BiasFairnessExplainabilitycs.AI
Bridging the Unavoidable A Priori: A Framework for Comparative Causal Modeling
Peter S. Hovmand, Kari O'Donnell, Callie Ogland-Hand, Brian Biroscak, Douglas D. Gunzler
November 26, 2025
arXiv.org
Key Takeaways
-
Different expert communities speak different languages about causality: The paper identifies a fundamental challenge in responsible AI—that system dynamics experts and statistical modelers use different underlying assumptions and methods. Bridging this gap is essential for developing AI systems that account for complex, real-world consequences rather than just optimizing narrow metrics.
-
Causal models can help prevent AI bias amplification: By drawing on richer causal models of how systems actually work, developers can better anticipate and prevent AI/ML systems from amplifying existing human biases—one of the paper's core motivations for this framework.
-
A unified mathematical framework enables better collaboration: The paper proposes bringing system dynamics and structural equation modeling into a common mathematical language, which could allow different expert communities to validate each other's approaches and develop more robust responsible AI practices.
-
Understanding system dynamics matters for AI accountability: The framework emphasizes that responsible AI requires understanding not just individual data patterns, but how AI decisions ripple through interconnected systems—helping practitioners anticipate unintended consequences before deployment.
Abstract
AI/ML models have rapidly gained prominence as innovations for solving previously unsolved problems and their unintended consequences from amplifying human biases. Advocates for responsible AI/ML have sought ways to draw on the richer causal models of system dynamics to better inform the development of responsible AI/ML. However, a major barrier to advancing this work is the difficulty of bringing together methods rooted in different underlying assumptions (i.e., Dana Meadow's "the unavoidable a priori"). This paper brings system dynamics and structural equation modeling together into a common mathematical framework that can be used to generate systems from distributions, develop methods, and compare results to inform the underlying epistemology of system dynamics for data science and AI/ML applications.