When Does Experience Become Development? A Developmental Uptake Criterion for Persistent AI Systems
Year: 2026
Type: preprint
Abstract: Persistent AI systems can preserve records, retrieve memories, and change later behavior, but these capacities do not by themselves show that an experience has altered how the system subsequently forms states and organizes action. We distinguish historical shaping, defined as any persistent cross-episode causal influence, from Developmental Uptake, a stronger phenomenon in which an experience becomes a recombinable structure that remains sensitive to current evidence and constraints within a declared boundary and scope. The proposal centers on the distinction between a history's propositional content and its current action-relevant standing. We introduce the Developmental Uptake Criterion (DUC), a candidate joint diagnostic and evidentiary protocol that tests: (D1) selective sensitivity to current standing; (D2) invariance across standing-equivalent surface histories; (D3) revision of an established influence under later defeaters or revocation; and (D4) structurally relevant transfer beyond trajectory replay within scope. The joint effect must also be causally attributed to the target experience through rollback, ablation, or a functionally equivalent intervention. DUC is architecture-neutral and does not presuppose improved performance. A pass supports only bounded system-level functional uptake; it does not establish internal representational change, understanding, subjecthood, or AGI. The conditions are a conceptual proposal and have not been shown to be minimal, independent, or universally necessary. This is a conceptual diagnostic framework and evaluation agenda. It reports no empirical results and has not been peer reviewed.
Author: Lv, Pascal
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7170089161