Wikidata:Property proposal/addresses

From Wikidata
Jump to navigation Jump to search

addresses[edit]

Originally proposed at Wikidata:Property proposal/Generic

   Withdrawn
Descriptionwhich need, impairment, process, risk or disaster the subject addresses
Data typeItem
Example 1needs:
Example 2impairments:
Example 3processes:
Example 4risks:
Example 5disasters:
See alsohas cause (P828), medical condition treated (P2175)

Motivation[edit]

The best property we currently have for these statements is has cause (P828). E.g. fire protection (Q41399)has cause (P828)fire (Q3196). While this is a true statement (without fire there wouldn't be any fire protection) the statement does not express that fire protection (Q41399) helps to address the problem of fire (Q3196) ... which is an important detail.

Other properties that are currently in use are:

So I think it makes sense to introduce a new "addresses" property. Possible aliases are "protects against"/"protects from"/"prevents"/"counters"/"reduces"/"helps with"/"helps against".

Note that a similar property that has been proposed is reduces, however that property could not be properly used for many of these examples, e.g:

Also many subclasses of protection aim to completely prevent the risk e.g. fall protection (Q23580296)reducesfalling (Q333495) would not be quite accurate ... the goal is to prevent falling.

--Push-f (talk) 13:57, 18 October 2022 (UTC)[reply]

Discussion[edit]

  •  Support Makes sense to me, and neatly subsumes (the intentional part of) the reduces proposal, and the protects against proposal (which has been withdrawn in favor of this one). --Waldyrious (talk) 18:18, 18 October 2022 (UTC)[reply]
  • Notified participants of WikiProject Influence; WikiProject Properties has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead.. --Waldyrious (talk) 09:23, 24 October 2022 (UTC)[reply]
    I see that a number of "need to" items have been recently created to support some of the examples (not very many, only ten right now).
    While creating items specifically to be used as subjects or objects in claims using existing properties is sometimes necessary, I'd like to warn against creating them before the properties they should be used with actually exist.
    Few properties enjoy the luxury of having items custom-made for them, and I intend to ignore those examples when determining whether the proposed property has the most useful semantics.
    In general, I prefer properties that promote massive interlinking of basic concepts, thereby enhancing Wikidata's usefulness to self-learning AI applications. Creating isolated, trivial link paths between items with complex descriptions don't do that. Consider:
    "Risk X" -P1- "Method to mitigate X" -P2- "Expert on mitigating X"
    "Risk Y" -P1- "Method to mitigate Y" -P2- "Expert on mitigating Y"
    To a self-learning AI system, these two item chains are identical, and will therefore in no way improve that system's "understanding" of X and Y. It's like trying to solve a system of equations:
    X + Y + Z + 0 = 0
    X + Y + Z + 1 = 1
    X + Y + Z + 2 = 2
    X + Y + Z + 3 = 3
    X + Y + Z + 4 = 4
    Regardless of how many such equations you list, the system remains under-specified and cannot be solved, the only conclusion being that X + Y + Z = 0.
    Better then to encode "risk", "mitigate", "method" and "expert" as properties and apply them as appropriate to existing items X, Y, Z etc.
    This makes me a bit wary of the long list of examples, and I find myself unconvinced that "addresses" has the most useful semantics. The label attached to the property is unimportant; it's just a reminder to humans what the property is about.
    I see that a few items indeed appear in more than one claim, which is a good thing, but it clearly doesn't dominate the picture, and I think it's insufficient.
    As few properties lend themselves to the more complex item networks, your proposal is no worse than any else, and I won't object to it. But I will refrain from taking a active position on this one.  Neutral SM5POR (talk) 09:24, 2 November 2022 (UTC)[reply]
    Most of the subjects of my examples are classes with several subclasses/instances. So by just stating e.g. fire protection (Q41399)addressesfire (Q3196) some Wikidata-based Q&A system could infer the answer to the question "What can be done to address the risk of fire?" by simply listing the subclasses of fire protection (Q41399) which currently are fire safety (Q2997557), active fire protection (Q4677559), passive fire protection (Q643227) and cleaning of chimneys (Q41324048).
    I do think that the proposed property lends itself to massive interlinking of concepts. There are hundreds if not thousands of notable concepts that humans came up with to address some other concept.
    Expressing these statements can surely only benefit a self-learning AI system. Yes of course if the AI system does not understand fire (Q3196) that strongly limits the knowledge that can be gained from fire protection (Q41399)addressesfire (Q3196), but I'd argue that this is a different problem. --Push-f (talk) 11:46, 2 November 2022 (UTC)[reply]
  •  Comment This looks very much like has goal (P3712).--Ainali (talk) 04:16, 24 November 2022 (UTC)[reply]
     Withdrawn Thanks! That indeed seems to cover the described use case, along with the object pertains to qualifier I just proposed. --Push-f (talk) 08:12, 24 November 2022 (UTC)[reply]
     Comment I don't think the proposal should have been withdrawn. As I discussed yesterday with Ainali in the Wikidata Telegram chat room, I see the two properties as sort of opposite/complementary. The has goal (P3712) property is about something that the subject is trying to create/expand, whereas the proposed "addresses" property is about something the subject is trying to eliminate/reduce (this is a gross simplification, but it helps frame the discussion IMO).
    Granted, it is certainly possible to have a goal to reduce things, but modeling things using (primarily) the opposite relation of what's intuitive seems inconvenient/inappropriate, not to mention it invites the use of the dreaded "of" qualifier (as shown in the antivirus example above).
    I understand that the object pertains to qualifier would improve things in regards to usage of the "of" qualifier, but it doesn't address (see what I did there? 😄) the issue of modeling things in a convoluted way. In particular, I think
    Are you confident we can find the right relation (preventing, reducing, terminating, working around, satisfying, resolving, combating, etc.) for all the classes of examples shown above? And even if we do, do you think the community would be able to internalize such a complex set of modeling principles in a way that wouldn't cause under-modeling or inconsistent modeling?
    I'll grant that the modeling using has goal (P3712) has the potential to be more precise, but IMO it's too complex to be widely adopted and therefore may actually lead to even less information overall than if we had the "addresses" property. --Waldyrious (talk) 17:17, 24 November 2022 (UTC)[reply]
Retrieved from "https://www.wikidata.org/w/index.php?title=Wikidata:Property_proposal/addresses&oldid=1778233925"