Proposal: Split out the familial relations into a domain ontology

[alanruttenberg]

There are a LOT of these - I stopped counting at 50 -looks like a third of all object properties are these - and they are so specific that I don't think they belong at the mid-level.

[alanruttenberg]

In addition

• They mostly don't have definitions, leaving the question of whether e.g. a step father is a valid value for 'has father', or an adopted person is a child - are these speaking to social roles or biological facts?
• Many should have disjointObjectProperty axioms, since I don't think you can be a mother and father at the same time (although maybe, if the steps, surrogates, and fatherness is considered a social role.

There's a tutorial about a familial relations ontology: https://oboacademy.github.io/obook/tutorial/fhkb/ that might serve as a basis for further enhancement if it becomes a domain ontology.

[johnbeve]

+1
@mark-jensen I know this has been on your chopping block for a minute...
FYSA @APCox @CarterBeauBenson @neilotte @oliviahobai

[BrendaBraitling]

@alanruttenberg @mark-jensen @johnbeve I am not an expert at FOL but perhaps CCO could embrace a midlevel approach to kinship as presented in this paper - which you may be familiar with - It claims to handle kinship and be aligned to BFO2020-FOL. It seems to be concise and immediately useful to real world systems as mentioned in the paper. I do not have the FOL expertise to evaluate it completely. Interoperability across domains regarding information about humans seems critical.

An Extendible Realism-Based Ontology for Kinship -2023 - Michael Rabenberg, Anuwat Pengput, Werner Ceusters

CCO probably should at least assert that a natural person will always be the biological child of its biological parents. If the above proposal from the paper is not sufficient, the domains can then deal with the kinship beyond child - parent.

In general I think Midlevel is the best place to provide human biological filial generations to allow for interoperability across domains. This would aid in untangling domain specific, diverse configurations of family arrangements, households, legal parents vs biological parents, etc...

[alanruttenberg]

@BrendaBraitling Thanks for the reference! Werner is the person who has most used the BFO axiomatization other than myself. So the FOL will definitely be compatible. It's undoubtedly well thought out and so worth paying attention to.

Translating to OWL will be difficult in a number of cases, because of the ternary predicates - time indexing and others. But at least it is well defined. It looks like it is limited to "blood relations", so no accounting e.g for an adopted child, or a grandparent by a second marriage, or non-biological, but legal parents.

Regarding where to put the familial relations, some few might be appropriate, but not to the level that CCO has. Having the relations in a domain ontology is no burden - just another import. But it's important, I think, to have CCO not be overwhelming large, and for a mid-level to exhibit some amount of consistency in how specific the terms get. We have, e.g. Algorithm, with no subclasses (e.g why not sorting, function fitting, caching, visualization). Or Quality specification that specializes only to size and mass. We have no term for law. We have Biological sex, but no term for Gender - even the most general. So is it really appropriate that we have relation for third cousin? Even the cited paper only goes to first cousin. SNOMED CT, which the paper includes links to, doesn't have third cousin and it has something like 300,000 classes. Seems it would be rather uncommon for that relation to be used outside some specialized area. Also, it's complicated - you mention legal parents. We don't have those in CCO AFAICT. And there are all sorts of other complications. So that's why the suggestion it be made a domain ontology, beyond the simplest relations, maybe spouse and mother, father, daughter, son - by role, not biology. Or maybe just spouse, parent and child.

[CarterBeauBenson]

@johnbeve I agree that these should be removed. The next question is where should the relations go? They should remain publicly accessibly which limits us to CPO , the person ontology, linguistics ontology, or breaking the relations out into their own module.

[cameronmore]

Classes for family and different extensions of family, family relations, and equivalence classes for family members might be best served all being put into a single domain together. Perhaps an Ancestry Ontology?