Search for: All records

Hugging Face is the definitive hub for individuals and organizations coalescing around the shared goal of “democratizing” AI. While open AI draws on the ideological values of open source software (OSS), the artifacts and modes of collaboration remain fundamentally different. Nascent research on the platform has shown that a fraction of repositories account for most interactions, ambiguous licensing and governance norms prevail, and corporate actors such as Meta, Qwen, and OpenAI dominate discussions. However, the nature of model-based communities, their collaborative capacities, and the effects of these conditions on governance remain underexplored. This work empirically investigates whether models—the primary artifact in open AI ecosystems—can serve as a viable foundation for building communities and enacting governance mechanisms within the ecosystem. First, we use interaction and participation data on Hugging Face to trace collaboration and discussions surrounding models. Second, we analyze governance variations across models with regular and growing community interactions over time. We describe three phenomena: model obsolescence, nomadic communities, and persistent communities. Our findings demonstrate that the absence of robust communities hinder governance in artifact-driven ecosystems, ultimately questioning whether traditional principles of openness foundational to OS software can be effectively translated to open AI. 

Variance in predictions across different trained models is a significant, under-explored source of error in fair binary classification. In practice, the variance on some data examples is so large that decisions can be effectively arbitrary. To investigate this problem, we take an experimental approach and make four overarching contributions. We: 1) Define a metric called self-consistency, derived from variance, which we use as a proxy for measuring and reducing arbitrariness; 2) Develop an ensembling algorithm that abstains from classification when a prediction would be arbitrary; 3) Conduct the largest to-date empirical study of the role of variance (vis-a-vis self-consistency and arbitrariness) in fair binary classification; and, 4) Release a toolkit that makes the US Home Mortgage Disclosure Act (HMDA) datasets easily usable for future research. Altogether, our experiments reveal shocking insights about the reliability of conclusions on benchmark datasets. Most fair binary classification benchmarks are close-to-fair when taking into account the amount of arbitrariness present in predictions -- before we even try to apply any fairness interventions. This finding calls into question the practical utility of common algorithmic fairness methods, and in turn suggests that we should reconsider how we choose to measure fairness in binary classification.