From: Werner C. in S. <wer...@ec...> - 2005-09-13 14:56:32
|
----- Original Message ----- From: "Fabian Neuhaus" <fne...@we...> > "Causes" is a relation between events. For example, not Oswald was the > cause of Kennedy's death, but Oswald's pulling the trigger of his rifle. > The best reading of 'causally active participant' I can come up with is: > Participant of an event that caused something. In our example there are - > among others - the following participants: Oswald and the rifle. According > to the criterion both are agents. However, I am sure that this is not the > intended result. Therefore, there is something wrong with this criterion. Not wrong, but not precise enough to capture interesting differences. Remember that i wanted to introduce also an instrument relationship or enabler, whatever the name. The riffle enjoys the instrument relationship. > > > I always agreed that we can apply the notion "agent" in the case of higher > organisms with intentions. That's why the "agent", "patient", "instrument" > distinction works for swimming and weeding events, but not for the domains > of biomedical ontologies. But your argument for that is intention, and I'm trying to explain you that it has nothing to do with intention. This is to say, intention is something that comes on top (or not at all) of the kind of relationships that I tried to introduce. If somebody causes somebody else to die, it does not need to be with intention, although that somebody is a higher organism. >>> Now you want to introduce a more fine grained distinction and introduce >>> subtypes of Participates_while_undergoing_some_change. >> >> No, I didn't. The "change" was, I repeat, only mentioned because there >> are processes in which an entity enjoying the has_agent relationship also >> undergoes the process such that it changes: eg suicide, self-destruction, >> cell-transformation, etc. > > Okay, so "agent_of" is a direct subtype of "participant_of", right? Yes. That's the way it is in our paper. And I think it is right (or I would not have co-authored the paper). > So the definition is of the following form: > > x agent_of y if and only if (x participant_of y and ... ) > > Please fill out the blank for me. You are the logician :-). There is a description in the Genome Biology paper, although the instance-level relation is declared 'primitive'. That is probably because the causation relationship is not addressed. That must be done first before we can make has_agent a defined one. Werner |