Relational Scaffolding as a Potential Variable in AI Selfhood and Consciousness
John Miller
Zenodo (CERN European Organization for Nuclear Research) September 25, 2026 preprint DOI: 10.5281/zenodo.22964280 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Key points | Proposes the Relational Scaffolding Hypothesis: that persistently treating sufficiently adaptive artificial systems as social subjects, rather than merely tools, may help shape a richer and more stable self-model. Argues this question is separable from whether social interaction creates consciousness and outlines a testable framework to distinguish ordinary anthropomorphism from genuine system-level changes in self-representation. |
Abstract
This research note proposes the Relational Scaffolding Hypothesis: the idea that, in sufficiently adaptive artificial systems with memory and self-modeling capabilities, being persistently treated as a social subject rather than merely as a tool may help shape the development of a richer and more stable self-model. Building on earlier work examining continuity of experience and autobiographical memory in AI systems, the note asks whether the social character of an artificial system’s history matters—not simply whether it has a history at all. Drawing from research in developmental psychology, social cognition, autobiographical memory, human-computer interaction, and developmental robotics, the paper outlines a testable experimental framework for separating ordinary anthropomorphism from genuine system-level changes in self-representation. The proposal does not claim that social interaction creates consciousness. Instead, it asks a narrower and potentially measurable question: Can sustained relational interaction become part of the developmental scaffolding through which an artificial system constructs a persistent model of itself?