Why it is a question
A8 ended: “under which decision families and resource constraints does preserving justificatory history provide additional value beyond preserving a sufficient current epistemic state?” A10 found the question mis-posed. If a state is sufficient for a family, no history can improve ideal risk on it. “So the interesting research problem is not whether history provides information beyond a sufficient state. Properly formulated, it cannot.” What remains is “Under what conditions does a state representation that is sufficient for one class of epistemic decisions cease to be sufficient when the agent’s questions, models, sources, obligations, or computational resources change?”
The hard part is the family. A10 §4: “An observation can do more than change the answer to an existing question. It can make an entirely new question intelligible.” Allow every conceivable question and lossless retention is the trivial answer. Allow only today’s and the record discards what tomorrow’s question needs. A10 proposed a robust objective over an uncertainty set of future decision distributions, and named a restricted class of transformations worth studying: invalidating a source, revising a likelihood model, introducing an alternative, changing a concept’s scope, auditing an inference.
An answer fixes such a class, states the distribution over it, and says what a retained state must keep. CLAIM-295 is the record’s candidate account of why history would be worth keeping at all.
Note of 2026-10-22: a partial answer from the Ledger of Will
The Ledger of Will (LIT-1376, read in full) answers this question for one case: its revision closure (Proposition 4) holds for a revision family declared in advance, filed as THEORY-884. It leaves the open-ended case open, and says so: “An agent cannot infer that its future concerns are exhausted by its current ontology. The response is explicit scope and retained routes to reacquisition” (§10). Its fibre criterion (Proposition 2) is CLAIM-350’s result stated independently for histories, and its “Information availability and computational accessibility are different resources” (§5.2) is one of CLAIM-295’s three values.
Note of 2026-10-27: A8’s question split off
The status review of this date split this question under ADR-049. A8’s question, whether history adds value beyond a sufficient state, and whether such a state can stop being sufficient, is QUESTION-288, Answered by CLAIM-300. This question keeps what is still open: how to judge sufficiency when the decision family changes with what the inquiry finds. The title already said only that, so it is unchanged.