Synthetic Players

ProtocolFinal

Units, estimands, and what marginal checks cannot identify

Deployment → persona prompt → condition → episode → seat → round → request; four estimand families kept distinct.

Sources: hierarchy.md, propositions.md · Paper §3, §4.3

Unit hierarchyIdentification propositions

The unit chain is deployment → explicit persona prompt (16 + bare control) → condition (6 cells, 160 arms) → episode (1,712 in Phase 5) → seat (2/episode) → round (54,276) → provider request (36,251). These nouns are not interchangeable; calls are operational scale, not subjects.

Four estimand families

FamilyDefinitionStatus in this program
Fixed-panelproperties of these sixteen sealed promptsall registered Phase 5 predicates attach here
Persona-generatora wider persona populationexploratory at n=16; two-stage bootstrap targets it
Prompt-indexedΔi for the same explicit string across conditionsidentified; a latent person’s effect is not (invariance untested)
Human-substitutionhuman–LLM equivalencenot claimed; references protocol-nonmatched

Identification propositions

Proposition A — accepted bands [ℓ₀,u₀], [ℓ₁,u₁] identify only Δ ∈ [ℓ₁−u₀, u₁−ℓ₀]: the interval can contain zero, an attenuated effect, or the wrong sign. Exact mean matching would force the aggregate effect by identity — the failure lives in the slack of coarse criteria. Proposition B — mean and variance identify neither the between/within split, the shape, boundary mass, nor the cross-condition coupling that defines prompt-indexed response (a Fréchet–Hoeffding/Sklar application, not a new theorem).