Formal Ontology · BFO · OWL DL · Reasoner-Validated
Four connected pieces of work, each grounded in the Basic Formal Ontology and validated against an OWL DL reasoner: BFO-Agent, which populates an ontology from source text and commits only assertions the reasoner accepts; the FOL / OWL Tester, which finds the consistency and coherence faults that pass silently in standard editors; A Structural Ontology of the Law, a 6,900-class formalization of legal structure; and a public library of validated ontologies. What each one does, and what it is for, follows.
§ 01 — BFO-Agent
BFO-Agent separates proposal from commitment. A language model, LoRA fine-tuned for extraction, reads a source text and proposes candidate axioms. Each candidate is evaluated against the BFO upper ontology and tested for logical consistency and entailment by an OWL DL reasoner. Axioms that pass are committed to the knowledge base with a provenance link back to the source span that produced them. Axioms that fail are logged and never enter the ontology.
The agent is reflexive: it evaluates its own proposed commitments rather than treating its output as authoritative, which is the mechanism by which it declines instead of inventing. The refusal behavior is a property of the pipeline, not of the prompt.
Measured: 93.3% refusal of unfounded assertions under the gate vs. 10% at the ungated baseline, on identical inputs, across SOoL, Spinoza’s Ethics, and the Leibniz corpus. Reported in “Confabulation is Architectural” (under review, Minds and Machines).
§ 02 — FOL / OWL Tester
An OWL ontology can look correct in an editor and still reason wrong. The Tester loads an OWL or RDF ontology and runs three checks. Consistency: a description-logic test of whether the axioms admit any model at all. Coherence: detection of category errors against the BFO backbone, including partition-straddle, where a class is entailed to be both a continuant and an occurrent, violating BFO’s top-level disjointness. Entailment: a first-order test of what the axioms actually imply, as opposed to what the author intended.
Partition-straddle is the representative case. It produces no warning in Protégé or most editors, yet it corrupts every inference that touches the affected class. As a concrete instance, the Tester flagged an aerospace ontology in which Force, Weight, Drag, and Lift were each entailed to straddle the continuant/occurrent partition.
§ 03 — A Structural Ontology of the Law
The largest application of the method. Law is formalized not as a body of language but as a structured region of social reality: roles, authorities, recognitions, obligations, powers, and the specific ways each can fail. The released ontology runs to nearly 6,900 classes, BFO-aligned and validated for consistency with the HermiT reasoner.
It supplies three constructs the rest of the work builds on: the eight-node Minimum Legal Chain, which a legal scenario can be walked through to test for closure; a thirteen-type contradiction typology; and the Irreversible Accountability Test. A legal system fails, on this account, when recognitions misalign or obligations lose their correlates, producing contradiction debt that has to be repaired.
Forthcoming · Palgrave Macmillan · 2026
§ 04 — The Ontology Library
Each ontology renders a body of thought into BFO and is validated for consistency with the HermiT reasoner. All are browsable, citable, and free, and each was built with BFO-Agent. Each one doubles as a reusable BFO module and as a worked reference for how a given source maps onto a formal upper ontology.
Law as a structured region of social reality: norms, powers, roles, obligations, persons, institutions. Derived from the forthcoming Palgrave book.
A formal ontology of Spinoza’s Ethics from the Elwes translation, with per-class provenance linking every concept to the proposition that produced it.
Drawn from Duncan’s 1890 anthology: the Monadology, the correspondence with Clarke, the New System of Nature, and more.
The structural skeleton of the 2025 book, made inspectable: directedness, recognition, fidelity, rupture.
§ 05 — Custom Ontologies
The same method that produced the library applies to any domain. The process is fixed and reasoner-checked at every stage. Analysis: identify the continuants, occurrents, roles, dispositions, and relations the domain actually contains. Alignment: map them onto the BFO upper ontology so the model inherits a principled backbone rather than an ad hoc hierarchy. Formalization: express the result in OWL DL. Validation: run consistency and coherence checks until the reasoner accepts the model. Provenance and documentation: deliver it as a modular, traceable, queryable artifact.
The result is not a diagram or a glossary. It is a logical model a reasoner has accepted, which is what makes the value below available rather than aspirational.
Value — what a validated ontology buys
§ 06 — Consulting
I take on formal-ontology and ontology-grounded-AI work for research groups, standards efforts, and teams whose data models have outgrown the schema they started in. Trained in formal ontology under Barry Smith, one of the founders of BFO. Engagements take one of these forms, and most combine several.
Build a domain ontology from analysis through BFO-aligned, reasoner-validated OWL, delivered with provenance and documentation.
Map an existing ontology to BFO and report where it straddles partitions, contradicts itself, or fails to entail what it claims.
Run a live ontology through consistency and partition checks and return a defect report scoped to what to fix and why.
Build the validators and reasoner-gated pipelines, on the BFO-Agent pattern, that keep unfounded assertions out of a knowledge base.