Commercializing Ontology: An Interview with John Beverley and Barry Smith
This piece extends a discussion with Barry Smith and John Beverley on the future role of applied ontology inside and outside the university. Barry Smith is Distinguished Julian Park Professor of Philosophy and Professor of Biomedical Informatics and Computer Science and Engineering at the University at Buffalo (UB). Barry is also a co-author of Basic Formal Ontology (BFO), now ratified as ISO/IEC 21838-2 — the first piece of philosophy to be declared an international standard. Joining him again is John Beverley, Assistant Professor in the UB Department of Philosophy and President of the National Center for Ontological Research Inc., a non-profit that provides validation and credentialing services for ontologies and ontology-based systems. John’s work has helped push applied ontology further into the worlds of data science, artificial intelligence and defense.
In my first interview with Barry and John we discussed new private and public sector jobs in the burgeoning ontology sector. We drew attention to the increasing commercial and industrial applications of ontology across diverse industries. We also explored the evolution of the field and how philosophers are ideally suited for developing positions, employing a variety of skills in logic and computing on the one hand, and in the design of classification systems and creation of definitions on the other.
In our October 2024 conversation, Barry and John described how applied ontology has matured from a niche research specialty into a high-demand profession attracting talent from philosophy, computer science, biology and the defense community. Since then, the landscape has continued to evolve: the role of large language models in enterprise data has deepened and AI productivity tools have begun reshaping academic and applied work. UB is also now, in addition to its Master of Science program in applied ontology, advancing a first-of-its-kind PhD program in this field. We are here to take stock of all of these developments
****
Field Update: What Has Changed Since October 2024?
Charlie: Barry and John, when we last spoke in the autumn of 2024, you described an ontology ecosystem experiencing explosive growth — new partners, BFO adoption across the defense industry and international community, and a sharp rise in commercial hiring. Now approaching the midpoint of 2026, I would like you to brief the philosophical community on how things look today. What are the most significant new developments in the field — whether in terms of institutional adoption, new partner organizations, theoretical advances, or the job market — and where do you see the most momentum heading into the second half of the year?
JB: Since our last conversation, three areas of momentum have become especially clear.
First, ontology is spreading in mission-critical domains. We see this in health care, behavioral and social science, defense, intelligence, homeland security, manufacturing, and enterprise AI. In health care and biomedical research, the need is often to make research data comparable across projects and institutions. In behavioral and social science, the challenge often turns on the fact that key terms – loneliness, solitude, social isolation, identity, intervention, motivation, flourishing – are used differently across disciplines. In defense and intelligence, the challenge is that decisions depend on data moving seamlessly and reliably across systems, agencies, classification regimes, and operational contexts. In each case, the practical problem remains the same; unless the meaning of the data is explicit, reusable, and governed, the institution cannot know what its systems are really saying. The trick is to find some common way of addressing this problem, a way that will be shared across many scientific domains and many organizations.
Second, ontology engineering is becoming more important because of AI. Large language models can summarize, draft, classify, and generate plausible answers at remarkable speed. But fluency is not semantic correctness. A model can produce a beautiful paragraph while collapsing distinctions that matter. It can say that two datasets are aligned when they are only superficially similar. It can generate a taxonomy that looks tidy but hides category mistakes. AI remains in all such cases on the level of syntactic patterns. The more organizations rely on AI, the more they need an independent semantic layer: some way of saying what the system is supposed to mean, what it is supposed to be about, what distinctions it must preserve, what counts as a valid inference, and what would count as an error.
Third, ontology is becoming a professional skill set. Under the previous cottage industry model an ontology for a domain X would be built by an expert in X with little knowledge of the wider ontology landscape. Any usefulness the ontology initially enjoyed would then be lost as soon as it was applied to data deriving from domains outside X. Approaches of this sort cannot meet the market demand for ontologists. The field now needs teachable methods, reviewable standards, repeatable workflows, and people trained to work across philosophy, logic, computer science, data engineering, and domain science. This is why we have established the MS and PhD graduate programs in applied ontology at UB’s Department of Philosophy, developing a range of skills in ontology engineering as the tradecraft of an applied ontologist. The field needs ontologies. But it also needs personnel skilled in the disciplined practices involved in building, testing, revising, documenting, governing ontologies, and also in ensuring that the ontologies are not only used but also used properly and consistently.
The story since 2024 is thus that ontology is becoming part of the infrastructure of modern work with knowledge, information and data. It is one of the places where philosophy is re-entering public life, not through op-eds or ethics panels, but through standards, data systems, AI governance, and the executable architecture underwriting decision-making.
BS: What has changed since our last interview is not merely that ontology has continued to grow, but that its importance has become more visible across a wider range of institutions. We are seeing increasing recognition that if data are to be shared, reused, and interpreted consistently across different communities, then there must be some common understanding of what the data are about. This is precisely where ontology enters. The issue is not simply one of terminology. It is a matter of categories, definitions, and relations, and of ensuring that these are treated in ways that support both human understanding and machine processing.
One especially important development has been the continuing expansion of ontology into the biomedical sphere. At UB we are currently involved in substantial work relating to aging, translational science, and Temporomandibular Disorders. In connection with the latter two, I am working with colleagues in the UB Dental School and with the National Library of Medicine on questions surrounding what are called Common Data Elements, or CDEs. These are units of data relating, for example, to pain, sleep, anxiety, and depression, which ought to be collected in consistent ways across different clinical studies. But in practice different groups have often introduced their own competing formulations, thereby undermining the very comparability that CDEs are supposed to secure. This is a good illustration of the sort of problem ontologists are increasingly called upon to address.
More generally, we are seeing a similar pattern in many domains. Organizations have large quantities of data; they want to combine these data; but they cannot do so effectively because they are using different ways of describing what they have collected. Ontology is needed not because it is fashionable, but because without it there is no stable framework for interoperability. This applies in health care, manufacturing, government, defense, and many other areas. It is one reason why the field has moved so rapidly from a relatively specialized research area into something with genuine strategic importance.
A second important development concerns the broader intellectual setting. Recent discussion of AI has made it easier for people to see why formal ontology matters. Statistical systems can generate useful outputs, but they do not by themselves provide an account of the kinds of entities in a domain, of the relations between them, or of what inferences are justified. For that one needs an explicit semantic framework. In this respect the current AI moment has not displaced ontology; rather, it has revealed more clearly why ontology is needed.
So the main story since 2024 is that ontology is becoming more deeply embedded in the infrastructure of serious knowledge work. It is no longer confined to a small circle of specialists. It is increasingly visible wherever institutions need disciplined treatment of meaning, especially when they want their data and systems to work together across scientific, technical, or administrative boundaries.
The Double-Edged Sword: How AI Is Reshaping Ontological Work
Charlie: You and John made a compelling distinction in our last conversation between the statistical black-box architecture of large language models and the proof-generating, justification-rich architecture of ontology-based systems. That contrast has only grown sharper as AI tools have proliferated. I would like to explore this from two angles. First, on the productivity side: are you and your colleagues using AI tools — whether LLMs, code co-pilots, or knowledge-graph assistants — in your own research and development workflows? If so, how? Second, and perhaps more interesting philosophically: do you see AI-generated outputs creating new challenges for the integrity of ontology development — whether in terms of definitions, shortcut taxonomy-building, or simply the pressure to move faster than rigorous ontological methods allow? Put simply: where is AI helping, and where is it a source of concern?
JB: AI tools are genuinely useful. We use them, and it would be silly not to. They can help draft code, summarize and, to a degree, validity-check documents, identify candidate ontology terms, suggest mappings, produce first-pass definitions, generate critical questions, and help teams move more quickly through the early stages of a project. There is, however, a danger in relying too much on AI for ontology engineering itself. Ontology involves disciplined representations of meaning. The central task is to decide what kind of thing is being defined, how it relates to other things, what distinctions must be preserved, and what logically follows if we represent it one way rather than another. A polished definition can still be wrong. Formal rigor is key to avoiding confusions over a process with its result, a role with the person who bears it, a plan with its execution, or an information artifact with the thing the artifact is about.
AI is both an accelerant and a stress test. It accelerates parts of the workflow, but it also exposes how much semantic discipline institutions lack. Many organizations now want to combine language models, knowledge graphs, dashboards, data lakes, retrieval systems, and analytic workflows but they regularly lack clear answers to very basic questions such as: what is the authoritative meaning of the terms being used? If an AI system says that two fields match, what makes that true? If it summarizes a policy, has it preserved the operative distinctions, or merely produced a readable paraphrase?
Recent discussions around companies like Palantir are instructive here, because they illustrate a broader issue. Palantir describes itself in terms of ontology, and its platforms may be very useful for operational data integration, workflow management, analytics, and decision support. But an operational platform is not the same thing as an authoritative semantic model. A platform can help users move data, visualize data, and act on data, but it can do all of this while still failing to preserve the formal meaning of the model that the government or institution needs to govern. Moreover, two systems can on this basis exchange the same data and still misunderstand each other. Conversely, two systems can use different internal technologies and still interoperate if they preserve the same semantic commitments.
For philosophers, this should be a familiar point in a new setting. The issue is not merely interpreting “data.” It is about tracking, reference, identity, classification, inference, context, and interpretation. Those are philosophical topics. What is new is that they now show up in procurement decisions, enterprise architectures, battlefield systems, clinical research platforms, and AI pipelines.
So AI is helping ontology work by making some tasks faster. But it is also making ontology more necessary. The more powerful our machines become at producing language, the more important it becomes to ask what that language means, how we know, and whether the meaning can be preserved across systems. That is not a problem AI has solved; it is one of the problems AI has made impossible to ignore.
A World First: The New PhD Program in Applied Ontology at UB
Charlie: In our October 2024 conversation, you mentioned that the Department of Philosophy at UB had recently created a graduate-level Applied Ontology track — and that UB would soon be home to the world’s first academic program in applied ontology. I understand that a new PhD program is now moving through the New York State approval process. Can you describe the program — its structure, the coursework, the skills it will confer, and the kinds of careers its graduates will be positioned to pursue? And what does it mean, institutionally and symbolically, for a philosophy department to be the home of what is in many respects an applied data-science credential?
BS: We have now had one full semester of teaching within the framework of our new Applied Ontology Master of Science program. We have also already received approval from the New York State Department of Education to initiate our PhD program. The demand signal for education of this sort of clear. Search any major employment website and you’re likely to find multiple occurrences of the word “ontology.” Google Jobs, for example, returns more than 200 openings when ‘ontology’ is used as a search term. But searching for universities that offer degrees in ontology will return a single result: the University at Buffalo. Many of the listed openings offer salaries in the low 6 figures.
I like to view our Applied Ontology program as resting on three pillars: logic, formal ontology, and what we might call ‘philosophical exploration’. As concerns logic, what the students learn will overlap almost completely with what all philosophy students with a logical bent will learn, though extended via treatment of the sorts of logic used in the ontology sphere.
Formal ontology of course draws its roots from a much older tradition stretching back to Aristotle’s table of categories and to Linnaeus’s classifications of species and diseases. We expect our graduates to have some familiarity with this tradition, and also with contemporary approaches to ontology under the heading of ‘analytic metaphysics’, here addressing the role played by issues such as ontological commitment, scientific realism, process philosophy, modal realism, four dimensionalism, use-mention confusions, truth and reference, vagueness, and many more. We expect familiarity also with the role played by formal ontology in early GOFAI (for ‘Good Old-Fashioned AI’) era of Artificial Intelligence research, above all through the work of Patrick Hayes and his associates in Stanford. Modern discussions of what is called ‘neurosymbolic AI’ draw in part on the logical, GOFAI background.
As concerns the third pillar of ‘philosophical exploration’, this relates not to the content of what our students learn, but rather to the ways they are trained in extending this content and in broadening its application into areas outside philosophy as described by John in the foregoing. For as he makes clear, the methods of formal ontology, are today constantly being extended into new areas. Philosophy exploration involves additionally interaction with communities with existing ontologies but which are, for one reason or another, in need of external review. Here philosophical argument comes into play. Our students might, for example, become involved where multiple factions within a given scientific community have different views as to how specific terms need to be defined (and thus how these terms will be positioned in the resultant ontological structure). Our students will learn how to they can help to resolve such arguments, for example by drawing on precedents in other fields or by drawing on their training in the logic of definitions. In all of this they will need to draw on the argumentative skills that students of philosophy need to acquire, but where they will be learned and applied in arguing not about, for example, hyper-grounding or paraconsistency or cosmic panpsychism, but rather about vaccine adverse events, or corporate personhood, or the proper ontological treatment of nation states.
Our program at UB is currently a partner in several large grants in the biomedical sphere, covering 1. aging, 2. translational science (meaning: finding ways to harvest the results of bench biology research in patient care) and 3. Temporomandibular Disorders (TMDs). Under 2. and 3. I am working with colleagues in the UB Dental School and with the National Library of Medicine to revise the current treatment of what are called ‘Common Data Elements’ or ‘CDEs’, The latter are units of data (relating for example to pain, sleep, anxiety, and depression) that should be collected in identical ways in different clinical trials. Unfortunately different research groups have thus far been allowed to submit their own lists of CDEs, thereby defeating the very goal of collecting CDEs in the first place. This illustrates one characteristic way in which ontologists are called upon to help: organizations of all kinds are collecting data; often they need to combine their data; but they cannot do so because they are each using different ways to describe the data they have been collecting. Philosophy here plays several roles: first as the moorage of logic; second as a place where issues of categorization and definition are addressed scientifically; and third as training ground for argument and persuasion – skills often required not only in ensuring development of robust ontologies, but also achieving adoption of ontology solutions by multiple different types of organizations.
JB: The program itself is an attempt to educate students in a hybrid discipline: part philosophy, part logic, part knowledge representation, part data science, part software practice, and part institutional governance. Students need to understand what categories are, how definitions work, how relations structure a domain. They need to know how formal representations support inference, and how ontologies are actually built and maintained in modern technical environments.
That combination is unusual, but it reflects the reality of the work we do. A good applied ontologist has to be able to sit with domain experts and ask clarifying questions. They have to notice when one term is being used in several ways. They have to translate informal expert knowledge into explicit design patterns. They have to understand enough logic to see the consequences of a representation. They have to understand enough technology to work with knowledge graphs, semantic web standards, databases, version control, validation tools, and AI-assisted workflows. And they have to be able to explain and defend their decisions to stakeholders who may not care about philosophy, but who very much care whether their systems work.
The first semester has been encouraging. The program enrolled 28 students, including students dialing in from outside the United States and several from the Washington, D.C. area. The audience includes working professionals, people in government and defense, people in health care and informatics, and people who see that ontology is becoming a marketable and increasingly necessary skill.
The PhD program deepens this model by training people to engage in research that will advance the field and to develop better modeling methods, better standards, better alignment strategies, better AI-evaluation frameworks, and better accounts of how formal ontologies should be governed. It will prepare students for roles in academia, government, industry, research institutes, standards organizations, health care, defense, intelligence, manufacturing, and enterprise technology.
Charlie: Finally, a follow-up. Will the program be open to students from outside philosophy — computer scientists, data engineers, biologists, intelligence professionals? And what is the expected demand — is UB already fielding inquiries from prospective students or institutional sponsors?
BS: In 2006 I gave a talk to my colleagues here in Buffalo with the title “Why I am no longer a philosopher”. Instead of an abstract for this talk I used an image a shining new cruise liner named “HMS Ontology” as it was leaving an old, decrepit Mediterranean harbor. The image was designed to communicate the message that ontology – following in the footsteps of the natural sciences, psychology, and many other disciplines – was now freeing itself from its roots in philosophy, and becoming a discipline in its own right. This has, in practice, only manifested in part. Our new Applied Ontology graduate program is founded on logic and parts of the philosophy of science which are recognizably parts of philosophy as traditionally conceived.
The answer, then, is yes: the program must be open to students from outside philosophy. Indeed, this follows from the very nature of applied ontology itself. Ontologies are built for use in domains such as biology, medicine, manufacturing, intelligence analysis, and administration. They must therefore be shaped in constant interaction with people who possess deep domain expertise, as well as with those who understand the technical environments in which ontologies will be deployed. A successful ontologist today may begin life as a philosopher, but may equally begin as a computer scientist, a data engineer, a biomedical researcher, or an intelligence professional. What matters is a willingness to learn a common discipline of rigor in categorization, definition, and formal representation.
This is one reason why our own program is built to draw students from multiple backgrounds. Some will come to us already skilled in philosophy and logic but needing to learn more about semantic technologies, databases, and AI-related workflows. Others will arrive with technical or scientific expertise but with less training in the philosophical questions that inevitably arise when one tries to build robust ontologies: questions of identity, granularity, dependence, reference, vagueness, and ontological commitment. The aim is not to erase these different starting points, but to bring them together within a single framework.
As concerns demand, the signals are already unmistakable. There is growing recognition in both the public and private sectors that ontology is needed wherever data from different sources must be integrated and used coherently. This applies not only in biomedical and scientific contexts, but also in defense, intelligence, logistics, manufacturing, and enterprise information systems. The demand is driven in part by the rise of AI, which has made it newly evident that statistical tools alone cannot supply the semantic discipline required for reliable decision-support. For this reason we are already seeing interest not only from prospective students, but also from working professionals and from organizations that recognize the need for personnel trained in this area.
JB: There is also a broader lesson here for the philosophical community. Philosophers sometimes worry that the public does not understand the value of philosophy. We know the old line from an unsupportive parent: “You’re going to study philosophy? Why, so you can go work at the philosophy factory?” Applied ontology provides a way to say, in a sense, “Yes.” It shows that philosophical skills can become professionally legible when connected to the right technical and institutional problems. Applied ontology is part of the future of philosophy, one in which philosophical training contributes directly to the systems through which governments, companies, scientists, and citizens increasingly understand the world.
Charlie: Barry and John, thanks so much for this update, which highlights how AI can be complimentary and productive – but also increases the need for people and, most importantly, philosophers! I really appreciate the comprehensive review and we can continue to track your amazing progress in this Public Philosophy Newsletter. Thank you!
From the APA Archive
What else am I reading/listening to
 | A guest post by
|