Bias by Default.: Neocolonial Visual Vocabularies in AI Image Generating Design Practices.

Abstract

The growing popularity of AI image generators solidifies a novel form of co-creation and image-making while concurrently unveiling concerns regarding the representational harms engendered by these models. By examining images generated by the prompt "imagine a religious person" it becomes evident how certain aspects of the predominantly Christian patriarchal culture of the West overwhelmingly dominate the generated images. Expanding the inquiry by incorporating diverse countries into the same prompt, the comparative outcomes of the images highlight an amplification of biases combined with cultural hybridities. As a critical exercise, this research employs the concept of data colonialism to prompt reflection on how social hierarchies and linguistic ideologies shape the default visual vocabulary of these models. Consequently, the article suggests that a decolonial approach to design can broaden discussions about biases generated by these generative models in dynamic social contexts, while simultaneously offering novel insights for design practice.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 15 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.