Answers are grounded in this paper and cite their sources.
Randomized trial examines the dynamics of position amplification in social hierarchies, suggesting limitations on capability identification.
Trajectories of positions are widely used to argue that a hierarchy was built by amplification: the early ordering predicts the final one better than diffusion allows, and the excess is read as compounding. We show that this inference fails. Entities that merely differ in a persistent trait produce the same excess at zero gain, and that null carries a free parameter tunable to any lock-in speed: magnitude identifies nothing, only a within-system contrast does. Inside a falsifiable normal form, a multiplicative cascade amplifying position, we separate what a trajectory settles from what it cannot. The dynamics are identifiable: a necessity theorem, a non-monotone dispersion-maximising gain, a Hill-verified closed-form Kesten tail index tying tail heaviness to contestability, and a sqrt(tau) early-lead law. The seed is not: the persistent input is recovered, but its decomposition into capability and position is not a function of the trajectory. The obstruction is the amplified form of one panel econometrics knows, state dependence against unobserved heterogeneity, exact here. This is observational equivalence, not imprecision, so no quantity of data bears on it; the split must come from another observable, and what an exogenous handle returns is a proportion, the manufactured share R = 1 - corr^2, not a verdict. Both points bite: a within-Wikipedia revocation contrast supplies the capability-free test (delta-rho = 0.11, 95% CI [0.09, 0.13]; a third of it heterogeneity by placebo), and an online-chess cohort with every appearance of manufactured order is at least ~62% revealed skill (R ~ 0.38, an upper bound).
No takes yet. Share an insight, caveat, or question.
Kristian Sestak (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: