…An application of non-monotonic reasoning,
Robert Hoehndorf et al.
Starts by describing the definition of BFO function. It’s rather lengthy, so I won’t reproduce it here – you can find it in the BFO file. Fuctions may “typically” not be realized, e.g. sperm cells. Functions may be aquired through transformation. What is a realization? What makes a process a realization of a function? Participation of the function bearer is not sufficient. The mode of participation in a process necessary to understand realizations, and these modes are called roles or qua entities.
The ontology of function (OF) hopes to sort some of this out. What is the relation between function and disposition? Function and causality? Larry Wright says that the function of X is Z and this means: X is there because it does Z; Z is a consequence (or result) of X’s being there. The heart is there because it pumps blood, and pumping blood is a consequence of the heart being there. But this doesn’t work for “the heart makes pumping noises”, as the heart isn’t there to make pumping noises. Therefore you need some refinements.
John Searle says all functions are socially ascribed by an agent to an object; and functions are always observer-relative; function ascription is based on brute facts. Nicolai Hartmann says 3 steps: first, a goal in the future has to be set; second is planning the goal’s realization; third is realizing the goal (causally). 1 and 2 create functionality. The third one isn’t necessary for function. If we don’t need 3 to have a function, must that realization be possible at all? What happens if we set a goal and plan, but then something went wrong? No realization is then possible, and we have a missing disposition. The entity is malfunctioning with respect to its function. Functionality does not imply causality because things can malfunction. But when you look in BFO, what they use in their explanation, what they use is almost always some form of typicality or normality: normally, entities do not malfunction. They have created a formal framework for malfunctiongs.
In non-monotonic inference, it permits a sort of default reasoning that X is not malfunctioning as long as it is not proven otherwise. The use of default logics, circumscription or auto-epistemic logis. It can be implemented using answer-set programming. For the implementation, start with an Ear that has function Hearing. There is a disposition D to participate in the hearing process (realization) as a reciever (role). If an instance e of Ear is not malfunctioning, it has a disposition d (instance of D). Normally, instances of Ear are not malfunctioning. Deafness implies a malfunctioning.
BFO’s and other’s definitions of function needs refinement. There are a number of philosophical explanations of functions that can be divided into: reductions to causality, and social ascription. Each explanation relates functionality to causality, and most use normality/typicality. Non-monotonic reasoning is necessary to formalize normality. He proposes that we focus on the necessary conditions and inferences instead of definitions for function.
FriendFeed discussion: http://ff.im/4wOyA
Please note that this post is merely my notes on the presentation. They are not guaranteed to be correct, and unless explicitly stated are not my opinions. They do not reflect the opinions of my employers. Any errors you can happily assume to be mine and no-one else’s. I’m happy to correct any errors you may spot – just let me know!