Is fidelity a dial or a consequence?¶
The question, and why it is live for us¶
brief: Frame from the group's decision, not from the literature. One paragraph on what a proposal would commit to either way.
What each perspective found¶
brief: One subsection per perspective the run interviewed, each in its own voice and each standing on its own citekeys. Do not harmonise them here -- the contradictions are the next section's material, and smoothing them early is how a multi-perspective report becomes an ordinary survey.
queries:
- digital twin model fidelity
- fidelity requirement decision support
- surrogate accuracy tradeoff real time
Where the corpus contradicts itself¶
brief: The report's centre. Take the contradiction map's five parts in order -- direct contradictions, strongest vs weakest evidence, the resolving question, universal agreement, the blind spot -- and give each its own short subsection. Name the sources on both sides of each contradiction; a contradiction stated without both sides is an assertion.
claim: Papers that treat fidelity as tunable are, with one exception, reporting on models whose deployment context they do not describe.
queries:
- fidelity adjustable runtime adaptation
- model accuracy fixed requirement safety
What every perspective agreed on¶
brief: Short, and worth its own section precisely because it is boring. Agreement across five differently-prompted searches is the strongest signal this method produces.
The question that would settle it¶
claim: Whether fidelity can be tuned after deployment depends on whether the decision the twin serves is itself fixed, and no paper in this corpus reports a case where the decision changed.
brief: Phrase this as a study someone could actually run, with what it would have to measure.
The blind spot¶
brief: What none of the perspectives' searches turned up at all, and whether that is the corpus being thin or the field being silent. Say which you think it is, and mark it as judgement.
What this does and does not license¶
brief: Back to the group's decision. What the corpus supports, how firmly, and -- per scope.md -- no recommendation.