Wikidata:Property proposal/fails compliance with
fails compliance with[edit]
Originally proposed at Wikidata:Property proposal/Generic
Description | fails compliance of the test defined in the associated item |
---|---|
Data type | Item |
Example 1 | Hackers (Q13908) → "fails compliance with" → Bechdel test (Q4165246) |
Example 2 | instant-runoff voting (Q1491219) → "fails compliance with" → monotonicity criterion (Q6902035) |
Example 3 | Copeland's method (Q5168347) → "fails compliance with" → independence of irrelevant alternatives (Q3150644) |
Motivation[edit]
We need the opposite of complies with (P5009) to state when an item doesn't comply with the criterion associated with an item. For example:
- We have a way of stating that Beauty and the Beast (Q19946102) complies with (P5009) the Bechdel test (Q4165246). But the only way to say that Hackers (Q13908) fails to comply with the Bechdel test (Q4165246) is to state that Hackers (Q13908) has characteristic (P1552) Q45172088. In this case, a workaround was employed (the "Q45172088" item).
- We can state that Copeland's method (Q5168347) complies with (P5009) the monotonicity criterion (Q6902035). But we can't currently state that instant-runoff voting (Q1491219) fails compliance with the monotonicity criterion (Q6902035) (per the w:Comparison of electoral systems#Comparisons table), since we don't yet have a "fails Monotonicity criterion" item.
- Speaking of the w:Comparison of electoral systems#Comparisons table, there are many columns in that criteria compliance table. I'm hoping we don't need to define "fails ___ criteria" for each column of that table to express the contents of that table.
Could someone create a "fails compliance with" property? -- RobLa (talk) 02:49, 31 January 2020 (UTC)
Discussion[edit]
Summary of Wikidata:Project_chat/Archie/2020/01 conversation: this is what was discussed over at Wikidata:Project chat/Archive/2020/01 as part of the "Reasonably quick way to resolve 'non-compliance property' issue?" topic back in January/February. User:Jura1 suggested modeling this after assessment (P5021) and test score (P5022), allowing us to specify other options besides "complies", "doesn't comply". User:Ghouston suggested instead that we can add qualifiers to this new property if we need to move beyond binary compliance/non-compliance. User:Ls1g suggested a change to the data model to allow statements which negate any existing property, and links to this paper: "Negative Statements Considered Useful" - Hiba Arnaout, Simon Razniewski, and Gerhard Weikum.
- Comment - Please edit the summary above if you believe there is a problem with it. I'd still prefer taking User:Ghouston's approach as I understand it, which would mean creating a "Fails compliance with" property. -- RobLa (talk) 05:21, 1 April 2020 (UTC)
- deprecated summary from 2020-02-01):
this is what was discussed so far over at Wikidata:Project_chat as part of the "Reasonably quick way to resolve 'non-compliance property' issue?" topic. User:Jura1 suggests modeling this after assessment (P5021) and test score (P5022), allowing us to specify degrees of compliance. User:Ghouston suggests instead that we can add qualifiers to this new property if we need to move beyond binary compliance/non-compliance. User:Ls1g suggests a change to the data model to allow statements which negate any existing property, and links to this paper: "Negative Statements Considered Useful" - Hiba Arnaout, Simon Razniewski, and Gerhard Weikum. I'd prefer taking User:Ghouston's approach as I understand it. -- RobLa (talk) 05:15, 1 February 2020 (UTC)
- I don't think "specifying degrees" should be done. Seems silly. See Wikidata:Project_chat/Archive/2020/01#Reasonably_quick_way_to_resolve_"non-compliance_property"_issue? for my actual comments. --- Jura 08:57, 1 February 2020 (UTC)
- @Jura1: - Your comments at "Wikidata:Project chat/Archive/2020/01#Reasonably_quick_way_to_resolve_"non-compliance_property"_issue?" weren't clear to me at the time we had that discussion. I corrected it above (as I understand it), updated the links to point to the archive, and decoupled the summary from my comment (immediately below). I would appreciate another evaluation from both you and @Ghouston:, as well as any further corrections to my updated summary. RobLa (talk) 05:21, 1 April 2020 (UTC)
- I don't think "specifying degrees" should be done. Seems silly. See Wikidata:Project_chat/Archive/2020/01#Reasonably_quick_way_to_resolve_"non-compliance_property"_issue? for my actual comments. --- Jura 08:57, 1 February 2020 (UTC)
- I do Support this property as preferable (and practical) compared to alternatives. Ghouston (talk) 06:25, 1 February 2020 (UTC)
- Oppose for the proposed usecase. See the discussion at Wikidata:Project_chat/Archive/2020/01#Reasonably_quick_way_to_resolve_"non-compliance_property"_issue?. The mentioned table is not binary. --- Jura 08:59, 1 February 2020 (UTC)
- @Jura1:, can you give details of how you think it should be done? You want to replace complies with (P5009) with a new property, such that there's only a single property for defining compliance? Then the statement itself would be meaningless without interpreting the qualifiers, which would make it harder to write queries. Ghouston (talk) 09:56, 3 February 2020 (UTC)
- I think the explanation on project chat is fairly clear. Please comment there if you think it needs more input. --- Jura 09:58, 3 February 2020 (UTC)
- @Ghouston:, thank you bringing the conversation to this proposal page. I agree with you that queries seem a lot easier with the addition of "fails compliance with" than queries involving a new regime modeled after assessment (P5021) and test score (P5022). This page seems like a better place to discuss alternatives to "fails compliance with" than the omnibus project chat page. -- RobLa (talk) 18:08, 3 February 2020 (UTC)
- @Jura1: I don't think it's correct to say that compliance is not binary. A voting system either complies with a criterion or it doesn't; there is no in-between. The reason that table has cells other than Yes or No is because it combines closely-related voting systems into the same row, and closely-related criteria into the same column. In other words, some of the rows in that table are actually classes of voting systems rather than instances of voting systems. Omegatron (talk) 18:56, 9 February 2020 (UTC)
- @Jura1:, can you give details of how you think it should be done? You want to replace complies with (P5009) with a new property, such that there's only a single property for defining compliance? Then the statement itself would be meaningless without interpreting the qualifiers, which would make it harder to write queries. Ghouston (talk) 09:56, 3 February 2020 (UTC)
- Oppose. There should be a better data model for both this and complies with (P5009) in such cases. Jura1's suggestion looks workable. --Yair rand (talk) 21:47, 5 February 2020 (UTC)
- Is there any precedent for a statement that's meaningless without interpreting the qualifiers? I can't think of one, but I don't know them all. It would be like X <compliance> Y. Ghouston (talk) 22:21, 5 February 2020 (UTC)
- @Ghouston: disjoint union of (P2738) only allows list of values as qualifiers (Q23766486) as a value. --Yair rand (talk) 22:48, 5 February 2020 (UTC)
- I don't think it's meaningless. It's like the "test taken" mentioned on Project Chat or "significant event". We know that's a valid criterion/test/event for the item, we just don't have full details. The approach seems more suitable for non-binary content like the voting systems description that is planned.
Also, I don't get why Bechtel test is mentioned. It isn't even used with the other property. --- Jura 23:03, 5 February 2020 (UTC)- Bechtel test is an example on complies with (P5009). Would new properties need to be defined as qualifiers, or would existing qualifiers suffice? What would Mercedes-Benz OM646 engine (Q1921277) complies with (P5009) EURO3 (Q13424865) be replaced with? Would there be multiple ways of saying that, e.g., conforms: yes and conformance: 100%? Would that make it harder to write queries? Ghouston (talk) 23:35, 5 February 2020 (UTC)
- Is there any precedent for a statement that's meaningless without interpreting the qualifiers? I can't think of one, but I don't know them all. It would be like X <compliance> Y. Ghouston (talk) 22:21, 5 February 2020 (UTC)
- Support - These are binary criteria, and so "test score" or "degree of compliance" or other qualifiers are irrelevant and make queries overly complex. "change to the data model to allow statements which negate any existing property" might be fine, too, but maybe more complicated to use. Omegatron (talk) 14:20, 20 December 2020 (UTC)
- Weak support. While "compliance" is binary, Wikidata properties explicitly allows "unknown" and "no value", and implicitly "we don't care" / "there is no verifiable info about this" (by not including a statement). So it will be helpful to have a property to record notable instances of test failures. I also think we should require references for uses of this property so we don't collect irrelevant "failures" and can back ourselves up if people complain we're bad-mouthing them. ("Weak" because I won't have time to help curate this property, but I support whoever wants to do this.) Deryck Chan (talk) 14:22, 1 January 2021 (UTC)
Notified participants of WikiProject Media Representation
- Comment We spoke about this at Wikidata_talk:WikiProject_Media_Representation#Fails_compliance_with and we would rather use Yair rand's and Jura's method (using assessment (P5021) or another, similar property that is not restricted to humans with a new qualifier
outcome
ortest result
). While Bechdel test results may be modelled using complies with/fails compliance with, this is not true for many other tests (we need a "not applicable" for some tests and some tests don't deliver a "fail" or "pass" result but just a score). The assessment (P5021) andtest outcome
approach has the advantage to be more flexible. Please have a look at the model presented at the discussion mentioned above to have an overview about tests that don't fit into the complies with/fails compliance with model and how we could model them after assessment (P5021) andtest outcome
. So please don't use the Bechdel test (Q4165246) as an example for this property as we are not going to model Bechdel test results like that.
- That said I won't oppose this property. I even think that
complies with
/fails compliance with
could be useful for us as a qualifier. There are tests that require certain criteria (e.g the Riz Test) or a number of criteria (e.g. the Feldman test) to be met. To be able to make it explicit which criteria are met and which not is, besides other advantages, necessary so that we can meet a minimal standard of sourcing (e.g. while there will be few sources for a certain film that it meets the Feldman test there are enough sources which can be used to decide if a film has a female screenwriter). We are currently using criterion used (P1013) to record the criteria which were crucial for the outcome but this does not fit nicely all tests - sometimes it may seem a bit shoehorned. So could we use this as a qualifier? In this case I would support this. @RobLa: (as this proposal is quite old...) - Valentina.Anitnelav (talk) 12:45, 5 January 2021 (UTC)
- To me as Valentina.Anitnelav points out, assessment (P5021) could be used instead, especially with assessment outcome (Property:P9259). Are there things that cannot be described with assessment (P5021) and assessment outcome (Property:P9259)? For instance if a hardware component tried to respect a standard and failed we could add that it failed compliance with that standard according to <the organization that reviewed it> GNUtoo (talk) 08:55, 3 February 2022 (UTC)
- @Valentina.Anitnelav, Deryck Chan, Yair rand, Omegatron, Jura1:@Ls1g, Ghouston, RobLa: Not done Given that there was no consensus for creation within more than a year and no progress in sight I mark this as not done. If you still would like the property, please make a new proposal. ChristianKl ❪✉❫ 17:49, 4 December 2022 (UTC)
- So how should we indicate compliance with voting system criteria? 24.44.18.109 03:08, 11 December 2022 (UTC)