A couple of weeks ago we started an interesting discussion about the issue of realism vs. relativism in relation to biomedical ontologies. As pointed by Werner Ceusters and Barry Smith in one of their joint papers, this choice is not merely theoretical -- it can have some significant implications for the quality of how we describe biology and health sciences and, as a result, make inferences about them.
When asked whether the issue of reality vs. relativism makes any difference, however, a pragmatist would probably say that this is irrelevant since experience will provide the adequate feedback to determine whether a biomedical ontology is adequate or not. This view is interesting in that it agrees with some of the current practices of those developing ontologies from a computational perspective. In their perspective, the development of computational ontologies should be started with a use case, which is a preview of what future experience will be while putting a computational ontology into practice.
Interestingly, way before use cases had even been conceived, Arthur Lovejoy made some related comments about this topic in his book The Thirteen Pragmatisms and Other Essays. Lovejoy, a previous student of William James at Harvard (Diggins, 2006), argued that one of the major problems with Pragmatism is that it relies on experience, which is a post-facto type of experience, therefore not providing much guidance to what should actually be done when no experience is possible. Although in biomedical ontologies a situation where "no experience is possible" is not frequent, we are constantly hunted down by our inability to predict how exactly our ontologies will be used, ultimately turning Lovejoy's prediction into reality. Having acknowledged this limitation in Pragmatism, partial inferences from previous experience can frequently provide some guidance, the ultimate guidance always been derived from the event itself. No guidance will ever achieve the status of deterministic guidance, in that every event always contains both fixed and random factors, a belief widespread among those following the probabilistic tradition.
Having or not a probabilistic view of the world that relies on previous assessments, the problem of purpose is still present. In other words, your purpose will probably dictate how you perceive and think about the world (Clarke, 1989; Smith, 1978).
Subscribe to:
Post Comments (Atom)
1 comment:
When Richard writes: "your purpose will probably dictate how you perceive and think about the world" he identifies what I believe is a core problem at the heart of current approaches to ontology.
Biomedical data is siloed to a degree which prevents the harvesting of important kinds of cross-domain information. To address this silo problem, cross-domain ontologies like the GO are being built, as controlled vocabularies to be used in annotations of data, in ways designed to ensure some degree of comparison and integration of siloed data.
But now the problem: because so many different groups are each building their own Pietrobonesque ontologies in light of their different purposes, the silo problem is raising its ugly head once more -- but now at the level of these ontologies themselves. Building purpose-dependent ontologies has tended to mean: building local, idiosyncratic, short-lived ontologies.
The solution, as I understand matters -- a solution first propounded by Cornelius Rosse and now embodied in the OBO Foundry initiative -- is realism-based ontologies: the key to interoperability lies in building ontologies which represent the common biomedical reality of cells and organisms, GABA transporters and VGLUT3-immunoreactive axon terminals. Whether these things exist is not, after all, a function of our purposes (and nor is it a function of the software tools we happen to have to hand at any given stage).
Post a Comment