Skip to content

Related Work: Upper Ontologies

A foundational (upper) ontology defines the most general categories of existence — endurants vs perdurants, objects vs processes vs qualities, dependence and parthood relations — before any domain. BFO, DOLCE, UFO, and GIST are the best-known. They answer "what kinds of things can exist at all."

Why arch:core is not one

Core's concepts — Element, Relationship, Viewpoint, Stakeholder, Concern, the Consideration hierarchy — are enterprise-architecture concepts, not categories of being. Core is neutral across EA frameworks, not across domains. It therefore plays the upper-ontology role in the Linked.Archi stack (the root every module imports) while being, technically, a domain core ontology aligned with ISO 42010. This mirrors how other EA ontologies (e.g. ArchiMEO) position a domain core relative to — but distinct from — a foundational layer.

The reasons for not grounding core in a foundational ontology — conflict with the minimal-OWL profile, accessibility, the open-world/SHACL mismatch, and EA-framework neutrality — are recorded in DD-4.

The debate, and where this work sits

Whether a domain ontology needs foundational grounding is genuinely contested, and the decision here is taken with both sides in view.

The foundation-first position holds that domain ontologies built without grounding accumulate modeling errors and representation gaps. Its clearest recent statement is Oliveira et al. (2023), which analyzes a real ungrounded ontology (the cybersecurity ontology D3FEND) and argues that foundational and (well-founded) core ontologies are ontology-engineering frameworks that systematically prevent such mistakes. This is the same school as the UFO-based analyses of ArchiMate (Guizzardi and colleagues). The concern is taken seriously here rather than dismissed.

The foundation-optional position holds that for most projects an upper ontology is unnecessary, and that a well-scoped domain ontology plus SHACL (for constraints and generation), SKOS (for shared vocabularies), RDF 1.2 reification (for claims and provenance), and named graphs (for context-local world assumptions) does the work that matters. This is argued, among others, by Cagle (2026), whose recommended stack — domain model + SHACL + SKOS + RDF 1.2 + named graphs — was arrived at independently here and matches the Linked.Archi architecture closely. It is cited as practitioner corroboration that this is a recognized position, not as the source of the design.

Linked.Archi takes the foundation-optional position, for the reasons recorded in DD-4. It does not ignore the foundation-first warning about modeling drift; it mitigates that risk by other means — SHACL shapes catch the structural and relationship errors that foundational disjointness would otherwise catch, ISO 42010 provides conceptual discipline and a shared backbone, and SKOS definitions and scope notes carry the meaning negotiation that foundational analysis would otherwise provide. The claim is not "grounding is unnecessary in general," but "grounding in a foundational ontology is not the only way to obtain its benefits, and for an EA-notation modeling stack the costs outweigh the gains."

UFO and ArchiMate

The most direct challenge to "why no foundational ontology?" is the body of work grounding ArchiMate in UFO (the Guizzardi and Almeida line on the ontological foundations of ArchiMate), which uses UFO to disambiguate concepts such as Capability, Service, and Role. Linked.Archi engages with this work as an input to concept definitions — where UFO analysis clarifies what a concept means, that clarity is reflected in core's definitions and SKOS scope notes — but does not import UFO's axioms, for the reasons in DD-4. UFO informs the semantics; it does not axiomatize them.

GIST

GIST is the foundational ontology closest to Linked.Archi in spirit and is examined in detail in its own note (see GIST vs Linked.Archi). In brief: unlike BFO/DOLCE it uses everyday business vocabulary rather than philosophical primitives and is minimalist in concept count (~100 classes/properties), but it is still a genuine upper ontology with deliberate DL axiomatization (fine-grained top-level disjointness for inconsistency detection). It was not adopted as a grounding because that disjointness-based reasoning is the constraint-style OWL the minimal profile declines (DD-2), and because grounding EA notations in GIST's generic primitives conflicts with Linked.Archi's notation-fidelity goal. ISO 42010 was preferred as an architecture-description standard at the right layer (DD-3).

One argument from the foundation-optional literature is deliberately not relied upon here: that LLMs have made domain-ontology construction cheap enough to erase the bootstrapping value of an upper ontology. That is empirical and contested; the decision here rests on the principled trade-offs above, not on the economics of ontology generation.

References

  • Oliveira, Í., Engelberg, G., Barcelos, P. P. F., Sales, T. P., Fumagalli, M., Baratella, R., Klein, D., & Guizzardi, G. (2023). Boosting D3FEND: Ontological Analysis and Recommendations. In Formal Ontology in Information Systems (FOIS 2023), Frontiers in Artificial Intelligence and Applications, vol. 377 (pp. 334–348). IOS Press. DOI: 10.3233/FAIA231138. (Open access, CC BY-NC.)
  • Cagle, K., & Shannon, C. (2026, May 13). Do You Need An Upper Ontology? The Ontologist. https://ontologist.substack.com/p/do-you-need-an-upper-ontology (Practitioner commentary; not peer-reviewed.)
  • ArchiMEO Enterprise Ontology — Related work on ArchiMEO's domain core with foundational grounding.
  • Azevedo, C. L. B., Almeida, J. P. A., van Sinderen, M., & Quartel, D. (2015). An Ontology-Based Well-Founded Proposal for Modeling Resources and Capabilities in ArchiMate. IEEE EDOC. ontological foundations of ArchiMate, if cited formally.]*