On the Ecology of Aesthetic Bias in GenAI Development and Use
Source of Publication
Proceedings of the 31st International Symposium on Electronic/Emerging Art (ISEA2026)
Conference Year
2026
Abstract
Generative AI tools for image production and animation have been observed to tend toward homogeneity – ‘slop’. While critical attention has focused on representational and demographic biases in these systems, less has been said about the specifically aesthetic biases that constrain GenAI systems. These biases manifest in formal image qualities such as exposure, composition, color treatment, and other material aspects of digital images, yet they share many of the same technical origins as other forms of bias. This paper approaches the problem from the perspective of a practitioner working with genAI tools in animation and digital production, where aesthetic constraints quickly become practical obstacles. The central argument of this contribution is that the objective functions driving genAI development such as optimizing for accuracy, efficiency, and reliability, are fundamentally at odds with the needs of visual culture, which proceeds through outliers and exceptions rather than the ‘mean’ of a cultural distribution. Aesthetic biases accumulate across an ecology of development stages: dataset curation, training processes, reinforcement by non-experts, and inference-level conventions. Making this ecology explicit is essential, both for engineers who may not recognize aesthetic narrowing as a problem, and for artists who must develop new literacies to retain creative control.
Document Type
Conference Proceeding
Publication Date
11-6-2026
Recommended Citation
Buchanan, Andy, "On the Ecology of Aesthetic Bias in GenAI Development and Use" (2026). ISEA 2026 Proceedings. 77.
https://zuscholars.zu.ac.ae/isea2026/77