A longitudinal pattern in which a single user's repeated working interactions with one AI system show increasing efficiency or stability over time, such as shorter prompt-to-acceptable-output paths or more reused conventions across sessions.
TREND. Tracked per user-system pair across successive sessions over a defined window (e.g. 8 weeks). Indicators logged per session: turns-to-accepted-output, count of reused user-defined conventions, and session-level task-completion rate. The construct is the slope of these indicators over the window.
Proposed measurement protocol (not yet empirically validated): Per-session telemetry over >=6 sessions: turns-to-accepted-output, reused-convention count, completion rate; trend estimated by linear mixed-effects slope per user; effect reported with 95% CI.
https://andreasehstandlicenseofclarityloc.github.io/augmanitai-periodic/#topic-relation-trust