Model documentation plays a crucial role in promoting responsible AI (RAI) development. The paradigm shift from traditional machine learning models to Generative AI (GenAI) models has reshaped the conditions under which documentation is produced, particularly on open-source platforms where models are hosted and shared. To investigate how this paradigm shift has manifested in developers’ documentation practices, we conducted interviews with 17 GenAI developers who document models on open-source platforms. Our findings illustrated that uncertainties have become the defining feature of developers’ GenAI documentation practices, which unfolds in three interrelated forms: (1) normative and epistemic uncertainties in determining documentation content; (2) methodological uncertainties in how to evaluate and communicate model properties; and (3) ecosystemic uncertainties in who should document. We argue that the uncertainties in GenAI documentation require coordinated interventions, including infrastructural support to address epistemic and methodological uncertainties, community-based mechanisms to cultivate RAI documentation norms, and collaboration across supply chain actors to address ecosystemic uncertainties.
ACM CHI Conference on Human Factors in Computing Systems