CSNM-EXT: A Longitudinal Extension for Tracing the Evolution of Human–AI Cognition

Memory systems for human–AI interaction have generally focused on preserving information relevant to maintaining contextual continuity across sessions. However, narrative continuity does not necessarily make it possible to reconstruct how ideas evolved, who introduced a given proposition, which elements were modified or discarded, what evidence produced a change, or what biases may have intervened along the way. This paper proposes CSNM-EXT, a conceptual extension of Cross-Session Narrative Memory (CSNM) oriented toward the longitudinal traceability of cognitive co-construction between a person and an artificial intelligence system. The proposed model incorporates five dimensions — provenance, temporality, transformation, validation, and attribution — and five falsifiable research hypotheses (H1–H5).