D.4 Representation Health

Can I trust this representation for the question I am asking?

Trust begins by stating the correspondence a representation claims. A derived service map may describe the implementation at HEAD; an ownership model may encode intent the implementation must satisfy; a generated policy may be the authoritative source from which implementation artifacts derive. These are distinct correspondence relations.

Read four things together:

• Correspondence. Does the claimed relation still hold? Descriptive models should still match observed reality; intent-bearing models should still be satisfied where authority applies; generated artifacts should still derive from their declared source.
• Coverage and relevance. Does the representation cover the engineering surface for which it is being relied upon? Missing coverage may be legitimate when implementation lies below the model's intended grain.
• Traceability. Can the reader move from the model claim to the territory it concerns, and back where that reverse relation is meaningful?
• Freshness. Has the claimed correspondence been re-established since the territory or the authored intent last changed?

  REPRESENTATION HEALTH     "Can I trust this representation for this question?"
  ────────────────────────────────────────────────────────────────────────────
  Correspondence   claimed relation still holds
  Coverage         intended surface is represented
  Traceability     claims resolve to their subjects
  Freshness        relation re-established after change
  ────────────────────────────────────────────────────────────────────────────
  Ask first: what correspondence does this model claim?

If the claimed correspondence fails, do not rely on the representation for that reasoning. The finding does not determine whether the representation or the represented system should change.

In DocAble, the Missing-Model surface fell from 56% to 7.89% over nine passes. Appendix H records the measurement and its limitations.