characterized by vibrant clothing, digital technology, and confident expressions. Together, these findings demonstrate the perpetuation of negative and reductionist age scripts in AI-generated art, such as older adults’ incompetence, loneliness, fear, and outdatedness. Implications of this research extend to societal representations of aging, emphasizing the importance of challenging age-related biases in AI-generated content. Future research should explore refinement of generative models which foster more equitable outcomes.}},
  author       = {{Allen, Laura and Xu, Wenqian and Nishikitani, Mariko and Atul Patil, Vaishnavi and Hule, Sunil and Bradley, Dana}},
  issn         = {{2399-5300}},
  language     = {{eng}},
  month        = {{12}},
  note         = {{Conference Abstract}},
  number       = {{S1}},
  pages        = {{925--926}},
  publisher    = {{Oxford University Press}},
  series       = {{Innovation in Aging}},
  title        = {{Art, aging, and artificial intelligence: Exploring age scripts in the universe of ai-generated art}},
  url          = {{http://dx.doi.org/10.1093/geroni/igae098.2986}},
  doi          = {{10.1093/geroni/igae098.2986}},
  volume       = {{8}},
  year         = {{2024}},
}