#LORE action relations
Action relations
Name: interacts with
Definition: For entities S and O, it holds that ‘S interacts with O’, if and only if 1) S and O are in spatial proximity over some duration of time, and 2) they act in such a way that they have an effect on each other.
Examples: A boy interacts with his father, a small molecule interacts with an enzyme.
Issues:
Algebraic properties: symmetric
Rule:
Name: creates
Definition: For entities S and O, it holds that 'S creates O', if and only if 1) an action of S directly or indirectly leads to the existence of O.
Examples: an artist paints a painting, a ribosome synthesizes a protein, a diver creates a wave
Issues:
Algebraic properties:
Rule:
Name: modifies
Definition: For entities S and O, it holds that 'S modifies O', if and only if 1) S and O are in direct or indirect physical contact, and 2) one or more qualities or attribute values of O changes as a result of its interaction with S.
Examples: An enzymes phosphorylates its target, a desiccator dries a sample,
Issues:
Algebraic properties:
Rule:
Name: destroys
Definition: For entities S and O, it holds that 'S destroys O', if and only if 1) an action of S directly or indirectly leads to the non-existence of O.
Examples: An enzyme cleaves its substrate, a secretary shreds a sensitive document,
Issues:
Algebraic properties:
Rule:
Ward on 'interacts with': would there also be a non-symmetric subtype for this (e.g. 'influences', or is 'modifies' meant as such?). The definition says 'effect on each other', but in some situations, only S or only O would get an effect. So I would suggest 'S has an effect on O, and/or O has an effect on S.'
Some general comments:
1) Returning to the discussion of "every relation can be represented as a class", I would say that every Action Relation could be described by some kind of Participation relation (e.g. destroys EqTo? some Thing participantOf some DestroyingProcess?), but I understand this is somewhat arbitrary. Can we formulate some criteria of what is really canonical and what would be on Ward's level 2? Some basic agreements could be a very generic set of classes, such as processes, material entities and dimensional spaces - whatever has some overlap with these classes is classified as a level 2 relation...
2)The "indirect" wording on the definitions make room for quite different understandings. While I normally think it is a good thing for users, this is like creating definition based on intentionality, as being directed at some object. For instance, if I start a chain of actions intending to destroy a cup, I am indirectly responsible for its destruction. However, if I slip on a banana and this starts the destruction chain, is the relation "banana destroys cup" valid? I am not proposing to solve this, but it could go to the issues part. Also, assigning roles (e.g. destruction role) to participants allow us to be more specific when required.
Another thing, will the relations "destroy" and "create" apply to both material and abstract entities? For instance, can I destroy the color of a car by painting it? If just to material entities, then it may be helpful to use Ontoclean's terminology of identity, such that destroying object x means that x lost its identity, but that its parts can still form different objects, whereas in modifying object x, x maintains its identity, though some parts may change.
Wouldn't the relations has_participant
belong to this group? They would be type 2 / 3 relations and therefore more basic than the relations already included.