Wikidata:Property proposal/context of death
context of death edit
Originally proposed at Wikidata:Property proposal/Person
Description | the person died in the societal context |
---|---|
Data type | Item |
Domain | human (Q5) |
Allowed values | instance/subclass of occurrence (Q1190554) |
Example 1 | George Floyd (Q95677819) → US police violence (Q96442197) |
Example 2 | Tamir Rice (Q57339968) → US police violence (Q96442197) |
Example 3 | Anne Frank (Q4583) → The Holocaust (Q2763) |
Example 4 | Wat Tyler (Q311820) → Peasants' Revolt (Q498871) |
Example 5 | John Jacob Astor IV (Q430672) → sinking of the RMS Titanic (Q2577588) |
Example 6 | Wilfred Owen (Q212719) → Battle of the Sambre (Q1345169) |
Example 7 | Michèle Kiesewetter (Q94744511) → National Socialist Underground murders (Q369336) |
Example 8 | James Scurlock (Q97328628) → George Floyd protests (Q95965839) |
Motivation edit
Prior discussion: https://www.wikidata.org/wiki/Wikidata:Project_chat#How_should_we_note_that_George_Floyd_(Q95677819)_was_killed_by_police?
We currently have no good way to model the context in which a person died. As a result it's not listed on items such as George Floyd (Q95677819) which makes it hard to query for all instances of US police violence. Sometimes constructions with instance of (P31) are used for holocaust victims but it would be more desireable to not overload instance of (P31) for those as well and have a decidated property.
I used the new item US police violence (Q96442197) to be able to neutrally describe a situation without making a judgement about whether the use of force is excessive (given that's criminal, such claims should ideally be sourced with court judgements). It might be possible to import data about fatal shootings from https://github.com/washingtonpost/data-police-shootings which make no judgement about whether or not the use of force was justified or excessive.
I put the country in the item because it's not just important that a US citizen was killed but also that the killing happened in the US. We might also have narrower items for individual cities. ChristianKl ❪✉❫ 11:07, 19 June 2020 (UTC)
Discussion edit
- Support. I've felt we needed something like this for a long time. I've added a couple more examples of how it could be used - someone who died during a specific battle, a rebellion, a disaster. Some items use eg "cause of death" for things like this but that's not strictly right, I think. Andrew Gray (talk) 12:31, 19 June 2020 (UTC)
- Those examples seem good. ChristianKl ❪✉❫ 22:39, 19 June 2020 (UTC)
- Neutral; sounds a bit vague. Nomen ad hoc (talk) 16:59, 19 June 2020 (UTC).
- Comment. I have never fully understood the difference between cause of death (P509) and manner of death (P1196), having always been trying to understand which one of the two is supposed to do what the proposal suggests a third property should. So I guess I have my answer now – none of them so far! Thierry Caro (talk) 20:56, 19 June 2020 (UTC)
- There's prior art in http://sparql.cwrc.ca/ontology/cwrc.html#DeathContext . Given our general naming I like expressing it like I did in the proposal. ChristianKl ❪✉❫ 14:14, 20 June 2020 (UTC)
- Comment for a specific event it makes sense, like a battle, however generally "(US) Police violence" doesn't seem to fit, why not George Floyd protests (Q95965839) ~~
- George Floyd (Q95677819) didn't die in George Floyd protests (Q95965839). His death caused George Floyd protests (Q95965839) which is a very different relationship. It also doesn't help with the use-case of querying for people who died to US police violence. ChristianKl ❪✉❫ 14:58, 20 June 2020 (UTC)
- Yes true, I was just seeing it the social context, so why US police violence, not police violence in general or why not Black Lives matter Germartin1 (talk) 16:52, 20 June 2020 (UTC)
- Given subclass relationships you can query for police violence (Q96441150) even when the item is noted as US police violence. If you only care for Black victim you can narrow your query via ethnic group (Q41710). I consider it very desireable that there's a clear way to query for all deaths by police violence in a certain city and that would not be possible by setting the context as Black Lives matter. That said, I don't think there should be a single value constraint and it's fine if somebody who likes to think in terms of Black Lives matter and maybe also care for violence that's not directly done by police officiers where Black people are victims to set Black Lives matter as an additional value. ChristianKl ❪✉❫ 17:27, 20 June 2020 (UTC)
- Yes true, I was just seeing it the social context, so why US police violence, not police violence in general or why not Black Lives matter Germartin1 (talk) 16:52, 20 June 2020 (UTC)
- George Floyd (Q95677819) didn't die in George Floyd protests (Q95965839). His death caused George Floyd protests (Q95965839) which is a very different relationship. It also doesn't help with the use-case of querying for people who died to US police violence. ChristianKl ❪✉❫ 14:58, 20 June 2020 (UTC)
- Support more to the point than "significant event" that is currently used a lot to model this. All for it! Moebeus (talk) 16:43, 22 June 2020 (UTC)
- Comment I'm not sure. I would use it for the project I'm working on collecting profiles from newspapers on Covid-19 victims but I think I would also continue to add it as a significant event since there are also profiles on people who recover, and it remains a significant event to those they leave behind. The entire group of people impacted by an event can be tagged the same way. That a significant event strongly correlated with the person's death can already be deduced from the proximity in time. 1Veertje (talk) 09:36, 1 July 2020 (UTC)
- I think systematic use of "significant event" would work as an alternative, but only if we can work out a consistent qualifier to clearly say "death connected to this event", and I don't think we currently have that. Relying on making the connection by date would get tricky for things with very long time-spans (eg it would be good to record people who died in the context of The Troubles (Q815436), but that lasted 30+ years) and might not be possible for generic events without specific timeframes like "police violence". Andrew Gray (talk)
- Comment I like the line of thinking. I do feel that it would be good to restrict the allowed values somehow, but can't figure out the common link between discrete events like battles/disasters and more general things like US police violence (but agree that they both fit in). --SilentSpike (talk) 22:35, 10 July 2020 (UTC)
- They're all instances or subclasses of occurrence (Q1190554) which might be a good fit. --SilentSpike (talk) 22:41, 10 July 2020 (UTC)
- @SilentSpike: I added it to the proposal. ChristianKl ❪✉❫ 08:57, 11 July 2020 (UTC)
Supportwith this clarified --SilentSpike (talk) 09:38, 11 July 2020 (UTC)
- @SilentSpike: I added it to the proposal. ChristianKl ❪✉❫ 08:57, 11 July 2020 (UTC)
- They're all instances or subclasses of occurrence (Q1190554) which might be a good fit. --SilentSpike (talk) 22:41, 10 July 2020 (UTC)
- Strong oppose in the current state. I don't think it's well defined. Example 4-6 are ok, and could be called "died in battle/event". Anne Frank could also have the values history of the Jews during World War II (Q2043277), or Final Solution (Q127013), George Floyd could also have colorism (Q5282081). Which one is it? Nepalicoi (talk) 13:46, 12 July 2020 (UTC)
- @Nepalicoi: Just to point out, as per the allowed values constraint, Anne Frank could not have a value of history of the Jews during World War II (Q2043277). You otherwise make raise a valid point, though I don't see why the property couldn't have multiple valid values? --SilentSpike (talk) 22:35, 13 July 2020 (UTC)
- Neutral Double posting, but I think the problematic cases are really the ones where the value is just a subclass (US police violence/discrimination against black crime victims). I suppose strictly those would not be the context of a death, but more of a category a death might fall into (for all other examples you can say "X died in Y", but not those ones). However, I see the original proposal intention in that it would be good to be able to easily link deaths to such topics. Not sure how to do so though. --SilentSpike (talk) 22:49, 13 July 2020 (UTC)
- I would Support this property as "died in event", what do you think? let's keep this discussion going, I think there is a need for this kind of property. Germartin1 (talk) 11:00, 25 August 2020 (UTC)
- Support. Good job! I was also trying to figure out how best to be able to aggregate deaths based on a particular event or situation. This seems like the best way so far. We just have to standardize, and have a larger selection of examples. --RAN (talk) 01:38, 1 September 2020 (UTC)
- Comment I support the idea but also think that ‘societal context’ sounds vague. I made up a quick list of possible case groups:
- genocide:
- Anne Frank (Q4583) → The Holocaust (Q2763) (example #3 from above)
- other killings by the State or similar parties:
- wars, revolts, short-term or long-term conflicts and similar cases
- Wat Tyler (Q311820) → Peasants' Revolt (Q498871) (example #4 from above)
- Pat Finucane (Q891652) → The Troubles (Q815436)
- specific battles
- Wilfred Owen (Q212719) → Battle of the Sambre (Q1345169) (example #6 from above)
- specific disasters or similar events
- John Jacob Astor IV (Q430672) → sinking of the RMS Titanic (Q2577588) (example #5 from above)
- James Scurlock (Q97328628) → George Floyd protests (Q95965839) (example #8 from above)
- serial killings
- Michèle Kiesewetter (Q94744511) → National Socialist Underground murders (Q369336) (example #7 from above)
- cases like US police violence (Q96442197)
- George Floyd (Q95677819) → US police violence (Q96442197) (example #1 from above)
- Tamir Rice (Q57339968) → US police violence (Q96442197) (example #2 from above)
- However, there are other cases that could be considered a context of death:
- Similar to “specific disasters” one might think of instances where the underlying context of a death is a tsunami (Q8070) or a flood (Q8068).
- Similar to cases like US police violence (Q96442197), there are other forms of underlying systemic causes for death:
- Marcus Omofuma (Q1666764) died of asphyxia after Austrian police officers gagged him during a deportation flight. A context of death = deportation (Q379693) might apply here.
- Death as a result of a specific activity:
- Ernst Balcke (Q99791) drowned while ice skating (Q779272) and
- Georg Heym (Q61601) drowned while attempting to save him (context of death = rescue operation?)
- Death during and as a result of a specific event:
- Nodar Kumaritashvili (Q193690) died during the 2010 Winter Olympics (Q9674)
- We should also consider if we want to accept reasons of suicide (if they are well-documented):
- persons driven to suicide by Nazi persecution like Hermann Umfrid (Q1498304) (killed himself to avoid persecution of his family), E. O. Plauen (Q67009) (killed himself to avoid a sham trial) or Walter Benjamin (Q61078) (killed himself out of desperation because he saw no way to flee Nazi persecution)
- underlying illnesses like major depressive disorder (Q42844) (but there are more complex cases: Ferdinand Raimund (Q45025) shot himself after being bitten by a dog because he erroneously thought the dog had rabies and he had a pathological fear of catching rabies)
- With the possible exception of #8/5, I do think that it would make a lot of sense to be able to query those contexts of death. --Emu (talk) 11:54, 16 November 2020 (UTC)
- genocide:
- Oppose. The property is sufficiently ill-defined that it will likely cause a mess. See also Help:Modeling causes and relevant discussions such as Property_talk:P828#A_better_way_to_model_causation and the discussion way back when in the project chat about modelling causes of death (can't find the link, sorry). There is a gap in the properties to be filled, but this current proposal does not do an adequate job of precisely defining boundaries. --Yair rand (talk) 04:50, 22 December 2020 (UTC)
- Comment I support the development of this concept, and I want to support a well-defined property like this, but I agree that this property seems to require extraordinary guidance on its usage restriction. If there were a way to get the benefits and prevent the drawbacks then this seems useful. Blue Rasberry (talk) 22:47, 13 May 2021 (UTC)
- I agree that the current properties aren't ideal for this and a new property could help. Maybe we could tackle some of the points raised above with a short decision tree:
- can we reference the statement (with one or several source combined): yes = proceed to #2, no = the property isn't applicable
- is the concept used as value used elsewhere (reference about the value): yes = proceed to #3, no = the property isn't applicable
- is it impossible to query the same with statements one or several other properties (yes= use these instead, no= use the new property)
- Samples:
- cancer death: #1/#2 we have a reference for both, #3: use "cause of death",
- red heads killed on Mondays: #1 references are available, if #2 would be a yes, #3 would indicate to query manner of death, date of death and hair color instead. --- Jura 17:15, 14 May 2021 (UTC)
- Strong oppose. The 'context of death' is rarely an obvious thing. Anne Frank, for instance, was clearly a victim of the Holocaust, but equally validly the victim of antisemitism in general, German antisemitism in particular, xenophobia at its most general and Nazism in particular. And that's a clear-cut case in comparison with instances where one context is the pretext or means of another, such as a regime forcibly conscripting a certain ethnic group, or where multiple parties have different contexts to effect someone's death. For the latter, consider the killing of Hafizullah Amin (Q63831) as part of Tajbeg Palace assault (Q975430): the proximate contexts alone are numerous – the dispute between the Khalq and the Parcham, rivalries inside the Khalq, the Soviet Union's desire for a more reliable leader in Afghanistan, Soviet expansionism and so on. The biggest problem is that these reasons can often be quite controversial in a way that is not capable of sensible resolution. If Country A attacks Country B in retaliation for Country B's claimed violations of its sovereignty, is the death of an A-ese soldier at the hand of a B-ian unit taking place in the context of 'B's aggression against A', the 'A-B war' or 'A's aggression against B'? There's a Pandora's box here that is best left unopened. Equally, there might be quite valid disagreement as to the attribution of systemic causes. Bonnie and Clyde were in a sense victims of police violence, and in a sense their deaths also took place in the context of a crime spree they were wholly responsible for. I am unsure if this concept, as broad as it is, could be useful. Narrower incarnations, such as 'died at/in' (for battles, events, terrorist attacks and the like) could cover some ground. So could 'notable events'. The issue of deaths is already controversial enough, e.g. a number of people murdered during purges or the Holocaust were killed under the color of law and so their manner of death lists 'capital punishment' when arguably murder might be more appropriate. Including the context of death would invite a much more complex situation. Ari T. Benchaim (talk) 00:21, 17 July 2021 (UTC)
- @Ari T. Benchaim: how would your samples pass through the tests suggested above? --- Jura 16:43, 17 July 2021 (UTC)
- The fact that you can make broader and less broad claims is generally true for a lot of relationships in Wikidata. We usually take by convention the most specific one. ChristianKl ❪✉❫ 21:10, 17 July 2021 (UTC)
- @ChristianKl: I appreciate that. My key concern is that there can be more than insignificant disagreement - and valid disagreement at that - as to what the most specific cause is. Is the Nazi repression of dissent or the German resistance the more 'proximate' context of the Scholls' deaths, for instance?
- Not done Too many open questions, lack of consensus --Emu (talk) 21:47, 13 August 2022 (UTC)