Wikidata:Property proposal/inferred from

inferred from statements on

Originally proposed at Wikidata:Property proposal/Generic

   Done: inferred from (P3452) (Talk and documentation)

Description	(use as reference) statement added based on related statement found on the following item
Data type	Item

Motivation

Currently we have a bots that automatically add inverse statements. They sometimes use "stated in". I think it would be worthwhile to have a more specific property for the use case. ChristianKl (talk) 10:07, 27 September 2016 (UTC)

Discussion

•  Comment: it can be useful, but such a source is easily lost. --AVRS (talk) 12:16, 27 September 2016 (UTC)

Especially in cases where the source it's lost, having a "inferred from" property could help with recognizing that the statement isn't justified anymore. Those are cases where data provenance is important. ChristianKl (talk) 14:35, 27 September 2016 (UTC)

•  Comment - this doesn't sound like a good idea to me. Can you give a specific example or two to illustrate though? ArthurPSmith (talk) 18:48, 27 September 2016 (UTC)

Let's say we have a property Alice (Qa) and a property Bob (Qa). We have the statement `Qa child (P40) Qb`. Now a Bot like User:LandesfilmsammlungBot comes along and wants to add the statement `Qb father (P22) Qa`. The current behavior of that bot is to give as a reference `stated in (P248) Qa` for such a claim (it also adds retrieved (P813). I think that's using stated in (P248) in a way it isn't designed to work and it would be good to have an alternative that make the data provenance more clear. ChristianKl (talk) 21:24, 27 September 2016 (UTC)

•  Support we had various approaches suggested for this and each time someone adds statements based on this some other group wants yet another solution. I think a more explicit label might even work better, e.g. "statement added based on related statement found on the following item". I don't think it should be used for properties other than symmetric or inverse ones. This new reference-qualifier could replace Q20651139.
  --- Jura 06:56, 5 October 2016 (UTC)

@Jura1: I think your sentence is a bit too long for a property name. I modified the name and added it as description. ChristianKl (talk) 09:50, 5 October 2016 (UTC)

•  Support. Thryduulf (talk) 12:15, 6 October 2016 (UTC)
•  Oppose. Automatic inverse statements should just copy the actual source, rather than referencing other Wikidata items. If there is no source, than perhaps the inverse shouldn't be automatically created in the first place. --Yair rand (talk) 18:11, 6 October 2016 (UTC)

@Yair rand:Simply copying statement means that there's no provenance that allows a data user to know that a bot created the entry based on reasoning about other Wikidata items. That especially an issue when the subject is more complex that a simple inverse statement. If you infer that Alice is the mother of Bob based on the fact that Alice is female and Bob is her child, that might be an error in cases of transgender individuals (e.g. the source might say that Alice is the father of Bob). ChristianKl (talk) 19:58, 6 October 2016 (UTC)

• Hesitant. It would be better if we could link directly to a real statement, or all this will look like a dirty hack. This will easily get ambiguous on which claim was used, so we actually would need to duplicate the claim main snak at least in some source of ours or something ... author  TomT0m / talk page 19:48, 12 October 2016 (UTC)
  • Could we use uses property (P3176) as a qualifier to record which statements were used to infer from? Thryduulf (talk) 16:53, 13 October 2016 (UTC)
    • This would not be enough, we would have to store also the value as there could be several statements with the same property. But this still could be several statements with the same property/main value pair (main snak) ... so maybe in some cases we would also have to store some qualifier snaks ... in the end this amout to copy the whole statement which is not really practicable and difficult to implement. Of course maybe a partial solution could be implemented right now but a more close examination of the needs of something like that should occur - and this is in the scope of WikiProject Reasoning so maybe the discussion should happen there. author  TomT0m / talk page 17:12, 13 October 2016 (UTC)
      • The information might not be completely comprehensive but it would still provide more information about the provenance than the status quo. We could have an additional property like "Automatic reasoning schema" that a bot can use if we want to provide more detailed information. ChristianKl (talk) 12:02, 22 October 2016 (UTC)
        • @ChristianKl: A usecase where this would actually be useful would help. Do you have something in mind ? If the only usecase is "complete information for a symmetric property", it's obvious that the item in case is the suject of the claim. Knowing the rule is enough to know that the item has been inferred either from statement in the item of a parent or from a brother/sister. Thinking about this, it would probably be more useful to have an item for the rule of inference where the rule is described and a property to link to this item, like we have a "based on heuristic (P887) " - that could be used in this case, by the way.

          (also, rereading the proposal, it should be more appropriate to say to the bot owners who does that to just stop putting items in "stated in") author  TomT0m / talk page 13:33, 22 October 2016 (UTC)

 Support. CC0 (talk)