Just-In-Time Ontology: Emergent Ontological Structure in Embedded Knowledge Graphs
Year: 2026
Type: article
Abstract: Contact: [email protected]. We document a phenomenon in which an embedded knowledge graph spontaneously generates ontological structure at every architectural layer without any layer being designed to produce ontology. In the NeuroCrystal system (83,610 biomedical concept grains across 106 corpora, embedded in 768-dimensional nomic space), ontological categories emerge from grain decomposition, coordinate geometry, traversal dynamics, and cross-domain bridging without top-down design, taxonomic input, or explicit ontological engineering. We formalize this as Just-in-Time Ontology (JiTO): structure that does not pre-exist in the system but crystallizes at the moment of query, measurement, or traversal. GitHub: github.com/mjgodat/cosine-drainage-theorem
Author: Godat, Michael
Language: English
Topic: Semantic Web and Ontologies
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204125239