Definition
Presenting an explanation written by one component as the reason for a decision made by another. It appears when the deciding component gives no reasons, as a classifier or typed decision model that returns only a score does not, and a language model is asked to write a rationale afterwards. The text describes a plausible decision, not the one that was made.
It matters because recourse depends on the reason being something the affected person can dispute and that would change the outcome if it were wrong. A substituted reason meets neither test, and it passes every readability check. STD-02 §1.4 requires reasons to state what the decision rested on, labels an explanation written afterwards as an account, and requires a decision object with no reasons to say so. It is a specific form of explainability decoy.
Phases: Deployment, Audit · Maturity: Core concept
Adjacent terms
References
- STD-02 Contestability and Recourse · Standard
- Typed decision models and where their authority lives · Explainer