Framework
I judge healthcare AI decisions on three things: explainable, challengeable, and correctable.
Three questions I ask before a healthcare team relies on an algorithmic or automated decision.
Three tests
Explainable
A person — clinician, compliance officer, or patient — can see what the system considered and what it recommended. Not a full model explanation; enough to understand the basis for action.
Challengeable
The recommendation can be overridden, escalated, or sent back for review without halting care or creating a compliance incident.
Correctable
When the evidence, policy, or model changes, the decision can be updated and the record corrected. Someone owns that correction.
Practices that make it real
- Who owns the recommendation, the override, and the correction.
- Escalation paths designed before the edge case hits.
- Evidence and assumptions visible enough to inspect.
- Records that help during audit, not only after a failure.
- The unhappy path tested with the same care as the happy path.
How I use this
It fits alongside NIST AI RMF and existing compliance programs, but it starts with the people who have to trust or challenge the recommendation.
Decision-readiness checklist
A one-page checklist I use to assess whether an AI/system decision is ready to rely on — or what still needs governance work.
Request the checklist →