Translating ontological distortion theory into implementable software architecture for clinical AI

Develop implementable software architectures for clinical AI systems that translate the theoretical analyses of ontological distortion—specifically the three-forces model of documentary enactment, the reification feedback loop, and terminology governance failures—into concrete engineering designs suitable for deployment in production pipelines.

Background

The paper argues that while prior theoretical work has characterized mechanisms that distort clinical data—such as the three-forces model of documentary enactment, the reification feedback loop that amplifies coding artefacts, and terminology governance failures—there is a lack of concrete, implementable software architecture to operationalize these insights in clinical AI systems.

To address this gap, the paper proposes a set of ontology-aware design patterns as a preliminary architectural vocabulary, but emphasizes that creating implementable software architecture from these theoretical insights is still unresolved at the time of writing.