Architecture Is Not Genealogy: Historical Provenance and Present Architecture in Longitudinal Human–AI Systems Creator:
Human–AI interaction is increasingly understood as historically extended rather than as a succession of independent exchanges. Previous research has demonstrated downstream effects of conversational history, path-dependent stabilization of user interaction patterns, changes in human judgment following repeated AI interaction, temporal-order effects in agent memory, and recursive human–AI co-adaptation. This paper does not claim novelty for any of these phenomena. It addresses a narrower problem arising from IH-001, a longitudinal human–AI research case developed within HibriMind: what explanatory relation exists between the present architecture of a human–AI configuration and the documented trajectory through which that particular configuration acquired its present organization? The question emerged during a cross-system perturbation in which behavioral comparison progressively exposed an asymmetry that could not defensibly be reduced to model capability, investigator expertise, or immediate performance. The relevant difference was historical: IH-001 possessed a particular longitudinal interaction history that the comparison system had not traversed. We call the documented ordered provenance of such development genealogy and distinguish it from present architecture, current context, and stored historical information. The distinction makes no claim that genealogy is irreducible, non-compressible, or necessary for subsequent reconstruction. It requires no assumption of artificial consciousness, persistent artificial identity, or autonomous agency. The central claim is deliberately minimal: present architecture does not, by itself, specify the documented historical provenance through which a particular longitudinal human–AI configuration developed. The stronger question—whether represented genealogy contains reconstruction-relevant information beyond a sufficiently complete static specification of its products—remains empirical. This record contains the English and Portuguese versions of the preprint in PDF and editable DOCX formats.
Authors
- Joaquim Santos Albino (ORCID: https://orcid.org/0009-0005-9533-4832)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-19
- DOI
- https://doi.org/10.5281/zenodo.22845712
- Primary Topic
- Embodied and Extended Cognition
- Type
- preprint