Published February 3, 2026 | Version v1.0
Working paper Open

Governing Adaptive Clinical Artificial Intelligence: Structural Failure Modes, Auditability, and Infrastructure for Decision Safety

Authors/Creators

  • 1. EDMO icon Boston University

Description

This collection reflects a structured research program examining clinical artificial intelligence as an adaptive sociotechnical infrastructure requiring explicit governance constraints. The included works develop a layered analytical framework addressing structural failure modes, deployment level safety considerations, reimbursement driven behavioral incentives, and externally constrained learning architectures.

Across these contributions, the Externally Governed Learning Systems (EGLS) framework is introduced as a formal model for separating adaptive computation from institutional decision authority and viability enforcement. The materials in this collection collectively explore how governance mechanisms can be embedded at the infrastructure level to support auditability, reproducibility, and deployment safety in clinical AI systems.

The intended audience includes clinicians, health informaticians, machine learning researchers, regulators, policymakers, and institutional leaders engaged in the deployment and oversight of adaptive AI systems in healthcare.

Technical info (English)

This record represents an archival synthesis document consolidating a sequence of scholarly works addressing governance architectures for adaptive clinical artificial intelligence systems. The manuscript is structured as a program level integration of related conceptual, empirical, and implementation oriented contributions.

Several component works referenced in this synthesis have been published in peer reviewed journals or deposited as publicly accessible preprints. Others function as theoretical or methodological extensions within the broader research program. Each manuscript remains an independent scholarly contribution; the present document organizes their shared assumptions, structural relationships, and cumulative analytical scope.

The synthesis emphasizes deployment realism, auditability, and institutional constraint in clinical AI systems, focusing on failure modes that arise during operational embedding within healthcare environments. Rather than introducing new predictive models, the work examines governance mechanisms that shape adaptive learning processes, decision authority separation, and accountability structures.

The document is provided as an openly accessible working paper to support continued scholarly discussion, critique, and refinement of governance approaches in clinical artificial intelligence.

Files

Borges_2026_Governing_Adaptive_Clinical_Artificial_Intelligence_v1.pdf

Files (12.0 MB)

Additional details

Related works

Cites
Journal article: 10.34297/AJBSR.2026.30.003888 (DOI)
Is compiled by
Dataset: 10.5061/dryad.fj6q57482 (DOI)
Is supplemented by
Journal article: 10.1093/jamiaopen/ooaf177 (DOI)
Dataset: 10.5281/zenodo.18356747 (DOI)

Dates

Created
2026-02-03

Software

Development Status
Active

References

  • Borges, Julian Y. V. "Auditing Shortcut Learning and Misclassification in AI-Based Breast Cancer Genomic Subtyping." JAMIA Open (2026). https://doi.org/10.1093/jamiaopen/ooaf177
  • Borges, Julian Y. V. "Multi-Armed Bandit–Based Adaptive Model Selection for Clinical AI Governance: An Infrastructure Approach to Safer AI Use in U.S. Healthcare." Preprint (2026). https://doi.org/10.21203/rs.3.rs-8651972/v1
  • Borges, Julian Y. V. "Adaptive FHIR-Native AI Governance for Clinical Decision Support." Preprint (2026). https://doi.org/10.21203/rs.3.rs-8714776/v1
  • Borges, Julian Y. V. "Externally Governed Learning Systems: Adaptive Computation Under Viability Constraints." SSRN Working Paper (2026). https://papers.ssrn.com/abstract=6160268
  • Borges, J. (2026). "Clinical Artificial Intelligence as a Sociotechnical System: Structural Failure Modes and Governance Requirements." Available at Am J Biomed Sci & Res. 2026 30(1) AJBSR.MS.ID.003888 DOI: 10.34297/AJBSR.2026.30.003888
  • Borges, Julian Y. V. "Payer-Mediated Artificial Intelligence Governance and Reimbursement-Driven Algorithm Behavior in United States Healthcare." SSRN Working Paper, January 31, 2026. https://papers.ssrn.com/abstract=6159287