The Open Collaboration — Governing Savoir Through a Disinformation Surge

A four-round, advanced platform-governance simulation set inside the OpenSavoir Foundation, the lean (90 staff, €24M budget) non-profit behind Savoir, a free volunteer-built multilingual encyclopedia with 3.1 million articles, 210 million monthly readers, 48,000 active contributors and ~1,900 privileged reviewers. Six weeks before a contested national election, a coordinated influence operation makes 41,000 edits across 1,200 politically sensitive articles in 72 hours, collapsing the quality index on those pages from 82 to 61/100, pushing the review backlog to 14 days, and prompting the headline “The encyclopedia anyone can poison.” You have 30 days to restore quality above 78 — or the press narrative hardens, two institutional grantors (€3.8M, ~16% of budget) freeze, and the board imposes top-down editorial control, ending Savoir's self-governance experiment. Playing the foundation's crisis team, you must work through the volunteer community, not over it (Jemielniak's Collaborative Society thesis), because the foundation can build tools and set terms but cannot decree content. Round 1 — Diagnose: read the surge, name the true constraint (reviewer capacity, not edit volume), separate coordinated abuse from good-faith activity, and weight the false-positive risk of mistaking trusted editors for attackers. Round 2 — Govern: choose a moderation lever along the openness-control spectrum (light throttling, targeted pending-changes review, or full lockdown), decide who holds authority (foundation fiat, expedited community consensus, or a hybrid), and set an explicit sunset for emergency measures. Round 3 — Mobilise: split a €600k emergency budget across tooling, contractor moderators and community support; decide whether and how to wire the €220k anti-abuse classifier (augment human judgment vs auto-revert); and design a recognition incentive that rewards verified quality rather than edit count. Round 4 — Recover & sustain: deliver a credible integrity plan to the grantors on time, defend the open model to the board, and leave a standing anti-abuse governance design. The math rewards proportionate, reversible moderation governed through the community, a classifier used as an assistant, quality-based recognition and on-time stakeholder management — and measurably punishes the five classic errors: locking everything (the blunt-instrument trap), ruling by fiat (forfeiting the volunteers), over-automating (machine overreach and false-positive harm), assuming volunteer goodwill is infinite (under-reaction), and gameable edit-count incentives plus ignoring the grantor and reputation timeline. Final KPIs track the quality index on targeted pages (/100), disinformation containment (%), active contributors and review backlog (days), alongside community trust and grantor confidence.