Does the dissolution of a group implies the leaving of its members?

[stephenhart8]

In the case of A. J. Casson who joined the Group of Seven in 1923 until the dissolution of the group:

• Do we need to create a dissolution event when the group is dissolved? or
• Is this dissolution event automatically implies the leaving of its members?

Is it necessary to have a leaving event? If we need to create leaving events after a dissolution, that would need implementing some scripts in order to create those triples.

[Habennin]

Do we need to create a dissolution event when the group is dissolved?

If you have the knowledge in the data, yes. Otherwise, no.

Is this dissolution event automatically implies the leaving of its members?

It is the logical conclusion that if a group has dissolved then all of its members have left (though in Canada for example there would be the counter example of the Progressive Conservative Party which was disbanded in order to form a new party the Conservative Party and yet some of its members continued to want to be part of the PCs. They lost their case though.

Though this it the logical implication, I am not sure if you would need to go so far as to instantiate it. The property p107 is already future proofed by saying former or current...

[KarineLeonardBrouillet]

Notes on verbal meeting 2020-02-17

Dissolution of a group: use recurrent properties that are well maintained. Would support writing a script would be right 9/10 times. Does bring up the discussion around the Open World and organizing knowledge bases. Models impose quantifiers to the world. But in the knowledge base, different opinions might build up to seemingly non-logical information so it is ok to escape such logic (e.g. Mary and Sally cannot both be the mother of John, but we cannot know which one is the mother).

We would need a script that creates the leaving group event for every member of a group that has a Dissolution Event.

[VladimirAlexiev]

Any such inference is prone to fall victim to non-monotonicity. Consider these cases:

• you learn of a new member X and add a Joining event. Do you also add the automatic Leaving event? Now your script logic is more complicated.
• you have added the automatic Leaving event for member Y. Now you learn that he left earlier. Do you delete the automatic event?
  • What if he rejoined, and finally left with the groups dissolution?

So you could have such script, but only under human control of the results.

[stephenhart8]

@VladimirAlexiev Indeed, a script could create more problems than solving some.

Another option, discussed with @illip and @KarineLeonardBrouillet on the 23.03.2020, would be to tackle this problem not by adding some data or modifying the model, but by having a specific query: When we search for the memberships of an Actor, and there is no end date of that membership, we could retrieve the dissolution date of the group.

[illip]

The Semantic Committee has decided on 2021-05-06 to test the benefits and the risks of these inferences.

Short term, we will explain in the Target Model that there are no semantic links between the Group Belonging and the Formation/Dissolution of groups patterns. So the users must be careful when it comes to querying the membership time-span as they must check both patterns. For instance, a leaving event might not be recorded but the dissolution event could give some information regarding the membership ending date.

When CHIN will have a sufficient amount of data, we will test if a semi-automated script could be relevant in order to general new knowledge. For instance, we could test how many errors are generated when we create several leaving events when a dissolution event is recorded. It will also be important to evaluate the resources required to maintain these generated data, especially regarding the update process.

As the SPARQL queries to get all the proper data to answer those membership time-span issues might be complex, CHIN will also offer a few of them in order to help end-users to analyze their results.