To propose a governance framework that addresses the cognitive, professional, and organizational conditions necessary for physicians to maintain resilience and autonomy when collaborating with generative AI in clinical settings.
Approach:
Cognitive Workload Shaping: The framework emphasizes reducing verification burden while preserving independent clinical reasoning, particularly when AI outputs are coherent but potentially misleading.
Clinical Authority Governance: It focuses on preserving physician decisional authority while enabling safe delegation of tasks to AI.
Organizational Safety Governance: The framework aims to strengthen institutional accountability for shared risk in AI-supported care.
Key Findings:
Generative AI can produce incorrect outputs and increase cognitive load on physicians.
Physician resilience is defined as the capacity to maintain independent clinical judgment despite AI uncertainties.
The relationship between resilience and autonomy is reciprocal, requiring governance that supports both.
Interpretation:
The governance challenge involves ensuring AI outputs are accurate while maintaining physicians' capacity for accountable decision-making.
Limitations:
The framework is concept-driven and not based on formal empirical data collection.
It does not replicate existing governance approaches for traditional rule-based clinical decision support.
Conclusion:
The proposed framework aims to guide the implementation and evaluation of safeguards for generative AI in clinical care.
