Issue 256: groups and relations between persons

Starting Date: 
2014-08-04
Working Group: 
4
Status: 
Open
Background: 

Posted by Christian-Emil on 4/8/2014 
 

Dear all, 
This is not a part of the discussion in April about groups and aggregations. It is groups as a way to model relations between persons (actors). I gave a presentation about CRM and prosopography at the DH2014 workshop "Ontologies for prosopography" (see http://edd.uio.no/artiklar/DH2014/C-E_Ore_prosopography.pdf ). 

The current CRM way to model relations between persons is to use the E74 Group. A relation is modeled as an instance of E74 Group and the type of relation is expressed via P2 has Type. In a non-symmetric relation each person is linked via 'P107 is current or former member of ' specified by 'P107.1 kind of member'. This is all according to the scope note in CRM. 

One may note that an instance of E74 Group used in this way represents an instance, an n-tuple, of a relation (seen as a set of n-tuples as in mathematics or in relational databases). The relation is identified by the type of the E74 group. 

I was a little skeptical when this way of modeling relations where introduced in CRM. My first thought was to define explicit, typed properties. After studying how for example the SNAP (Standards for Networking Ancient Prosopographies, http://snapdrgn.net/) tries to cope with their at least 65 identified relations between persons by introducing a relation class in RDFS, I realized that the CRM solution is very good. 

Since this is not meant to be a statement about me and CRM, I will raise two issues which I think need some discussion. 
1) E74 Group scope note "This class comprises any gatherings or organizations of two or more people that act collectively or in a similar way due to any form of unifying relationship.[...]" Will all related persons fulfill the requirement " act collectively or in a similar way due to any form of unifying relationship", that is, is E74 Group too narrow to be used to model all kind of relations between persons like the ones we find in prosopography? 
2) The modeling of relations by 'P107 is current or former member of ' specified by 'P107.1 kind of member': If this is to be implemented in RDF(S), should we in the CRM definition recommend or at list hint to a good solution to implement the .1 E55 Type properties? 

Posted by Martin 4/8/2014 
Dear Christian-Emil, 

I could quite well imagine having a sort of more general Group describing a social bond that would not involve members potentially "acting as one" or one speaking for them. 
In that case, that Group would no more be "one Actor". 

Would you regard http://snapdrgn.net/ as a good practical scope? Do you have other sources to map from?
If we have a practical scope, we can model things. Do you propose an amendment to the CRM or a "social" extension? 

Posted by Christian Emil on 4/8/2014 
Snapdrgn and the associated projects for prosopographical information (prosopographies) can be a case study and serve as a source of information/evidence. It is only a 2-3 years project. However, it can be a task to see how to map the snapdrgn ontology (which is expressed in rdf(s) I believe) to CRM. If we cannot do that, CRM needs adjustment or amendments. I will try to make the mapping and study the matter further. 

Posted by Detlev Balzer on 5/8/2014 Dear Christian-Emil

by the way, a more modest approach to prosopography (compared to snapdrgn) has been taken here: 

http://d-nb.info/standards/elementset/agrelon.owl 

Unfortunately, this hasn't yet made it beyond the proposal stage. It may, however, serve as an example of what kinds of relationships are considered important in the library sector. 

Posted by Christian Emil on 5/8/2014 
Dear Detlev, 
The agrelon demonstrates clearly that there is a lot of possible relations. It could be interesting to see the set of relations if one tried to model the traditional peasant family in Russia. Traditionally there is a very large numbers of terms for describing the relations in the extended family. 

Posted by Martin on 5/8/2014 
Anthropologically, there seem to be an immense number of variants of kinship, as George Lakoff describes.
So, the challenge for us is to find the generalizations that would be relevant for recall in an intergated information system. What would be a reasonable distinction in a query? When would be the answer set too large? I read that all Chinese with last name Wang (or another) assume a sort of kinship, of "we". Where are the limits to "minorities" ? Are there reasonable delimiters to more immediate forms of kinship? 

Could we classify social relations by 
* kinship & kinship equivalent (like adoption, marriage), 
– immediate ?? 
– relevant for social interaction 
– spiritual/political relevance 
* by business & interest groups, 
* by acquaintance&neighborhood, 
* by employment 
* by dependency of power (liege, slavery, military) 
???? 
Which of these could appear as a selection in a query? Do we have research questions and queries for prosopography and other social relations? 

Posted by Christian Emil on 5/8/2014 
I think we should accept that ethnographical collections and others have a need for expressing relations between persons. To make the CRM simple we should model a schematic way to express such. A solution could be to say that these relations are defined by humans and create a relation class somewhere under conceptual objects. 

Posted by Martin on 5/8/2014

On 5/8/2014 2:28 μμ, Christian-Emil Smith Ore wrote: 
I think we should accept that ethnographical collections and others have a need for expressing relations between persons. To make the CRM simple we should model a schematic way to express such. A solution could be to say that these relations are defined by humans and create a relation class somewhere under conceptual objects.

Good idea. In case of associated laws, norms etc. one could specialize plans in the wider sense, plans how people should behave. 

Does anybody on this list have an opinion about if we should develop an extension for social relations? 

Do we have any anthropologist among us? 

Posted by Vladimir 5/8/2014

more modest approach to prosopography (compared to snapdrgn): 
http://d-nb.info/standards/elementset/agrelon

I like the analysis done in AgRelOn, see MTSR 2012 paper (p222). 
It abstracts away the gender (to reduce combinatorial complexity) and analyzes the rest hierarchically (e.g. professional vs familial relations).

E74 Group used in this way represents an instance, an n-tuple, of a relation

There are inter-dependent fields that need to be kept in sync 
– the type of group and the P107.1 props need to be correlated. If the P107 props are specific, maybe one can skip the type of group 
Something like this: if a group has P107_1_mother and P107_2_father and P107_3_child, then its P2_has_type is "family". 
– the group is a short-cut for life events, as shown in C.E.Ore's presentation

Current Proposal: 

In 31st joined meeting of the CIDOC CRM SIG, ISO/TC46/SC4/WG9 and the 24th FRBR - CIDOC CRM, the sig decided that we should Collect examples of vocabularies and consolidate and map them  to CRM.  CEO will send examples.

Heraklion, Crete, October 2014

Posted by Christian-Emil on 7/2/2015

The SIG decide 8-10 years ago to use  E74 Group to model relations between persons.  From a practical point of view this is very handy. A relation is simply an instance of E74 group of E21 Persons where the persons in question  are members. The intention behind   E74 Group is that the group as such can act as one unit.  As Martin pointed out :"I could quite well imagine having a sort of more general Group describing a social bond that would not involve members potentially "acting as one" or one speaking for them.  In that case, that Group would no more be "one Actor". (Martin Doerr 4/8/2014).

In prosopography (see for example http://snapdrgn.net/) and in social anthropology one may be interested in relations between persons not naturally acting as a single actor . For example, my great grandfather aunt (born in the end of 18th century) and I may be related but do not have the potential to acting as one actor.

The excerpt from ULAN below, shows besides family relationships a  student/teacher relationship. I would claim that it is not necessary  so that a student and a teacher should be seen as  members of a E74 Group.

Still we need to be able to model such relations. It is of course possible to extend CRM with new properties for each relevant relationship. One may also extend CRM with a generic relationship property with a .1 E55 Type property indicating the kind of the relationship. The one need one generic property for binary relationships, one for trinary  etc etc.

A solution may be to introduce a single relationship group which is not a subclass of E39 Actor, but with the similar membership properties as  E74. One may argue that (social) relationships are abstract entities originating in humans' minds. Therefore the class of such "relationship groups" should be a subclass of E28 Conceptual Object. A relationship is not a type but has a type. It is hardly a legal object(?) It could be a subclass of P89 Propositional Object?

We need some discussion here. 

Posted by  karl Grossner   on 7/2/2015

In CRM terms, a sub-class of P89 Propositional Object makes sense to me.

In a 'Spatial History Ontology' I developed a few years ago (and never wrote on outside my dissertation), I developed six modeling patterns and extended DOLCE to account for them, one is Groups and Membership.

I also studied CRM patterns closely at that time. One gap in both is what you describe, and I called it a FunctionalGroup (FGRP). In DOLCE terms it sub-classes Collective:

Endurant
- Social Object
-- Non-agentive social object
--- Collection
---- Collective
----- Functional Group

Examples I gave included 'central European farmers circa 1200 CE,‘ Bauhaus artists, U.S. Presidents, French speakers. These groups are only together incidentally or by fiat, in many cases distributed through time and not capable of agency. There are some cases that strain this conception, for example the people present in a square, who later became a (very agentive) mob.

There are many more types...an excerpt from that work:

******
Functional groups are necessary to specify and discover meaningful aggregations of persons without requiring declarations of identity. For example, set (group) membership could be a function of:
- Legal or otherwise formalized membership in a role: citizens(country); residents(place); employees(organization); members(organization); registrants(conference); etc.
- Event participation: participants (event)
- Relationships: friends-of (person)
******

Posted by Martin on 9/2/2015

Dear Karl, Christian-Emil,

My concern is that we may put in one pot things that may turn out later as incompatibly distinct.
My concern is not to introduce too many relationships. I become concerned when a single relationships is proposed without its ontological substance been understood. Such a thing any application can do on its
own account. I would be much more happy if we could analyse different senses.
I'll give examples below>

On 7/2/2015 8:41 μμ, Karl Grossner wrote:

In CRM terms, a sub-class of P89 Propositional Object makes sense to me.

In a 'Spatial History Ontology' I developed a few years ago (and never wrote on outside my dissertation), I developed six modeling patterns and extended DOLCE to account for them, one is Groups and Membership.

I also studied CRM patterns closely at that time. One gap in both is what you describe, and I called it a FunctionalGroup (FGRP). In DOLCE terms it sub-classes Collective:

Endurant
- Social Object
-- Non-agentive social object
--- Collection
---- Collective
----- Functional Group

Examples I gave included 'central European farmers circa 1200 CE,‘ Bauhaus artists, U.S. Presidents, French speakers. These groups are only together incidentally or by fiat, in many cases distributed through time and not capable of agency. There are some cases that strain this conception, for example the people present in a square, who later became a (very agentive) mob.

This construct:

'central European farmers circa 1200 CE,‘, 'French Speakers'

reminds me situation logic as described by Aldo Gangemi in his "OIO" ontology.
Basically it appears as a set in the mathematical sense, confining a base class "Actor" to an arbitrary
set of constraints. To my opinion, these sets do not qualify as an ontological category in their own right, because they do not have any observable unity criteria as an individual.

I'd argue, that you only need Description Logic to create the group you need by rule, as long as the
constraints are expressed in terms of the CRM. You can do it with OWL, and it is CRM compatible.

People becoming an agentive mob would actually form a "superindividual", one Actor.

The "Bauhaus Artists" actually acted together to my best knowledge. I'd need to be convinced by
historical evidence that this was not the case.  Note, that being member of a Group does not mean that
you cannot do things that are outside the Group.

In CRM, we model US Presidents as a sort of pseudoperson, another kind of "superindividual",
which has a unity criterion and identity criterion across its office bearers. We regard them as a Group
with one member at a time. Members do not interact, they continue work of previous members similar to an
individual continuing to act as long as living.


There are many more types...an excerpt from that work:

******
Functional groups are necessary to specify and discover meaningful aggregations of persons without requiring declarations of identity. For example, set (group) membership could be a function of:
- Legal or otherwise formalized membership in a role: citizens(country); residents(place); employees(organization); members(organization); registrants(conference); etc.
- Event participation: participants (event)
- Relationships: friends-of (person)
******
Citizen of a democratic country elect their representative. The citizenship is a membership

document. It can be granted and withdrawn.

"friends of" is a different story. In a narrower sense, "Friends" do reciprocal activities together. This is a
relationship mediated by a continued activity, such as "business partners". May be this would be
a "functional Group" as you have designed?

My grand-grand father is a different issue. On one side, there are "elementary" kinship relationships
which are either biologically motivated or by social contracts of equivalent value, such as adoption and marriage.

These relationships have transitive interpretations which have different social status according to
the culture. George Lakoff cites American cultures that do not layer them by generations.
I do not know if you can adopt someone as second-degree uncle in any culture.
Such kinship claims are normally regarded as fraud if not justified by elementary ones. They may
be discovered by their members via elementary ones.
So, we could introduce a "kinship relation" or "family relation" based on birth, parenthood, marriage
and any degree of transitivity. That would be clear and objective?

I agree with Christian-Emil that they are a social contstruct, similar to legal constructs. Transitive closures of
binary relationships would not require n-ary relationships.

We could introduce an activity-mediated relationship, a sort of "partnerhood" or so, clearly
distinct from the kinship?

As CRM-SIG, I think we should seek evidence from anthropology which distinctions are persistent through cultures. 

In the 32nd joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 25th FRBR - CIDOC CRM Harmonization meeting, the crm-sig decided to elaborate more this issue by considering to introduce an activity induced relationships. Anja Masur and Christian Emil will work on this.

Oxford, February 2015

Meetings discussed: