Delphi Consensus Engine
Structured, multi-round expert consensus is the front door to every published instrument on Opendelphi — not an add-on.
Open Round 0 by Default
Rounds open with experts generating items, not rating AI suggestions. In Round 0, panelists propose what should be collected in their own words; AI only clusters and organizes the submissions. This anti-anchoring default keeps the consensus genuinely expert-driven.
Multi-Round Convergence
In each subsequent round, panelists review aggregated feedback, rate items, and refine their positions. The round lifecycle is managed durably by the platform — opening, closing, and carrying results forward — so the moderator runs the study, not the logistics.
Statistical Convergence Detection
Agreement is measured, not asserted: per-item interquartile range and panel-level Kendall's W quantify how close the panel is to consensus, round over round. Consensus is declared against explicit statistical thresholds, and the record of how it was reached is preserved.
Anonymous by Architecture
Panelist identities are structurally separated from their responses — anonymity isn't a display setting, it's enforced in the data model. That prevents authority bias and groupthink while every round, rating, and rationale remains fully documented for the study record.