Synthetic offline pilot. All answers came from explicitly fictional fixtures. No inference calls, Humanize publication, or invitations were executed. This validates the local workflow; a live cohort pilot remains pending.
Each mode assessed options in batches, collected a cross-batch ranking of selected choices, exported preferences, and solved the assignment. Every package was built as native EDSL Jobs, its fixture answers were validated, and its Results were saved, reloaded, and imported with respondent/package checks.
| Mode | Respondents | Native packages | Outcome |
|---|---|---|---|
| Teams and projects | 12 | 60 | Four teams of three; optimum proved |
| Peer review | 12 | 24 | 36 assignments; three reviews per essay and student; optimum proved |
Assessment surveys · Cross-batch rankings · Exported preferences · Solved result
Assessment surveys · Cross-batch rankings · Exported preferences · Solved result
Hard exclusions survive export. No-preference/insufficient-information answers remain assignable at zero score; missing responses block completion. Synthetic source labels remain synthetic.
Machine-readable pilot summary · Workflow and reproduction commands · Package guide