This paper addresses how generative AI systems encode and reproduce cultural values and
biases, focusing on AI-generated bodies as a site where such norms become visible and
affectively powerful. It develops conceptual strands from an interdisciplinary project that
designed and tested a video-based intervention to raise awareness among young people about
potential harms of AI-generated idealised body images on social media. In the video,
photorealistic images generated with text-to-image models illustrated over-idealised bodies,
while a friendly, gender-neutral avatar explained their generation and potential effects.
During acceptability testing via an online survey with university students, participants often
recognised the images as AI-generated, expressing detachment. While our study did not
empirically test the impact of awareness, previous research indicates that even when viewers
know body images are manipulated, they can nevertheless shape perceptions of what is
“normal,” “healthy,” or “attractive” (McComb and Mills 2020).
Drawing on this research, the paper argues that AI-generated bodies operate through
insidious normative influence, shaping aesthetic norms and desires even when their constructed
nature is recognised. They reflect and amplify culturally dominant ideals embedded in training
data and media ecologies. Their effects arise through repeated exposure that embeds narrow
ideals within social contexts, shaping self-evaluation and expectations without overt coercion.
The normative power of such imagery extends beyond body shape to racially biased, sexist, and
ageist and lifestyle biases. AI-generated bodies are youthful, slim, and situated in scenes of
happiness, sociability, and consumption, producing homogenised narratives that fuse physical
form with moral and social worth. Synthetic bodies thus contribute to hermeneutic and affective
injustice (Fricker 2007; Gallegos 2021) by constraining interpretive space and marginalising
alternative experiences. The paper examines how generative AI systems mirror and distort
cultural values, contributing to critical analysis of bias and normativity in AI.
Author: Paula Muhr