envo
envo copied to clipboard
EnvironmentOntology
[#1251] 'astronomical body part' - change parent
I heartily approve this, but a high level change requires consensus - what do you think @pbuttigieg, are we ready for this?
I ask the question what benefit do we get from keeping astronomical body part and environmental material as subclasses of fiat object part?
My understanding of fiat object part is that we're saying there is some 2D fiat boundary drawn up and bounding things. I can see that being more true for astronomical body part than environmental material. If there is a good rational that we want to keep such constraints on those hierarchies I'm all ears, otherwise perhaps we should consider pruning top level BFO terms such as fiat object part.
I think that would be over-interpreting 'fiat object part', it says nothing about 2D boundaries. But this serves to illustrate that there is a lot of subjective interpretation in these BFO classes! (that the definitions are subjective is not a novel observation -- there has been a debate about boundaries and granularities in BF for as long as I can remember...)
If there is some property relating to boundaries that we feel is important for some use case, then one compromise would be to instead assert this as a has-characteristic axiom at whatever level of ENVO we feel appropriate. But I am not sure even this is necessary.
On Wed, Nov 24, 2021 at 2:45 PM Simon Cox @.***> wrote:
@dr-shorthair https://github.com/dr-shorthair requested your review on: #1252 https://github.com/EnvironmentOntology/envo/pull/1252 'astronomical body part' - change parent .
— You are receiving this because your review was requested. Reply to this email directly, view it on GitHub https://github.com/EnvironmentOntology/envo/pull/1252#event-5669699745, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMONOOO62QFZQEV7CWULUNVTHBANCNFSM5IXEJZ3A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
@cmungall I'm certainly no expert but I was getting that from the elucidation of fiat object part. Again I'm not sure what this buys us to continue including it in ENVO.
b is a fiat object part = Def. b is a material entity which is such that for all times t, if b exists at t then there is some object c such that b proper continuant_part of c at t and c is demarcated from the remainder of c by a two-dimensional continuant fiat boundary. (axiom label in BFO2 Reference: [027-004])
ah yes, my mistake, sorry! but the point remains, if it's important for users to be able to know which terms in ENVO represent things with fiat 2D boundaries we could add this as a property.
I'm not sold on this - as we have seen time and again, most geographical entities are delimited by fiat, and different communities will set different boundaries (coasts, riparian zones, ...). Including different definitions to align with different community usages means we should acknowledge the fiat nature of the boundaries.
I suppose that depends on whether you see
- defined according to rule A, conditioned by observations b, c, d
as a fiat choice.
I agree that different communities will set different rules and supporting observations. But the general sense is that a delimitation that emerges from rules that are designed to encode natural phenomena is qualitatively different to a rule that is based on drawing an arbitrary line on a map/landscape.
On the other hand, if there is no 'natural' delimitation, then is there any partition of space that is not fiat? In which case 'fiat' is not a particularly useful distinction.
Seems to me that there is a subjective continuum between what is 'fiat' and things that have 'hard' boundaries. Perhaps the distinction is based on the degree to which multiple observers would all agree on the same boundary. Consensus being the criteria for 'hard' boundaries, and valid alternate interpretations being the criteria for fiat: some authority makes a decision abut the interpretation that will be used. So the boundary of the 'Empire State building' (as a non-fiat example) and the boundary of the Rhine River flood plain (fiat--requires adoption of one interpretation of 'floodplain' and the available observation evidence).
Can ontological distinction be made between an 'object' (non-fiat, there is consensus, or close enough, about its boundaries) and 'fiat objects', with boundaries based on interpretation of entity definition and observation evidence. In this view, there are likely astronomical body parts that are 'objects' and parts that are 'fiat objects'.
FWIW, there seems to be unending debate about these matter. E.g., see Barry Smith and Mark Do Mountains Exist.
If we want to classify mounts, deserts, caves, etc. as fiat objects, it would be useful if the fiat nature served some purpose in terms of reasoning. Currently, fiat object doesn't seem to serve such a purpose.
@wdduncan thanks for the reference - I vaguely recall seeing it before, but now I have a more immediate interest in the question.
Nevertheless, I agree with your proposition that the classification 'fiat' is not useful unless it assists reasoning. Since http://purl.obolibrary.org/obo/BFO_0000024 currently has no axioms, then it currently does not. I'm generally skeptical of the utility of any sub-class that doesn't formally differentiate itself from the parent.
Nevertheless, I agree with your proposition that the classification 'fiat' is not useful unless it assists reasoning. Since http://purl.obolibrary.org/obo/BFO_0000024 currently has no axioms, then it currently does not. I'm generally skeptical of the utility of any sub-class that doesn't formally differentiate itself from the parent.
I mostly agree with this too. I think there is probably some value in keeping the BFO term as it add a differentia to humans readers, but it does not add anything to machine systems or reasoning capabilities.
Perhaps I could suggest two possible options?
-
keep
BFO:0000024but add an axiom to it presumably linking to BFO:two-dimensional continuant fiat boundary -
@cmungall's compromise suggestion of dropping
BFO:0000024but instead "assert this as a has-characteristic axiom at whatever level of ENVO we feel appropriate ... relating to boundaries that we feel is important for some use case[s]"
Thoughts?
Not sure if has characteristic would work as an axiom. Perhaps something like has part some continuant fiat boundary might work too. But, such a change should be communicated with the BFO editors for feedback.
Seems simpler to classify entities under material entity rather than wading into murky issues about what fiat boundaries are.
Do we have any clear examples of fiat objects? military training area may be such a case, since its boundaries are decided upon by fiat. Note, this sense of 'fiat' is different than entities having fuzzy boundaries, which is how Smith et al. using the term 'fiat'.
On Thu, Dec 2, 2021 at 7:24 PM Bill Duncan @.***> wrote:
If we want to classify mounts, deserts, caves, etc. as fiat objects, it would be useful if the fiat nature served some purpose in terms of reasoning. Currently, fiat object doesn't seem to serve such a purpose.=
Exactly! It should serve some purpose. Reasoning is one of many possible use cases
Another use case might be that curators may find it useful to separate fiat objects from true objects. However, this is not true of any curator I know
Maybe there is some other purpose for these upper level divisions we are not aware of?
I always like to think of ontologies in terms of genetics, and consider doing a "deletion test". What is the phenotype of deleting an axiom from an ontology?
If an axiom can be deleted and it doesn't affect any computational use, or any use by human beings, and doesn't affect any projected future use, and if the ontology can be simplified by deleting the axiom, then it should be deleted.
On Fri, Dec 3, 2021 at 12:10 AM Kai Blumberg @.***> wrote:
Nevertheless, I agree with your proposition that the classification 'fiat' is not useful unless it assists reasoning. Since http://purl.obolibrary.org/obo/BFO_0000024 currently has no axioms, then it currently does not. I'm generally skeptical of the utility of any sub-class that doesn't formally differentiate itself from the parent.
I mostly agree with this too. I think there is probably some value in keeping the BFO term as it add a differentia to humans readers, but it does not add anything to machine systems or reasoning capabilities.
Which human readers are served by this differentia? Perhaps there is a community of ENVO users I am not aware of, but I am not aware of any way in which this philosophical subjective distinction serves either end-users or those of us editing the ontology
Perhaps I could suggest two possible options?
keep BFO:0000024 but add an axiom to it presumably linking to BFO:two-dimensional continuant fiat boundary http://purl.obolibrary.org/obo/BFO_0000146
It is bad practice to inject axioms into someone else's ontology
See: https://github.com/OBOFoundry/OBOFoundry.github.io/issues/1443
The route to go here is to request the upstream ontology to make the change. If the upstream ontology is not able to respond to community requests in a reasonable way, then we need to consider whether we want a dependency on that ontology. Every dependency has a combination of benefits and costs, in this case the costs may be higher than the benefits.
But I think your proposal can be adapted, see below
@cmungall https://github.com/cmungall's compromise suggestion of dropping BFO:0000024 but instead "assert this as a has-characteristic axiom at whatever level of ENVO we feel appropriate ... relating to boundaries that we feel is important for some use case[s]"
Yes. The exact nature of the axiom is not particularly important, so long as it doesn't interfere with the hierarchy
If the intended recipients of the axiom are humans, then we should add this as an annotation axiom. For example, a rdfs:comment may be perfectly appropriate
If the intended recipients include machines, then we should add as a logical axiom. This could be something like 'has-characteristic some fiat', where fiat itself has complex axioms pertaining to boundaries. This would have no loss of reasoning power (in fact there are hooks for far more reasoning power, as has been observed the current axiom in ENVO is entailment-silent).
Or it could follow your suggestion above, only we would place the axiom on the ABP class rather than injecting into BFO
E.g.
- ABP SubClassOf material entity
- ABP SubClassOf has_part some BFO:two-dimensional continuant fiat boundary http://purl.obolibrary.org/obo/BFO_0000146
This is logically much better than the current ENVO axioms because we are explicitly linking to the most relevant BFO class.
It is also far superior from a UX perspective, it simplifies the hierarchy and doesn't alienate users (most ontology browsers will show axiom 2 above in a side panel rather than part of the hierarchy)
I still don't think it passes the deletion test, but this seems like a reasonable compromise
On Fri, Dec 3, 2021 at 6:23 AM Bill Duncan @.***> wrote:
Not sure if has characteristic would work as an axiom. Perhaps something like has part some continuant fiat boundary http://www.ontobee.org/ontology/BFO?iri=http://purl.obolibrary.org/obo/BFO_0000140 might work too. But, such a change should be communicated with the BFO editors for feedback.
exactly - but seem my suggestion to move this axiom down to ABP
Seems simpler to classify entities under material entity rather than wading into murky issues about what fiat boundaries are.
Yes, yes, yes!!!! it's a philosophical quagmire/minefield, and takes us into discussions that serve no use case for the ontology
(do quagmires and minefields have fiat boundaries?)
Do we have any clear examples of fiat objects? military training area http://www.ontobee.org/ontology/ENVO?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FENVO_00000561 may be such a case, since its boundaries are decided upon by fiat. Note, this sense of 'fiat' is different than entities having fuzzy boundaries, which is how Smith et al. using the term 'fiat'.
I have no idea what objective criteria to apply to decide whether an ENVO class is fiat or not.
Why are planets considered objects? Isn't the earth's atmosphere part of the earth? Where is the boundaries of the earth? Why isn't everything fiat?
(note: this is rhetorical - I don't think there is a useful answer to the above question)
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/EnvironmentOntology/envo/pull/1252#issuecomment-985560876, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOMQ36BCYS633FUHDY3UPDHHJANCNFSM5IXEJZ3A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
@pbuttigieg how would you feel about @cmungall's suggestion of removing fiat object part but instead adding the following axioms to astronomical body part?
- ABP SubClassOf material entity
- ABP SubClassOf has_part some BFO:two-dimensional continuant fiat boundary
I can only support the proposal to ignore the distinction between fiat object part and object. The question, what characterizes an object, is an open question (I think, the question is not, what a fiat object part is, but rather what a bona fide object is).
BFO 1.0 has characterized object and fiat object part in reference to fiat and bona fide boundaries, with the latter being defined as mind-independent demarcations (i.e., they exist independent of any human partitioning activities) that rest on some physical discontinuity or qualitative heterogeneity. The problem with the second criterion (physical discontinuity & qualitative heterogeneity) is that these are very much granularity dependent - at some granularity level, every physical entity becomes a fiat object part. I have discussed this problem here: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0048603
As a consequence of this critique, in BFO 2.0, the focus has been changed from boundary types to causal unity. But this is still work in progress, I think, because you still find some reference to boundaries (see elucidation for fiat object part). Also,the list of causal unity types suggested in BFO 2.0 is confined to a synchronic approach to causal unity that is associated with a spatio-structural frame of reference - it does not take the dynamic nature of reality into account. I think, a human heart is a functional unit and thus, in a functional frame of reference, a bona fide functional object that exists independent of any human partitioning activities. In the same way, I can demarcate developmental units and evolutionary units. As a consequence, further causal unity types must be added to the list. The main problem here, however, is, that a given entity can be both a fiat object part and a bona fide object, depending on which frame of reference one is applying: spatio-structurally, the human heart is a fiat object part, but functionally it isn't. If you are interested in this discussion, you can read about it here: https://jbiomedsem.biomedcentral.com/articles/10.1186/s13326-019-0196-2
Since this is all very complicated to manage and communicate to users for annotation purposes, I agree with the suggestion to take a pragmatic approach and ignore the categories of object and fiat object part and add respective information as axioms where needed, thus making this classification an implicit one.
I believe @cmungall @wdduncan and myself are in favor of this. However, we don't have a full consensus at the moment.
I'll consider the arguments and make a decision for the next release in the next few days.
Implementation:
🚧 In progress 🚧
- [ ] Modify magnetosphere for illustration on limits of material parthood
- [ ] Incorporate example material part with both fiat and discontinuity-based boundaries
- [ ] Incorporate example of part with entirely fiat boundaries - MPAs, testing sites, ...
- [ ] Modify positioning of environmental zones
- [ ] Establish and document position on fuzzy vs fiat
Do you have a sense of where it impacts, in practice?
I totally agree that it is critical not to break things. So am curious where and how the semantics of 'fiat object part' actually come into play. (Particularly since there is no axiomatization of significance.)
Do you have a sense of where it impacts, in practice?
How we write definitions, classes and branches where boundaries matter and have fiat and/xor non-fiat components, the need to identify what kind of fiat boundaries exist (for axiomatisation but also to link with environmental authorities and capture how humans manage environmental systems).
So am curious where and how the semantics of 'fiat object part' actually come into play.
Where the boundaries of a part of something like Earth are declared by fiat. A marine protected area, chunks of the planet designated for weapons testing (as noted above), etc.
(Particularly since there is no axiomatization of significance.)
This can be said of many upper-level terms that are in wide use. It's also not the primary objective: representation of knowledge is a greater priority relative to whether or not some reasoner understands it (yet). If we define useful subclasses of the fiat boundaries (e.g. designated by national authorities), then this changes
I don't think anyone is arguing that fiat objects don't exist. Rather, the issue that the default operating procedure seems to be that every astronomical body part is categorized as fiat, and this is too strong (for reasons already stated).
I don't think anyone is arguing that fiat objects don't exist. Rather, the issue that the default operating procedure seems to be that every astronomical body part is categorized as fiat, and this is too strong (for reasons already stated).
Yes, that has been demonstrated.