Object Oriented Ontologies?
Only by Occident.
I am not a practicing programmer any more, and I never got deeply into the intricacies of the art and theory, because, well, Life. So I am going to be quite simplistic in my comments about it here. At me if you wish to.
Ontologies are most widely mentioned, if not used, in the wider community in the context of business and computing. I’d like to make a few comments about both and then finish with a final flourish and the usual over-generalisations.
Abstract
The use by computing of the term “ontology” is at best metaphorical and analogical, and at worst of pure hyperbole. Its use in general society is usually overselling ordinary ideas in social institutions such as business (particularly in management theory), health (nursing and disease classification) and policy (definitions of groups that then are used as metrics).
Ontology does not arise from ordinary language, practice or belief, but since Quine, the ontological project is aimed at explicating scientific theories.
In 2003, a consortium of scientists from some top ranked (whatever that means) universities published the first version of what they titled the “Gene Ontology Project”. Billed as “Gene Ontology: tool for the unification of biology”, it was fundamentally a translation manual between many different ways of naming and referring to genetic products. Philosopher Barry Smith and colleagues named it a form of Formal Ontology, and extended that to include the Open Biomedical Ontologies consortium a few years later. As he and his colleagues wrote in 2015:
One increasingly dominant strategy for the organization of scientific information about the world in computer-friendly form is associated with the term “ ontology ” (or sometimes “ ontological engineering ” or “ ontology technology ” or “ applied ontology ” ), understood as meaning (roughly) a controlled vocabulary for representing the types of entities in a given domain.
This use of ontology for what is effectively a database specification protocol blurs some lines. An ontology is, as Arp et al. call it, a “representational artifact”. To be fair, Arp, Smith and Spear, distinguishes between three kinds of ontology – Formal, Material, and Instrumental Ontologies – and they require of an ontology that it be, at least in a scientific domain, “semantically interoperable”, which is to say at the formal level it defines types or universals, at the material level it describes the entities that realise these universals, and at the instrumental (they use the term “organon”, meaning instruments, which comes from Aristotle’s commentators and Francis Bacon), that it be useful or convenient. The idea is that we need a unified representation to communicate within a discipline. I think we encountered this issue some time before, with the Universal Language Project.1 The justification for universal schemas, whether ontological, epistemological or instrumental, is as Arp et al. put it:
Above all, the very success of ontology-based approaches to the integration of data has led to a multiplication of ontologies in ways that threaten to recreate in a new form the very problems of interoperability that ontologies were themselves designed to solve. [xvii]
They call this the Tower of Babel problem. It also gets called taxonomic incommensurability and a range of other names in different contexts.
Now I use this example case to talk about something that I see more frequently: the notion of object-oriented ontologies, particularly in the humanities, basing the idea on programming techniques. As I noted above, I am not currently a programmer nor do I know a lot about current techniques, so I’m going to use database setup instead; but so far as I know, they are logically much the same thing.
A database is a table of data, which has a singular record index. A relational database has a number of tables linked by one or more fields within the tables.2 Now if a database is set up for a business functioning under one jurisdiction and set of circumstances, then the items of information (contained in the fields) are cemented in, to a degree. Change jurisdiction, or institutional practices, and the structure may become inefficient, misleading, or even fatal to the enterprise. This database structure forms the domain ontology of (the representation of) the business.
I think the analogies with scientific and social ontologies are rather obvious, as are the flaws and shortcomings. A business that cannot adapt its informational systems to meet the needs of a new environment will fail. A scientific discipline that cannot change its domain ontology will become either stagnant or falsified. A social ontology that sets, say, diseases in concrete will become ridiculous. At least, if there is a global ontology in those domains. And example for medicine is the DSM, arguably founded upon psychoanalytic categories in the 1910s (as the Statistical Manual for the Use of Institutions for the Insane) before the first edition in the 1950s. It is now in its fifth edition (2013,, and a revision in 2022).
Now in medicine and psychiatry, a classification of diseases (an ontology if you like) is called a nosology, which is simply a taxonomy. Sometimes it is a social categorisation, sometimes a biomedical one, and sometimes both. Sometimes it will be an economic category: that to which a certain drug or technology can be applied.3 But what an ontology of this kind is not is arbitrary and inconsequential. And it must be allowed to change as the conditions in which it was once a fit-for-purpose notion change overtime, because sure as eggs it will change.
So are constructed ontologies a lost cause? In IT they may very well be. At the least they need to be revised on a regular basis, and that costs, in time and money, and possible (likely?) failures. In science, any science, setting up an interoperable OOO to fix the reference terms for many subfields in a domain of research is both a short-term benefit (albeit a costly one) and in the longer term, an obstacle. It can work well and it can work badly, and all the intermediate outcomes, in all kinds of sequences. One thing it is not: a universal panacea.
Next I’ll discuss holisms some more.
See my Species book for information and references. Basically this is the problem of scientific taxonomy.
When I was learning to computer, this was based on a set of 13 rules set by E. F. Codd. We called this the Gospel According to Codd. One practice we were taught was to “normalise” the data: to make sure it was only entered and checked once, to avoid version errors. To this day I wish more people outside IT knew of this.
For example, Nic Rasmussen, in his On Speed: The Many Lives of Amphetamine (2008) argues convincingly that the drug was a solution to a problem that had to be devised to make money. The uses so invented were: pilots in Germany during the second world war; for stay at home mothers with (understandable) depression; children with learning disabilities; bedwetting; etc. Under the label of Adderall and other brand names, amphetamines are prescribed for ADHD diagnoses even now.