Key concepts

Ontological commitments are the explicit or implicit choices an ontology makes about the nature and structure of reality. These commitments fall into three levels. At the general level, TLOs may be classified as “generic” (making no real-world commitments) or truly “ontological,” and—among the latter—as lightweight or heavyweight depending on the breadth of their commitments. TLOs also differ in their subject focus, ranging from natural-language ontologies that mirror linguistic primitives (nouns, verbs) to foundational ontologies grounded in philosophical realism about material reality. Categoricity—whether a TLO aims to exhaustively enumerate every top-level category—is another dimension of general commitment.

Formal structure arises from how core relations—whole-part (mereology), type-instance, and super-sub-type—are organized vertically (shape of hierarchies) and horizontally (stratifications). Vertical features include constraints on parent-arity (single vs. multiple inheritance), boundedness of the type-instance relation (whether universals-particulars are delimited to fixed levels), stratification of type hierarchies (ranked vs. unranked), and whether relation classes (e.g., the sub-type links themselves) are treated as first-class objects. Horizontal stratifications introduce or omit distinctions such as space vs. time vs. spacetime, objects vs. locations, substances vs. properties, continuants vs. occurrents, and material vs. immaterial objects.

Universal commitments concern core philosophical choices, including the adoption of mereological theories (from minimal to general extensional mereology), the treatment of interpenetration (whether spatially coincident but ontologically distinct entities share parts), the embrace or avoidance of abstract particulars, the handling of possible vs. actual worlds, criteria of identity (extensional vs. intensional), views on time (eternalism vs. presentism), support for indexicals (the “here-and-now”), and whether relations of arbitrary arity are admitted.

Mechanisms

The assessment framework proceeds in three stages. First, candidate TLOs—from BFO, DOLCE, and Cyc to ISO standards (e.g., 15926-2, 21838-1) and domain-specific models (e.g., CIDOC, FIBO)—are collected through broad desktop research and expert recommendations. Second, a unified assessment framework is constructed by synthesizing prior ontological taxonomies, philosophical criteria for scientific good-theory (simplicity, explanatory sufficiency, fruitfulness), and formal properties of classification hierarchies. Third, each TLO is evaluated against every general, formal, and universal criterion, with results tabulated to reveal patterns of commitment.

Quantitative summaries illustrate that, among the surveyed TLOs, heavyweight ontologies more consistently document and enact broad commitments, while some generic models intentionally eschew deep ontological commitment. Formal-structural analyses show a spectrum from highly stratified (e.g., UML’s meta-model layering) to unstratified (e.g., BORO’s extensional approach), and horizontal-stratification maps reveal clusters around either unified or separated models of fundamental distinctions. Universal-commitment charts confirm that extensional identity criteria and eternalist time views predominate, whereas indexical support remains rare.

Examples

The Basic Formal Ontology (BFO) exemplifies a heavyweight foundational TLO. BFO splits reality into continuants (static, enduring particulars) and occurrents (dynamic processes), employs single-inheritance hierarchies with bounded temporal layers, and treats spatiotemporal regions as primitives despite also retaining separate space and time categories. BORO, by contrast, adopts a unifying, extensional ontology: elements are 4D spacetime extensions, types are sets of instances, and tuples capture relationships. DOLCE embodies a descriptive, natural-language ontology, anchoring its categories in human cognition and linguistic patterns—nouns map to substances, adjectives to qualities—while carefully distinguishing particulars from universals. ISO 15926-2 and the HQDM frame large-scale asset information exchange through a four-dimensional, extensional ontology optimized for data integration. UML’s meta-model, though not traditionally considered ontological, demonstrates stratified layering (model, metamodel, meta-metamodel) with limited use of higher-order “powertypes” for classification. Among generic models, Schema.org illustrates minimal ontological commitment, providing a multiple-inheritance class-property graph aimed at web data annotation rather than deep semantic realism.

References

• Kant, I. Critique of Pure Reason. London: J. M. Dent & Sons Ltd., 1964.

• Casati, R., & Varzi,. C. Parts and Places: The Structures of Spatial Representation. MIT Press, 1999.

• Smith, B., & Ceusters, W. “Ontological Realism as a Methodology for Coordinated Evolution of Scientific Ontologies.” Applied Ontology, 5(2), 139–188, 2010.

• Guarino, N., Oberle, D., & Staab, S. “What Is an Ontology?” In Handbook on Ontologies, Springer, 2009.

• Partridge, C. Business Objects: Re-engineering for Re-use. Butterworth-Heinemann, 1996.

• Fine, K. “Ontology: A Guide for the Perplexed.” In The Oxford Handbook of Metaphysics, Oxford University Press, 2003.

• Schaffer, J. “What Not to Multiply without Necessity.” Australasian Journal of Philosophy, 93(4), 644–664, 2015.

• Moltmann, F. “Abstract Objects and the Core-Periphery Distinction.” In The Bloomsbury Companion to Analytic Philosophy, 2020.

• A survey of Top-Level Ontologies - Digital Twin Hub