Goal Formation as Cultural Evolution

Question. Does training make operators form their own goals that diverge from the crowd's, and keep generating new ones, rather than competing harder on inherited goals?

Analogy. It works like a gold rush: most diggers crowd the creek where someone already struck colour and split a take that thins as more arrive, while a few strike out, stake their own ground away from the crowd, work it while the rivalry is low, and move on to stake the next claim once the rush catches up. The study asks whether training builds that second habit, the capacity to form goals that are the operator's own and divergent from the peer group's rather than copied from it, and to keep generating new ones as earlier goals get crowded, measured as a trait of the person and not a bet on whether any one goal pays out, with the explorer told from the crank by whether the operator tests the divergent goal against real feedback rather than defending a fixed one.

What's at stake. Model a population as agents who mostly copy their goals from a prior generation that tuned those goals to its environment. The environment drifts; most agents keep competing on the inherited goal, on a peak sliding out from under them. A few mutate the goal, explore, and land somewhere better adapted to the current environment, where low rivalry yields high return until the goal is discovered and copied, at which point the population as a whole has relocated to a better-adapted goal. The explorers are the population's adaptation mechanism, and a population that only competes on inherited goals has stopped tracking its environment. This is the proposal's innovation-and-sovereignty claim stated as human performance rather than economics: the trainable capacity is self-authored divergent wanting, and its absence, goals mirrored from the environment, is exactly what strategic programmability means. Whether the platform builds this capacity is asserted across the goal-setting-agency and innovation material and never measured.

The two answers it decides between. Either goal-formation is fixed (training improves how well operators pursue and revise goals, but the goals themselves stay copied from the reference group, so divergence does not rise with training), or it is trainable (operators trained on real-stakes problems form goals that measurably diverge from their cohort's and regenerate over time, so the platform produces the population's exploration capacity, not just better exploitation of inherited goals). The discriminator is whether goal-divergence and goal-regeneration rise with training, holding goal-pursuit skill constant, so the variable is what they want, not how well they chase it.

What a null result would mean. If training does not raise goal-divergence and regeneration, the proposal's claim that it produces self-authored wanting, and therefore the sovereignty-against-programmability and bottom-up-innovation cases that rest on it, needs revision: a finding about the theory, not a platform defect, since the platform would still train goal-pursuit and goal-revision, just not goal-origin. If it does, the platform builds the population's adaptation mechanism, which is the strongest form of the national-multiplier and cognitive-sovereignty argument.

Why this matters to defence. A population whose goals are mirrored from its environment is steerable: an adversary need not win an argument if it can keep the obvious goals salient and let the population compete itself into exhaustion on contested ground (DRDC Objective 6, the cognitive level; the offensive twin of divide-and-conquer). A population that forms its own divergent goals is both harder to steer and the source of the adaptation and innovation a country needs as its environment shifts (DRDC Objective 2; the leadership and innovation shortage the Impact chapter names). It changes a concrete decision: whether self-authored goal-formation is a trainable readiness capacity worth instrumenting and developing, rather than an assumed trait.

How we would run it. Using the consented researcher role, classify operators' formed goals over time on two axes: origin (self-authored versus adopted or mimicked from the reference group) and divergence (within-cohort overlap, measured relative to the operator's own peer set, never against a global population want-map, which would be the targeting-map liability). Compare trained operators against matched untrained and placebo-trained controls, holding goal-pursuit skill constant so the variable is goal-origin, not execution. The crank and the explorer both want what the crowd does not, so divergence alone scores them identically; what separates them is observable without judging the goal's value, the explorer iterates the divergent goal against real feedback while the crank defends a fixed bet against it, which is the contrarian-and-tested-against-the-world test made operational. Track regeneration longitudinally: does the operator move to a new divergent goal as an earlier one is discovered and crowds, or settle and compete. The only-us condition is a population working real-stakes life problems with naturalistic goal-formation over time, which a managed cohort cannot reproduce.

Earliest start. Stage 8: the study needs longitudinal goal-formation data from the researcher role across a maturing cohort.