Wikidata:Requests for comment/Best practices for statement ranks for disappeared entities

An editor has requested the community to provide input on "Best practices for statement ranks for disappeared entities" via the Requests for comment (RFC) process. This is the discussion page regarding the issue.

If you have an opinion regarding this issue, feel free to comment below. Thank you!

Current practices for ranks of evolving knowledge, for example the population of a city, is to put the historic numbers as normal rank and the most current one as preferred. It’s the same for a number of cases, such as the country of citizenship (P27) of someone : the country of citizenship s/he has no more are with normal rank, and the country of citizienship(s) the person has at present time are preferred. Help pages summaries this, for example ranking and evolving knowledge (created by the writer of these lines). A first question of this RfC concerns the relevance of the latter page.

What I’m not sure this actually answers the question of ranking of historical informations for historical entities or dead people. A good example is the countries of citizenships of Albert Einstein (Q937) View with Reasonator View with SQID, who were Switzerland and USA at the time of his dead, but had values before.

A case for « keeping stuffs as the time of disappearance of the entity » : it gives a natural way to find the most recent statements for evolving values for a property. If clients wants to get all the datas and not the most recent one they have to explicitely ask for them, just as with existing entities, thus keeping a kind of consistency. Less maintenance is required for when an entity disappears.

A case for « moving the preferred ranks of evolving datas to normal rank » : all the datas are historical. There is no data that is actually valid at present time, for example if we list all the current citizens of Switzerland Einstein is not one of them as he’s dead.

A last option is to not decide anything here, but it implies that things are less predictible for data reusers, which do not make the data reuse easier.

An example of situation in which community asked itself the question is on the spouse property. Should the last spouse be on preferred rank ?


Help:Evolving knowledgeEdit

Is evolving knowledge page a good summary of our practices and a helpful page ? Is it just a duplicate of ranking ? What change should be made to make it better ?

What to do when an entity disappears / for disappeared entitiesEdit

Please comment of the section which is your most preferred choice in the general case, and why.

Option 1 : « keep as isEdit

Keep stuffs as they would have been on Wikidata if Wikidata existed when the entity disappeared or do not move statements ranks (generally) when an existing entity disappears.

  Support I'm leaning to supporting this rule. The preferred status still has a potentially useful meaning in referring to the "final state", or "state at time of death/dissolution". I could be swayed by more comprehensive example cases in favor of the other option though. ArthurPSmith (talk) 14:20, 2 November 2018 (UTC)
  Weak support over switching to normal after dissolution. If disappearance requires some general changes to an item, I feel like this might end up being another area where Wikidata has to take a firm opinion on something we really don't want to, when the destruction/dissolution/death/disappearance of an entity is unclear or disputed. --Yair rand (talk) 05:16, 8 November 2018 (UTC)

Option 2 : « back to normal »Edit

The most recent value for a statement is no more important or no more current than the older one when the entity do not exists anymore. We just move back the « preferred » statement to « normal rank ».

  Support, since once somebody is dead, their citizenship, etc., presumably cease, and should not longer be considered "current". Ghouston (talk) 00:14, 26 October 2018 (UTC)
  Support The maintenance problem can be solved through a bot.--Malore (talk) 12:53, 26 October 2018 (UTC)
  Support No reason to favour eg their last spouse over any others, once they are dead. Jheald (talk) 00:53, 1 November 2018 (UTC)
Hmm, well actually that might be a case where the relation should remain preferred - pensions, inheritance, etc. for a surviving spouse usually depend on if the relation was there at time of death. ArthurPSmith (talk) 13:43, 1 November 2018 (UTC)
I doubt that Wikidata will ever be used to assign such things. Ghouston (talk) 23:29, 1 November 2018 (UTC)
Very likely, but that's still more than "no reason" I think. ArthurPSmith (talk) 14:17, 2 November 2018 (UTC)

Option 3 : no general ruleEdit

Decide this on a property or case by case basis, no Wikidata wide guideline

  •   Oppose. Consistency is important. --Yair rand (talk) 19:27, 25 October 2018 (UTC)
  •   Oppose. +1 to Yair rand. - Jmabel (talk) 01:16, 26 October 2018 (UTC)
  •   Oppose I think I agree we should be consistent on this. ArthurPSmith (talk) 14:20, 2 November 2018 (UTC)

Discussion and other suggestionsEdit

  •   Comment I see good arguments on both sides here; I guess that means it's a good question. I'm wondering if there's some way we can better understand the possible issues with the two different approaches. I see the "current citizens of Switzerland" is one example where the second approach may be preferable - though you could just filter out all the people who have died from that query, so I don't see that making a huge difference. Are there some other good examples of queries where this could be an issue one way or the other? ArthurPSmith (talk) 20:53, 25 October 2018 (UTC)
  • You will still have to filter out the dead if they are changed to normal rank. Ghouston (talk) 23:58, 25 October 2018 (UTC)
  • Option 1 has the advantage that it requires no further action at time of death. - Jmabel (talk) 01:17, 26 October 2018 (UTC)
  • I was wondering about that on Property_talk:P26#Preferred_rank_and_dead_people as well. --- Jura 06:24, 30 October 2018 (UTC)
  • An interesting usecase I found afterward : The case of the production of a car model that is stopped, see . At current time, the production rate is 0 and will always be, as the statement has « no value » for « end date ». Seem pretty natural to keep it at « preferred rank » actually. We can say however that a car model do not cease to exist as it’s not a physical object, it just stop beeing produced. author  TomT0m / talk page 19:51, 20 November 2018 (UTC)