Domain Language¶
Decision language must reveal policy and uncertainty rather than make a score sound like authority.
| Term | Meaning |
|---|---|
| candidate | A bounded entity being evaluated, with identity, evidence references, metrics, flags, provenance, and confidence |
| candidate cohort | The exact comparison set to which a ranking applies |
| fingerprint | Stable identity for the contract-relevant candidate state used in an evaluation |
| evidence posture | A structured account of support, gaps, contradictions, provenance, and readiness |
| falsifier | An observation or test that would weaken or overturn a favored interpretation |
| scenario | A declared decision context with assumptions, actions, confidence, and unresolved questions |
| policy | Explicit gates, criteria, thresholds, weights, and uncertainty treatment used to reach an outcome |
| score | A policy-dependent numerical evaluation; not an intrinsic property of a candidate |
| ranking | An ordering valid for one cohort, evidence snapshot, and policy |
| hold pressure | The degree to which scenario evaluations favor delaying action |
| confidence spread | Variation in confidence across scenarios, used to expose instability |
| downgrade chain | Ordered reasons a stronger recommendation became conditional, held, or refused |
| refusal | A governed non-recommendation that names the unmet condition and minimum missing evidence |
| escalation | A requirement for human arbitration because conflict or uncertainty exceeds policy limits |
| advisory output | Decision support with no execution authority |
| enforced policy | An advisory output explicitly promoted by a named actor under a named policy |
| regret | Retrospective cost of a decision compared with an alternative after outcomes are known |
A high score is not high confidence, a top rank is not evidence truth, consensus is not certainty, and enforcement is not scientific validation. Each statement answers a different question and must retain its policy and evidence context.