Each matrix counts answers by gold label (rows) and ember's answer (columns). The outlined diagonal is agreement.
Figure 14 intent (choice).Figure 15 specific_enough (noul).Figure 16 failure_kind (choice).Figure 17 retry (noul).Figure 18 risk (score).Figure 19 needs_review (noul).Figure 20 owner (choice).Figure 21 effort (score).Figure 22 approach (choice).Figure 23 dominant_red (noul).Figure 24 dominant_colour (choice).Figure 25 alert_level (score).Figure 26 colour_changed (noul).Figure 27 domain/tooling (noul).Figure 28 layer/infra (noul).Figure 29 type/code (noul).Figure 30 type/decision (noul).Figure 31 type/design (noul).Figure 32 type/discovery (noul).Figure 33 type/moc (noul).Figure 34 type/reference (noul).Figure 35 type/session-log (noul).Figure 36 type/spec (noul).Figure 37 type/system (noul).