Integrity Through Convergence

Integrity through convergence is the integrity mechanic by which a hackathon's result is established through agreement across independent teams using methodologically distinct paths on a shared dataset, applied in cases where no external ground truth is available on the day of judging. The mechanic shifts the integrity question from "does the answer match the truth" to "do the answers match each other across methods that should not have produced spurious agreement." The Event Horizon Telescope's four-team blind imaging protocol embodied the mechanic at canonical depth; Klein and Roodman's 2005 Annual Review of Nuclear and Particle Science paper had been arguing for the underlying logic for decades; and MacCoun and Perlmutter's 2015 Nature essay extended the case to the empirical sciences more broadly. See integrity-through-convergence.

The mechanic and the format archetype it grounds (the Blind Replication Sprint) are conceptually distinct things — the mechanic is the integrity logic, the archetype is the hackathon format that operationalizes it — and the distinction matters because the mechanic generalizes beyond the format. Any hackathon whose outputs make claims about something with no external referee — scientific measurements, contested historical facts, novel design decisions, predictions of future states — can in principle structure its evaluation around convergence even if the format is not a fully-implemented Blind Replication Sprint. The four operational conditions (team isolation, methodological diversity, pre- registered parameter sweeps, designed convergence event) are applicable in scaled-down form to events whose duration or institutional commitment cannot match the EHT's seven-week sprint. See blind-replication-sprint and blind-analysis.