Issue 616: shortcuts in P50 has current keeper, P52 has current owner, P55 has current location
Post by Wolfgang Schmidle (16 October 2022)
Dear All,
I am trying to understand how one can infer the current custody / ownership / location of a Physical Thing / Object.
Let's assume that there has been a E10 Transfer of Custody / E8 Acquisition / E9 Move to an Actor or Place y. If there was no later event at all, it is inferred in the scope notes of P50 has current keeper / P52 has current owner / P55 has current location that y is, in fact, the current keeper / owner / location. For example, the scope note of "P52 has current owner" says: "This property is a shortcut for the more detailed path from E18 Physical Thing through P24i changed ownership through, E8 Acquisition, P22 transferred title to to E39 Actor, if and only if this acquisition event is the most recent."
There is a stronger-sounding but actually weaker requirement that there was no later event that included a "P28 custody surrendered by / P23 transferred title from / P27 moved from" y. The owner / location scope notes use the stronger requirement, the keeper scope note uses the weaker requirement. It would be good to explain in the respective scope notes the reasoning behind this difference.
The FOL encodes the weaker requirement in all three cases. I assume the discrepancy between scope notes and FOL is an oversight. (This was actually my starting point.)
The scope notes not only say "if" but "if and only if". Is there a way to encode the "only if" part in FOL? This seems to be quite tricky. For example, if there were three Moves: 1. from somewhere to A, 2. from A to B, 3. from B back to A, then one can infer that A is the current location, but only Move 3 (and not Move 1) is actually the long form of the shortcut "P55 has current location". On the other hand, it does not follow from Move 1 and 2 that A is not the current location.
Should we worry about negative statements and incomplete knowledge in our knowledge base? Or do we assume here that if there has been such an event, then the knowledge base knows about it? (Or equivalently, if the knowledge base does not know of any such event, then there was indeed none?) Of course one can infer e.g. the current location based on a possibly incomplete list of Moves in a given knowledge base, but whose opinion would it represent? Can one still claim that the inferred statement is the opinion of the knowledge base maintainers?
In particular, what happens if an object disappears or gets destroyed? One may infer the last keeper / owner / location before the destruction, but both the scope notes and the FOL will happily argue that the destroyed object nonetheless has a current owner / keeper / location. Perhaps the destruction implies an implicit Transfer Of Custody where the custody has been surrendered, but there is probably no implicit Acquisition or Move. E64 End of Existence and E6 Destruction offer no concrete help, although E64 says: "It may be used for temporal reasoning about things … ceasing to exist".
I assume this has already been discussed somewhere, but the discussion didn't find its way into the scope notes.
Best,
Wolfgang
Post by Christian-Emil Ore (17 October 2022)
Dear Wolfgang,
It is clear (at least to me) that the FOLs in the 'current' properties are too weak. A complicating factor is that the FOL describes what we explicitly know, that is, the status in the knowledge system. In a closed world system, all shortcuts will imply at least one instance of the corresponding long path. This is not the case in an open world view, I think.
P55(x,y) ⇐ (∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)]
˄ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
If the premise in the FOL above is false, then P55(x,y) is trivially true. This is ok if [E9(z) ˄ P25i(x,z) ˄ P26(z,y)] is false, but it is not ok if
(∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)] ˄ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
is true.
We need an additional axiom, something like
(∃z) [P55(x,y) ˄ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)] ⇒ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]]
?
Best,
Christian-Emil
Post by Wolfgang Schmidle (18 October 2022)
Dear Christian-Emil,
I am not sure I understand your additional axiom. How would it be expressed in normal language? Are you saying "if the knowledge base knows that x has current location y and that were was at least one Move of x, then there must be a Move of x to y after which there is no more Move of x away from y"?
Best,
Wolfgang
Post by Christian-Emil Ore (18 October 2022)
I tried to say that
P55(x,y) ⇐ (∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)]
˄ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
is not sufficient since the above implication is true if the premise is false. So if there exist a newer move ( (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]] is true) it is consistent with P55(x,y). The question is what should the additional axiom be ?
The following is too strong since we do not require knowledge about a move
P55(x,y) ⇒ (∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)]
˄ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
That was what I thought.
Best,
Christian-Emil
Post by Martin Doerr (18 October 2022)
Dear both,
I think the discussion was that the "current" status cannot be inferred, but it is based on a local "closed world" knowledge, and can only be "true" until the time of the last respective update. So, I think the "no other move" since time X, or "no other move without back move" since time X exceeds the scope of logic.
Isn't it?
I fear the "if and only if" statements are wrong anyway. Better you raise an issue. I fear we have not understood circumstances that can lead to a custody or loosing etc., including death, heirs etc.
Best,
Martin
Post by Rob Sanderson (20 October 2022)
Given this, I feel that we should re-open Issue 473: https://cidoc-crm.org/Issue/ID-473-normal-custodian-of
If we cannot infer the current keeper, then nor can we infer the current normal/permanent keeper, despite what we concluded previously.
Rob
Post by Wolfgang Schmidle (19 October 2022)
The shortcuts given in the scope notes / FOLs of P50 has current keeper, P52 has current owner, P55 has current location are wrong. It is unclear how to correct them, or if they can be corrected at all.
Post by Christian-Emil Ore (19 October 2022)
In the scope notes of the "current" properties it is written that, say, P55 has current location, P55(x,y) is the case if and only if the physical object x is not moved from the place y at a later date than it was moved to y. Expressed in ordinary FOL (without any requirement of open world this will be
P55(x,y) ⇔ (∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)] ˄ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
that is, P55(x,y) if and only if there exists a move z such that x was moved to y in this move invent and it does not exist any other move event moving x away from y where the move event z ended before (or with) the start of the move event w. Under an open world view we do not postulate the existence of such a z or require that the knowledge base should contain information about such a (anonymous) event. Therefore the equivalence is weakened to
P55(x,y) ⇐ (∃z) [ [E9(z) ˄ P25i(x,z) ˄ P26(z,y)] ˄ ¬ (∃w) [E9(w) ˄ P25i(x,w) ˄ P27(w,y)˄ P182(z,w)]]
postulating the existence of the shortcut if there exists a move z such that x was moved to y in this move invent and it does not exist any other move event moving x away from y where the move event z ended before (or with) the start of the move event w.
This implication is true also in the case the premise is false, that is, there exists a newer move event taking x away from y. Therefore our knowledge base will be consistent even in this case which contradicts the scopenote text. So that is why I suggested an additional weaker axiom:
Ff the shortcut P55(x,y) exists and there is information about the move event moving x to y, then there cannot be information about any other move event moving x away from y at a later point in time.
May be I am wrong.
Best,
Christian-Emil
Post by Carlo Meghini (19 October 2022)
Dear Christian-Emil
I don't think that the axiom contradicts the scopenote text. In those worlds where there is a move w taking x away from y and w is after z (so the antecedent is false), P55(x,y) (the consequent) is false as well but the axiom still holds.
Carlo
Post by Christian-Emil Ore (19 October 2022)
Dear Carlo,
If the scopenote text describe the a world or a model for the theory, it is true. More formalistic: If I add this second move to a knowledge base, then there are no formal way to detect that the knowledge base is inconsistent?
Best,
Christian-Emil
Post by Carlo Meghini (19 October 2022)
I think there is no inconsistency, the axiom always derives the current location of the object. At different times, the object may be located in different places, but this is no inconsistency, as the current location is never written into the KB, it is always derived (via the axiom) on the basis of the story that is written in the KB. So at 5PM the current location is Rome and at 6 PM (after you add the second move) the current location is Pisa, and that is fine.
If I got everything all right, I mean.
Carlo
Post by Christian-Emil Ore (28 November 2022)
Dear all,
Wolfgang points to the fact that the 'current' properties is not defined in a consistent way, which of course they should have been. The textual scope notes says 'if and only if' which should be expressed as bidirectional implication, ⇔ (equivalence). Below I quite from an email exchange between Carlo and me. This may explain the issue:
- C-E:
There are several axioms in CRM of the form lefthandside(x,y)⇒ (∃z)[righthandside(x,y,z)], which is not a good thing, if I understand you right, due to efficient machine reasoning and the time it will take to find a needle z in the haystack.
- Carlo:
Precisely. The computer enters into a combinatorial examination of cases and basically may never come back.
and Carlo writes earlier in his reply, about claiming the existence of some individual on the right hand side of the implication:
' it's not outside of the model, it's that we do not know what to do with it, as you said, having a bunch of these "unknown" guys in the KB breaks efficiency. I understand that efficiency is an engineering issue, but in the end, we are engineers. '
In my earlier days when I worked with formal logic and models, I didn't care very much about efficiency. However, I fully understand Carlo and also see the point that we formulate the FOL so that it can be efficiently computable. This is one reason to drop the bidirectional implication implication in the properties P50, P52 and P55.
There is also another issue. The current properties P50, P52 and P55 need external curation and also break the basic assumption that a CIDOC-CRM KB/database store accumulate history. It would have been better if the current properties are implemented as named and stored queries with the ting/person as argument. The original reason for introducing the current properties was they were used in some museum databases in the 1990ies,. Maybe it is time to get rid of them?
Best,
Christian-Emil
P50 has current keeper
This property is a shortcut for the more detailed path from E18 Physical Thing through, P30i custody transferred through, E10 Transfer of Custody, P29 custody received by to E39 Actor, if and only if the custody has not been surrendered by the receiving actor at any later time.
FOL:
P50(x,y) ⇐ (∃z)
P52 has current owner
This property is a shortcut for the more detailed path from E18 Physical Thing through, P24i changed ownership through, E8 Acquisition, P22 transferred title to to E39 Actor, if and only if this acquisition event is the most recent.
FOL:
P52(x,y) ⇐ (∃z)
P55 has current location
This property is a shortcut. A more detailed representation can make use of the fully developed (i.e., indirect) path from E19 Physical Object, through, P25i moved by, E9 Move, P26 moved to to E53 Place if and only if this Move is the most recent.
P55(x,y) ⇐ (∃z) [ [E9(z) ⋀ P25i(x,z) ⋀ P26(z,y)] ⋀ ¬ (∃w) [E9(w) ⋀ P25i(x,w) ⋀ P27(w,y)⋀ P182(z,w)]]
Post by Detlev Balzer (28 November 2022)
Dear all,
I fully agree with Christian-Emil's observation that
The current properties P50, P52 and P55 need external curation and
also break the basic assumption that a CIDOC-CRM KB/database
store accumulate history.
This is a point I had to raise over and over again in discussions about database design: there is no notion of "now" if we are dealing with persistent data.
Maybe it is time to get rid of them?
Definitely. The reason why the idea of a "current" state of affairs is so deeply rooted in the database world seems to come from the fact that almost any textbook on the subject has silly examples such as "student: { name: Carla Jones, age: 23 }".
Best,
Detlev
Post by Martin Doerr (28 November 2022)
Dear All,
I would also suggest to deprecate them, if there is no community actively using them. Alternatively, we may think of a general, efficient mechanism to assert that a property is still valid? This may go into the Situation discussion.
Best,
Martin
In the 55th joint meeting of the CIDOC CRM and SO/TC46/SC4/WG9; 48th FRBR/LRMoo SIG meeting, CEO gave a summary of the issue:
- the FOL notation for P50/P52/P55 (cast as a weak shortcut) is inconsistent with what is stated in the scope-notes (cast in terms of “if and only if”).
- he weak shortcut notation implies the existence of something unknown, which from an implementation point of view is not good, but from an ontological point of view should make no difference.
- “current” properties need external curation and also break the basic assumption that a CIDOC CRM KB/database store accumulate history. CEO proposed to deprecate them. They were originally introduced because they were used in museums back in the 90s. Time to re-examine this.
Discussion: “Current” relates to the definition of Situation in a sense, and specifically to the statements in the knowledge base that its maintainers claim to know. It is a question for CRMinf really.
How to proceed:
- Before resolving to deprecate the properties, there should be an alternative, i.e., a way to maintain and exchange knowledge about the current validity of what has been documented.
- Start a new issue (634) to discuss the requirements for expressing current knowledge, and once there is agreement on that, consider the repercussions of deprecating P50/P52/P55 in the CRM.
Issue closed
Belval, December 2022