Explainability in Lending and Clinical Decisions: One Question, Two Regulators

A declined credit line and a deprioritised referral are both automated decisions about a person. Both require an explanation a human can give, under pressure, to someone who is upset.

A credit decision and a clinical triage decision look like different problems with different vocabularies. Both come down to the same question: can a human, under time pressure, explain to an affected person why the system produced this outcome?

What 'explainable' actually means Not a feature-importance chart, and not a description of the model architecture. It means a person can give a coherent, specific answer that the affected individual can understand, and, if it is wrong, act on.

Finance: the conduct frame In lending and insurance, explanation is tied to conduct and fair-treatment obligations. A decline must be attributable and contestable, and in many jurisdictions it is also a formal disclosure requirement. The audience is the customer and, eventually, the supervisor.

Healthcare: the safety frame In clinical settings, explanation is tied to the clinician's professional accountability. A clinician cannot safely act on a recommendation they cannot interrogate, and the record has to show the reasoning that was actually followed, not the reasoning the system would have preferred.

The shared practice For any system making or informing a decision about a person, write down the explanation path and rehearse it, including the case where the person disagrees. If producing the explanation requires a data scientist to be in the room, the system is not ready for production.

Conclusion Explainability is not a model property you can purchase. It is a capability you build, test against real scenarios, and assign to the business, and it is the part of governance that affected people will actually notice.