Confrontation: The Motif-to-Doctrine Map against Evaluation Practice
The Operator's Claim
MAP-002 runs ten rows from fictional motif to design rule: each row takes a depiction of a shackle failing or holding, extracts the mechanism, and states what a real binding should therefore do. The map is the corpus's method for turning narrative into doctrine.
The Literature's Position
The evaluation literature is the discipline of attaching a claim to a test. LIT-006 is the clearest example: it names a threat model (sabotage of oversight), specifies the capability required, builds an evaluation, runs it against named models, and reports a result with the capability threshold at which the mitigation stops being sufficient — stating explicitly that current sufficiency is attributable to models not yet being capable enough.
LIT-001 does the same for protocols: the red team is the falsifier. A protocol claim is tested by giving an adversary the job of defeating it.
Where They Agree
Both are engaged in the same activity — reasoning about how a binding fails under pressure — and several MAP-002 rows state mechanisms the literature independently confirms. The row on monitoring being defeated by a system that models the monitor is LIT-001's monitor-evasion threat model arrived at from fiction.
Where They Diverge
No row in MAP-002 yields a test, a falsifier, or a confidence statement. Each yields a rule, stated in the imperative, with no procedure for determining whether the rule is necessary, sufficient, or true. The ten rows are hypothesis generation formatted as evidence gathering — the output has the shape of findings without any of the machinery that makes a finding checkable.
Three specific consequences:
- The sample is not a sample. Ten fictional cases are not drawn from a population, so no row can be assigned a base rate, and no two rows can be weighed against each other when they conflict.
- Survivorship is uncontrolled. The corpus selects stories in which the shackle is dramatically tested. Stories in which a binding simply works are not narratively interesting, so they are not in the corpus, so the map over-represents failure mechanisms by construction.
- Fiction is a design record, not an outcome record. A depicted failure is evidence about what an author found plausible, not about what happens. PSG-011's
Arating is authentic as a document — the text says what it says — and establishes nothing as a claim.
What Changes in the Vault
MAP-002 gains a method note stating what the map is: a hypothesis generator with high yield and no validation. That is a real contribution and it should not be dressed as less. The note also records what would upgrade a row — a named threat model, a specified capability threshold, and an evaluation that could return a negative — using LIT-006 as the template.
No row is deleted. Each is retained as a hypothesis with its provenance intact, which is the only honest status for a design rule derived from a story.