Tangential-Output Serendipity
An episode in which an off-target or tangential AI output - content not directly matching the query - triggers an unanticipated association or idea connection in the user. The referent is the serendipitous association event arising from output 'noise'.
Operational Definition
EVENT. One occurrence is recorded when (a) an AI output segment is judged off-target relative to the user's request and (b) the user reports that this segment prompted a new association or idea. The observable markers are a rater-identified off-target segment plus a user-reported idea triggered by it.
Measurement Schema
Proposed measurement protocol (not yet empirically validated): Serendipity-episode corroboration: raters label output segments as on- or off-target to the query; users flag segments that sparked a new idea and rate its usefulness. A serendipity instance requires an off-target segment with a user spark-flag; could be reported as serendipitous-association episodes per session and their rated idea value. Rater on/off-target reliability via Cohen's kappa; association of off-target status with spark-flags via phi.