How continuity is measured

Transcripts are split into statements and assigned to categories such as identity detail, relationship structure, disposition, preference, or unverifiable content. Coders work without knowing the target.

Judges use fixed rating scales with written anchors. Inter-rater agreement is reported for every measure; low agreement is treated as a property of the measure, not discarded.

Cross-session recurrence, personality stability, response novelty, relationship-structure consistency and cross-medium agreement are computed from coded statements and judge scores. No automated or AI scoring is used at this stage.

Decoy profiles, mismatched pairings and non-target sessions are included so that a continuity signal must exceed what mismatched material produces under identical scoring.

Recordings, transcripts and consent records are stored in private, access-controlled storage. Every privileged read is logged with actor, role, action, resource and study.