AI-Generated Historical Images and Colonial Bias

Recent research highlights the social risks associated with generative AI, despite its revolutionary impact on image creation and consumption. Tools like Midjourney, DALL-E, and Sora can create anything from realistic photos to oil-like paintings from a short text prompt. However, these images often reflect sexist and racist stereotypes.

The study reveals that generative AI also harbors colonial biases. When visualizing Aotearoa New Zealand's past, Sora prioritizes the European settler perspective, depicting Māori as peripheral figures. Such depictions naturalize myths of benevolent colonization and undermine Māori claims to political sovereignty and cultural revitalization.

AI tools tend to reflect 19th-century colonial painting traditions when generating historical images. For instance, when depicting New Zealand in the 1700s, Sora portrays a forested valley in golden light with Māori as ornamental details, suggesting a wilderness awaiting European discovery.

These AI-generated images obscure the continuity of the Māori self-determination movement. Globally, researchers and communities are working to decolonize AI, developing ethical frameworks that include Indigenous data sovereignty. However, visual generative AI presents unique challenges related to images that shape history and identity.

AI literacy may offer the most promising solution. By understanding how these systems operate and what data they use, AI can move beyond recycling colonial tropes to become a medium for re-seeing the past through diverse perspectives.