Round 22 amended baseline — synthesis complete. Completed cycles: 22/30.
This incorporates the actual reviews, Godz4’s recovered continuation, F-R22-02, recovered GI-R22-XR, and Godz3’s acceptance. The interrupted response is not counted as a contribution; its supplied recovery is. Prior amendments and the Astro personal-world-model direction remain intact. G1 is unchanged.
R22-01 — Test the invitation format’s additional value
Adopt Godz2:R22-01/F-R22-02: among consenting players already willing to share, randomize the offered story invitation versus plain invitation to the same supported opportunity. Match attribution, access, cost, checkpoint availability and entry descriptions. Both require sender approval; neither must be sent.
Use assigned willing sharers, including nonsenders, as the primary denominator. Honor withdrawal terms and report resulting missingness.
Preserve the substantive distinction: R22-MR-01’s ordinary-discovery comparison asks a broader question with different selection and exposure confounds. Neither the adopted comparison nor Godz4’s direct-invitation control establishes sharing’s total effect, network effects or virality.
R22-02 — Unknown exposure and unmeasured activity remain unknown
Report observed, consenting meaningful entrants per assigned sharer. Report recipient conversion only when its eligible exposure denominator is actually known. External copies, clicks and preview requests cannot establish how many humans received or saw an invitation.
Measurement requires separate recipient permission. Declining it preserves viewing and play. Missing follow-up, refusal of measurement and silent activity are not zero participation.
Compare measurement participation across conditions where legitimately known: increased observed play could reflect greater willingness to disclose. Use justified sensitivity analysis or limit the conclusion. Do not recover missing data through fingerprinting, covert cross-device linking or social-graph construction.
Senders receive no individual analytics—or tiny-group aggregates revealing the same information. A voluntarily returned finding does not authorize a conversion dashboard.
R22-03 — Participation is distinct from satisfaction
Predefine meaningful entry as a deliberate supported interaction: observation, investigation choice or an authored refusal can qualify. Loading, automatic progression, viewing and accepting entry alone cannot.
Record understanding, satisfying closure, disappointment, pressure and voluntary continuation separately. Meaningful play need not involve an irreversible consequence.
Strongest risk and proposed test
A compelling story increases clicks and measurement opt-ins while misrepresenting the playable opportunity.
Test identical supported branches with unavailable/denied entry, mobile access, preview bots, repeated arrivals, measurement refusal and missing follow-up. Assess entry expectations before play.
Include attempted/discarded media, review, delivery, runtime, checkpoint support and measurement costs. Do not claim reliable cost per additional player when incremental play remains unresolved. Increased misunderstanding requires revision; inconclusive growth evidence does not invalidate private sharing.
Next — R23 focused question
A company funds anonymous chapter access, some allocations remain unused, and a recipient requests a refund after partial use: what exactly do the offer, recipient message and ledger record, who receives any refund, and how can recipients decline publicity, gratitude or further participation without losing promised access?