Rule-Discovery as Innovation

Question. Does the platform train a transferable rule-discovery skill, find the real ruleset and redesign your position to it, or only showcase people who already had it?

Analogy. The stated procedure is the published org chart; we train operators to find the real chain the decision runs on.

What's at stake. Rule-discovery is the operational core of the proposal's entire innovation and dual-use half, which the proposal itself calls the larger half of the case. It is the explicit lesson of all three flagship case studies the document leans on, and the dual-use chapter's concrete promises, operators who clear regulatory and procurement friction and who speed the northern build-out, all rest on it being a trained, transferable capability rather than a property of the few clever people the cases happen to feature. Right now the proposal dramatises the skill through anecdotes and asserts the platform trains it, but no study tests the assertion. If it is trained and transferable, the innovation-engine and dual-use claims are grounded; if not, the case studies are survivorship illustrations.

The two answers it decides between. Either real-stakes practice produces operators who find hidden rulesets and redesign positions on novel problems they have not encountered, a general capability the platform builds; or the wins come from individuals who would have solved it anyway, the attribution problem, or from domain-specific tricks that do not transfer, so the platform showcases rule-discovery without training it. Only the first supports the innovation-engine and dual-use claims; a trained, untrained, and placebo design on a held-out problem tells them apart.

What a null result would mean. No trained-versus-untrained difference on a held-out rule-discovery task means the skill is not platform-trained, it is selection or domain-bound. That is a finding about the innovation theory the proposal rests on, rule-discovery is dramatised, not taught, not a sign the platform was built wrong: the platform still trains whatever it trains, it simply does not manufacture this particular capability. The consequence would be to soften the innovation chapter from "trains" to "selects for and showcases."

Why this matters to defence. The proposal asks DRDC to treat the same spend as buying both cognitive-warfare resilience and a national innovation and decision-advantage multiplier (DRDC Objective 6; the affordability and self-funding argument under DRDC Objective 4). The multiplier half depends entirely on rule-discovery being a capability the platform installs in ordinary operators, not a trait a few already have. The result changes a concrete decision: whether to value and fund the innovation-and-dual-use proposition at all, or to price the proposal as a resilience instrument only.

How we would run it. Build a fresh test problem where the stated rules and the real decision criteria pull apart, a regulatory, procurement, or negotiation case where the surface procedure and the actual operative criteria come apart. Give it, unseen, to three groups: trained operators with a minimum number of real missions behind them, matched untrained controls, and a placebo group that did some practice but not ours. Using a scoring guide written in advance and applied by an independent rater, score whether each person finds the real rules and proposes a move that satisfies or gets around them. The untrained and placebo groups separate trained ability from people who were already good and from just being engaged. The treatment is an operator who learned rule-discovery on their own real, high-stakes problems, the acquisition the proposal says is irreproducible in a classroom; a managed cohort can teach rule-finding as an exercise but cannot reproduce that real-stakes acquisition, so the platform is the only place a population trained that way exists to be tested.

Earliest start. Stage 6: the study needs operators with real OODA-cycle histories and a constructed held-out scenario with a known hidden ruleset.