Advertisment design study for Pierce Arrow automobiles (1915). By Edward Penfield. Via Library of Congress collection.

What Makes a Good Ontology

Forrest ChristianOverachievers 2 Comments

Since we’ve been talking about naming, and Glenn Mehltretter posted such a great example in his comment, it’s a good time to consider Ontologies. Since Kinston’s Taxonomy is what the knowledge management people would call an ontology, I’ve been looking at them.

Barry Smith of the Basic Formal Ontology project at Institute for Formal Ontology and Medical Information Science (IFOMIS) has a some strong thoughts against the linguistic turn in most ontological development. Linguistics doesn’t really care if a statement maps to something in the real world, or something that is provable or repeatable. It only cares about the structure of the statement. From their point of view, statements about the imaginary world of my fantasy series (the novels have been on hold for long enough to say that they will never, ever be completed) are the same as statements about high-energy physics.

Smith thinks this is all so much idiocy where ontologies are concerned. He writes:

Bad ontologies are (inter alia) those whose general terms lack the relation to corresponding universals in reality, and thereby also to corresponding instances.

Good ontologies are reality representations, and the fact that such representations are possible is shown by the fact that, as is documented in our scientific textbooks, very many of them have already been achieved, though of course always only at some specific level of granularity and to some specific degree of precision, detail and completeness.

[Thomas Bittner and Barry Smith, “A Theory of Granular Partitions”, in: M. Duckham, et al. (eds.), Foundations of Geographic Information Science, Taylor & Francis, London, 117-151, 2003; quoted in Barry Smith, “Beyond Concepts: Ontology as Reality Representation“. (PDF)]

The article this was quoted in, also by Smith, is pretty good at describing why our work is so important. All models are false: that is, the map is not the territory. All maps make choices in what to ignore and what to focus on. A map made for demonstrating demographical variance in Siberia will not be that useful to me when I am lost in the Siberian winter. But that doesn’t mean that all maps are equal. Some maps represent nothing in the physical world: Tolkien’s maps of Middle Earth, for example. Other maps, such as the early cartographers’ attempts at defining the land masses of the world — or the map of Belgrade that NATO used and “inadvertently” bombed the Chinese embassy — are not false so much as incomplete. They do not as accurately achieve their goal as do other maps that more accurately represent the current configuration of streets and land borders.

The “map” created by Elliott Jaques and Wilfred Brown (and those who followed on) is a symbolic representation, and like all symbolic references it is not the thing itself but removed conceptually from it. It is not that it is fully accurate in representing all reality but that it provides a more accurate and complete representation of that than other models, and it maps certain areas that had previously not been mapped successfully.

Image Credit: Advertisment design study for Pierce Arrow automobiles (1915). By Edward Penfield. Via Library of Congress collection.

Comments 2

  1. I would suggest that the statement that promotes that all models/ maps are false is in fact a false statement. It would be more accurate to state that all models/ maps are incomplete and in some cases inaccurate.

    Granted, they concentrate their focus on the elements the cartographer is attempting to bring attention to. This is not to say that their purpose is irrelevant, or that the relevance of the particular model in view is insignificant in its unrelated purpose to something else.

    Every scientific verifiable fact began as an idea, a concept, a postulation, that was subsequently verified. It is worth noting that the concept can be obscured by a particular belief or value system and that on more than one occassion science has been discounted because it is contrabelief to mainstream thought.

    The process of cognition is one that creates structure. It assimilates the interrelationships between disjointed and otherwise nonrelated ideas, events, models and conversations. It elevates complexity to a level of simplicity. The extent to which one is able to make sense of the occuring world and establish the causal and effectual relationships among otherwise disconnected occurrences to charter a course is what has been labeled complexity of mental processing by Jaques. It also relies upon the skills and knowledge (experiences) of the individual rationalizing his or her interpretation to advance in the direction of the intended goal.

    Consider that what has been promoted as “good ontology” is merely what is presently known. While it may offer a systematically repeatable approach to what is being pursued in and of itself it does not offer anything new and as a consequence will become obsolete. What is being promoted as “bad ontology” may simply not yet have had its relevance related and assimilated to what is already known. It may in fact offer a breakthrough in technology that ensures competitiveness, or survival, in an ever changing and evolving world.

    The responsibility of leadership is complex. Human beings by their very nature are drawn to the attention of one goal or another sometimes without understanding the future implications of current actions in the creation of future problems. The solution to the present problem may very well become the source of a subsequent problem.

    In respect of Jaques’ stratified systems theory he has offered the proof with respect to the complexity of mental processing and the prevalence of various strata in society. His requisite organization managerial practices are more practical than philosophical. To argue that what he has observed in respect of a model/ map or structure for managerial leadership is false is indefendable. To suggest it is incomplete would be perfectly valid.

  2. Post
    Author

    Yeah, this is exactly right. All representation are “false” in that they are representations and not the thing. Not that some more accurately represent the thing than others do.

    Some ontologies are simply ridiculous: I’d call these bad. All are incomplete, but some are either more complete than others or provide more useful data than others. Things that are purely symbolic representations themselves (and I think that organizational hierarchy is at least partly this) are more difficult because there is no physical reality out there. “Felt-fair pay” may more accurately represent an external reality (or a reality in which it seems all humans participate across cultures and language systems) but not a physical external reality. It probably represents something inherent to the human mind, which is a reality that isn’t just within me but across all humans (or most).

    I guess you could say that a fake map of the world could be complete, but it doesn’t have as much useful data as the national map that came on the child’s placemat at a restaurant. All representation breaks down. Even though there seems to be a series of built-in pre-structures or structures to the human mind that allows us to automatically identify things that seem to share characteristics as a class (say, “ducks”) we can still have problems when we need to check whether or not something like a swan is part of that class. And then there’s the platypus, which is clearly not a duck but you’d have to stop and ask the question.

    Maybe part of the problem is that Natural Language and Formal Language are both necessary for the totality of human communication but they don’t ever meet. I probably get into more trouble by switching willy nilly between the two, even here.

    Anyway, I think I speak for the entire writing team when I say that this level of lucid thinking is why we miss having Al post his thoughts.

Leave a Reply

Your email address will not be published. Required fields are marked *