Conference Presentation at DHCH 2025

Conference: DHCH 2025 Date: 17–18 June 2025 Location: Istituto Svizzero di Roma

---

Revealing Fault Lines of Our Visual Culture: The Implications of Text-to-Image Model Personalization

The rise of open-source text-to-image (TTI) models has transformed AI-generated visual content, making powerful generative AI tools widely accessible. With recent advancements, users can personalize large open-source models to suit specific creative goals, expanding artistic possibilities in unprecedented ways.

However, TTI model personalization practices raise complex ethical concerns, including the proliferation of non-consensual deepfakes and the amplification of societal biases embedded in models. Central to this shift is the growing ecosystem of model-sharing platforms, often framed as democratizing AI but primarily shaped by engagement metrics and commercialization. These incentives influence the types of content created and shared, frequently prioritizing sensational or controversial content over equitable representation.

Our study presents an exploratory sociotechnical analysis of CivitAI, the most active platform for sharing and developing open-source TTI personalizations. Drawing on a dataset of over 40 million user-generated images and more than 230,000 models, we identify systemic patterns in visual output, model visibility, and user behavior. By examining model creation pipelines, we analyze how discrimination can be introduced and amplified through tools and workflows in open-source TTI personalization.

This research contributes to a critical interdisciplinary understanding of computational bias in open-source generative AI and proposes proactive approaches to mitigate downstream harm and support more equitable and responsible development.