Cognitive Sovereignty Self-Audit for Healthcare

1. When a clinician receives a diagnostic suggestion from Epic AI or Google Health, how often do they document their own clinical reasoning before accepting the AI recommendation?

Clinicians document the AI result and mark it as reviewed without noting their independent assessment Clinicians sometimes record their own thinking but often defer to AI when it matches their first impression Clinicians routinely document their reasoning but the record shows they checked AI output rather than built their own case Clinicians document their clinical reasoning first, then compare it against AI findings and explain any disagreement in the record

2. When junior doctors or medical students learn diagnostic skills, how much time do they spend reading investigations and forming opinions before seeing what an AI tool recommends?

They learn primarily by observing what AI suggests, then understand the reasoning behind it They review cases with AI tools present from the start of their training They learn to read investigations independently but AI tools are introduced early in their workflow They complete diagnostic reasoning exercises without AI for at least the first six months, then learn to use AI as a second opinion

3. If IBM Watson Health or DeepMind AlphaFold produces a recommendation that conflicts with a clinician's assessment, what typically happens?

The clinician escalates to a senior colleague because they trust the AI output more than their own judgement The clinician documents both opinions but the AI recommendation is weighted more heavily in the final decision The clinician and AI output are treated as equally valid and a senior reviews both before deciding The clinician explains in writing why their reasoning differs from the AI and makes the final call on that basis

4. When your organisation updated patient safety protocols after introducing AI diagnostic tools, what changed in how clinicians are expected to work?

Patient safety protocols were not updated because AI tools were integrated quietly into existing workflows Protocols were updated to require clinicians to check AI outputs but not to document their own independent reasoning Protocols were updated to require documentation of both the AI finding and the clinician's assessment, but liability still rests mainly on the AI output being documented Protocols were updated to make clear that the clinician's documented reasoning is the legal and clinical foundation, and AI is an aid only if it adds to that reasoning

5. How much clinical liability training do staff receive specifically about when they must overrule or question an AI recommendation?

No specific training. Staff assume the organisation is liable if they follow the AI suggestion Generic AI governance training but nothing specific about documenting disagreement with AI findings Training on how to use AI safely but limited guidance on when and how to reject an AI recommendation in writing Specific training with real case examples of when clinicians must document independent reasoning and how to evidence that they did not defer to AI

6. In your organisation, when Microsoft Azure Health or similar tools handle administrative or triage decisions, who reviews those decisions and how often?

No review occurs. Patients are routed or prioritised based on what the AI decides Spot checks happen but most AI administrative decisions run without human review A senior clinician reviews high-risk decisions but routine administrative AI decisions are not monitored regularly All triage and routing decisions made by AI are reviewed by a clinician before they affect patient pathways, and that review is documented

7. How do patients in your care learn whether a diagnosis or treatment recommendation came primarily from a clinician's reasoning or from an AI tool?

Patients are not told. The clinician presents recommendations as their own Patients may see AI was involved in their record but there is no clear explanation of how much the recommendation relied on it Patients are told an AI tool was consulted but the clinician's role in reviewing it is not clearly explained Clinicians explain to patients how they used their own reasoning, what role AI played, and what they decided based on their clinical judgement

8. If a clinician could not access AI tools for a week, how confident would they be in making diagnostic or treatment decisions alone?

Clinicians would feel significantly less confident because they have become reliant on AI for core diagnostic reasoning Clinicians would manage but would work more slowly and second-guess their decisions without AI to confirm them Clinicians would work at normal pace but would miss the efficiency benefit of AI and feel less supported Clinicians would work independently with the same confidence and quality, treating AI as a useful check rather than a foundation

Cognitive Sovereignty Self-Audit for Healthcare

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You