ARTIFICIAL INTELLIGENCE AND INDIGENOUS (SELF-)REPRESENTATION: A MODEL FOR AGENCY AND AUTONOMY

Abstract

The rise of generative AI (Gen-AI) models in cultural production has enabled broader access to image-making, particularly for underrepresented communities. Tools like MidJourney, DALL-E, and Stable Diffusion allow users with little technical expertise to create images from text prompts. However, these models reproduce biases embedded in their training datasets, raising concerns about representation, visibility politics, and data exploitation. This paper investigates these issues through the lens of Indigenous self-representation in Abya Yala (the Americas), and draws from initial findings from a project involving Indigenous artists from South America that explored Gen-AI’s limitations and its potential for greater autonomy in self-representation. The project identified that Gen-AI often reinforced stereotypes but also provided participants with a sense of control over their images. To address these limitations, the team developed a custom Indigenous Gen-AI model emphasizing autonomy and ethical data use. The model, based on Stable Diffusion, integrates Indigenous design elements and operates offline to ensure data sovereignty. The project also emphasizes a co-creative process involving scholars, Indigenous artists, and technologists. The initiative aligns with the principles of "Digital Buen Vivir," advocating for ethical, sustainable, and community-driven technology use. While the model empowers Indigenous creators some challenges remain, including technological accessibility, digital literacy gaps, and the structural constraints of Gen-AI. We conclude with recommendations for Indigenous engagement with AI and the future development of self-representational Gen-AI technologies.