This paper introduces Structuralist Artificial Intelligence (SAI), a research vision that defines intelligence as the capacity to form, reorganize, and transfer relational structures from experience. Grounded in biological observations of Hebbian consolidation, synaptic pruning, and structural plasticity, the framework argues that the primary product of learning is a change in relational topology, not merely the adjustment of scalar weights over a fixed structure. The vision translates these observations into three falsifiable architectural requirements: structural formation for extracting latent patterns, topological reorganization for dynamic adaptation without catastrophic interference, and structural transfer for generalization across domains sharing relational geometry. Operationalized through the AI Implicit paradigm and evaluated via the Experience-Compressed Intelligence (ECI) framework, this programme proposes a measurable path toward artificial systems with genuine structural competence and epistemic awareness.
Authors: Momen Ghazouani
DOI: https://doi.org/10.5281/zenodo.19979509
Publish Year: 2026
Download PDF