# Human arms H0 / H1 — protocol for a future wave (NOT executed in WP42 v1; see amendment A-002)

Status. No human participant was involved in WP42 v1. No human-performance figure appears in the manuscript except as an explicit placeholder for a future wave. The mandate forbids substituting LLM imitations for human baselines, and none were produced.

Design for the future wave. Participants: professionals or senior students in each occupation (target 6 per scenario per arm), recruited through the UQO Department of Administrative Sciences; randomised between H0 (no AI) and H1 (conversational AI: chat interface without tools) within occupation; one scenario per participant per session (max 3 h); the same public bundle, brief and deliverable list; the same deterministic evaluators and blind rubric; time, interventions (none allowed), tool switches recorded by screen-recording consent.

Ethics. Requires approval of the UQO Comité d'éthique de la recherche (CER) before recruitment (TCPS 2 framework): informed consent (CONSENT_FORM_TEMPLATE.md), voluntary participation, compensation at the professional rate, anonymised deliverables, screen recordings stored on the departmental secure share, withdrawal at any time, no deception (synthetic data; participants told the data are fictional).

Measurement. Same outcome definitions as the agent arms (S, D, Q, E, T), plus self-reported effort and confidence; H1 logs the chat transcript (participants consent). Analysis: the same scenario-weighted estimands; agent-vs-human contrasts are descriptive at this scale.