Archives
Archive 1

It seems obvious to me that sex and gender (and possibly gender identity) should be distinct properties, if they are indeed different things. It'd be like combining a 'height or weight' property. It just... doesn't make much sense. Likewise, the related male/female/man/woman terms are also all horribly conflated, vague, and unclear. Which again seems like it ruins the entire point of the dataset. If "male" can refer to someone who is biologically female, how is that useful? Ultimately it seems the intended usage is for gender identity, and then assuming that unless otherwise stated, a person's sex is gender identity. IE an intentional conflation to avoid acknowledging that some people have a gender identity that differs from their sex. But... that's the entire definition of transgender, is it not? that their gender identity differs from their sex? Yet instead we see a solution being "transgender male" and "transgender female" which makes the distinction obvious, but the data more unwieldy. I also see that there's a distinction between human males and non-human males. This doesn't make much sense. It's not like the concept of maleness is somehow different in humans vs other animals.

It's obvious that all of this gender identity stuff should have it's own property and values, which should ideally be named distinctly and differently from sex in order to differentiate them. I think it's obvious that some people are intentionally trying to conflate them, which causes problems with unambiguous datasets like what wikidata is attempting to achieve.

Also noteworthy: Transsexualism should be distinct from Gender Identity Disorder. It appears that "transgender" typically refers to people with GID, however the definitions don't match this. Instead it seems that a fictitious concept of gender identity is coined and applied to everyone (despite it not being a thing) and then people using this as the justification for how GID and Gender Dysphoria result, when this is untrue. There is then further attempts to conflate Transgender/GID/Dysphoria with Transsexualism, and Gender Identity with Sex. And that is what has lead to the confusion and horribly unclear labeling. Each concept should have a distinct entry with clear and unambiguous definitions.

Kafke (talk) 11:14, 2 February 2021 (UTC)Reply

Yup. Ditto. I agree. Same question. I'm surprised this question is unanswered. Jerimee (talk) 22:04, 3 May 2024 (UTC)Reply

You are right and I think this question is going unanswered because a lot of us have been railing against P21 for such a long time with no results that we are just utterly exhausted and disheartened. Sex and gender are completely separate concepts and putting them together into one property has never made sense. Disentangling messy data on one of the most frequently used properties in Wikidata is a huge undertaking, especially when there is little to no institutional support behind the idea and a lot of political opposition. People have been working on it, though, and your support would be welcome! See https://www.wikidata.org/wiki/Wikidata:WikiProject_LGBT/gender --Crystal Yragui, University of Washington Libraries (talk) 20:24, 16 May 2024 (UTC)Reply

Sorry, have I correctly understood you, by "completely separate" you mean that they can differ in ~0.4% of cases [1] ? For me, 99.6% means "completely the same". Infovarius (talk) 22:07, 18 May 2024 (UTC)Reply

I cannot access the link you cite, but speaking as one of the allegedly 0.4% of transgender people, that our numbers are small does not make us insignificant. Funcrunch (talk) 16:52, 19 May 2024 (UTC)Reply

Hi all, I was adding someone to Wikidata, who is Akava'ine - which closely corresponds to the Samoan fa'afafine, a third gender in Pacific culture. Wikidata lets you add fa'afafine, but doesn't recognise Akava'ine as a constraint for gender. Please could it be added? Lajmmoore (talk) 15:06, 3 June 2021 (UTC)Reply

@Lajmmoore: Looks like akava'ine was added in November 2021, so you should be able to add akava'ine as a field to biography items now. --Prosperosity (talk) 23:26, 26 May 2022 (UTC)Reply

@Megs, Le Petit Chat: Since the property is highly used (7M+), I think it's better to discuss the possible addition of citation-needed constraint (Q54554025) before applying it.

  Notified participants of WikiProject property constraints and I have reported the issue in WD:PC. --Epìdosis 19:21, 8 June 2021 (UTC)Reply

I think that the real problem is for living people. It's better to create a Complex constraint instead of a citation-needed constraint (Q54554025). Fralambert (talk) 22:14, 8 June 2021 (UTC)Reply

See Property_talk:P21/Archive_1#Gender_is_central_for_forming_grammatical_sentences_and_as_such_needed.--GZWDer (talk) 04:40, 9 June 2021 (UTC)Reply

Yeah, this property is essential for many Wikipedia language versions. We should be very careful not to complicate addition of this information too much, because almost all infoboxes in many Wikipedias (and probably other non-Wikipedia users in these languages too) rely on this property to display proper gender forms of words based on the subject's sex/gender. Requiring references would really complicate stuff.Vojtěch Dostál (talk) 04:47, 11 June 2021 (UTC)Reply

Dear all, sometimes I add some new zoologic authors but it happens (eg. Indian ones) that the first name can't be used to define the correct gender. For many times I putted "Male" with "hypothetical value" (like Q111133708) but I will prefer say "unkowned". Have you an idea to solve my problem. Thanks in advance. Givet (talk) 17:25, 24 March 2022 (UTC)Reply

@Givet: just don't add any value! "Unknown value" means something different - that there are sources which have done some investigation and didn't find their gender. "Hypotheses" should be based upon some arguments, which do you have? --Infovarius (talk) 15:05, 25 March 2022 (UTC)Reply

@Infovarius: - First of all, thanks for your reply. I add this information because if we try (on a species article) to connect an author without any gender information, we receive an error message. So how can I do? And, sorry, but I have no idea if, for example M. N. Acharji (like Q111133708) is a man or a woman. Givet (talk) 16:35, 25 March 2022 (UTC)Reply

@Givet: Click edit, then an icon appears that looks like three boxes placed on top of one another. Click that icon and select "unknown value". See Help:Statements#Unknown or no values for more info. /Autom (talk) 10:19, 6 September 2022 (UTC)Reply

@Autom: Of course ! You are totally right, I will use this method for the next times. I don't know why I was not thinking of this possibility, really better than an input of "Male/Female" with "perhaps". Thank you so much :-) Have a great day. Givet (talk) 15:26, 6 September 2022 (UTC)Reply

@Autom: I tried a few seconds ago. First time I received an error message but when I retried another time, it works! (see Q113815418). Givet (talk) 16:30, 7 September 2022 (UTC)Reply

For example, both Q22983 and Q7823740 are listed as male despite both not being human.

I would recommend adding some qualifier (eg. "or human-like") to the description, but I'm not sure. Binarycat64 (talk) 15:20, 30 March 2022 (UTC)Reply

Currently sex or gender (P21) is called "sex or gender". Lots of people have proposed separating these. Anyone can discuss more at wd:WikiProject LGBT.

I want to share this 2018 discussion started by the government of the United Kingdom to manage gender or sex statements in government forms. It seems that wiki and the British government do the same thing of combining these concepts. They take comments in GitHub, and there I see arguments similar to the ones we have had on wiki.

• https://design-system.service.gov.uk/patterns/gender-or-sex/
• https://github.com/alphagov/govuk-design-system-backlog/issues/69

I do not see a solution here, but I see the UK government faced with the same problem. Bluerasberry (talk) 14:36, 9 June 2022 (UTC)Reply

• Thanks @Carlinmack: for sharing this. Bluerasberry (talk) 14:37, 9 June 2022 (UTC)Reply

The description of the property states that for humans, it is possible to use "male, female, non-binary, intersex, transgender female, transgender male, agender." However, intersex is a biological sex rather than a gender identity. Besides, "trangender" is not a gender identity but the condition that a person does not identify with the gender that was assigned at birth. Somehow, saying that a "transgender female" is different from a "female" can be interpreted as an exclusion of trans women in the category of women (same applies to transgender males). Sjoel (talk) 14:23, 19 August 2022 (UTC)Reply

• One possible solution for this exclusion issue would be to record two values, e.g. both female and transgender female, for those people who explicitly identify as transwomen. That allows searching for all women regardless of biological sex as well as trans women. AdamSeattle (talk) 22:01, 5 April 2023 (UTC)Reply

In the 'Wikidata property example' section here https://www.wikidata.org/wiki/Property:P1855 is listed as Female when at https://www.wikidata.org/wiki/Property:P1855 they are listed as Genderfluid. Suggest someone with edit rights make this light - but important! - edit. Drjwbaker (talk) 15:39, 5 December 2022 (UTC)Reply

Editors have been observing a pattern of sex or gender (P21) statements being added for living people in ways that cause harm by misgendering. This is a gender issue, which has overlap with the LGBT+ community and with other types of personal demographic information which may violate privacy.

Broad, ethically questionable additions of P21 statements using bots or Quickstatements is a problem mentioned in many of the discussions at Wikidata:WikiProject LGBT/gender. A recent group of large Quickstatements batches (example) demonstrated a need for a discussion surrounding best practices for assigning gender identities to living people on Wikidata. The editor responsible for these batches of edits collected Wikidata items for humans who did not have P21 values already, and who had been assigned given names listed in Wikidata as “female given name”. Given names are very frequently assigned to items for humans based on nothing more than their labels (see related discussion on user talk page for User:Jura1). The P21 values assigned were “female”, with a reference “based on heuristic: inferred from given name”. The edits were well-intentioned, with the stated purpose of increasing the representation of women on Wikidata.

While the intention of increasing womens’ representation on Wikidata is a good one, the rationale behind these edits is dangerously flawed. A person’s given name is not a reliable indicator of that person’s gender identity. Performing blanket edits based on labels inevitably misgenders people. When the editor was approached, they responded by halting further ingests (how many edits were planned is unknown), but were not immediately willing to revert the tens of thousands of edits already made, saying, “I haven't mass-reverted yet, as I'd like to find wider consensus.”[1] Conversation continued on the user’s Wikidata talk page. Several Wikidata editors expressed and explained their ethical concerns about the edits that were made. One editor explained that the batch had added incorrect statements about many Tibetan people, and those errors were corrected individually. Another editor expressed that they themselves had been misgendered on Wikidata in the past by similar edits. The editor decided to revert the batches based on this feedback.

A couple of subsequent conversations took place, with editors expressing pushback against reverting the edits in question. One user asked, “Instead of reversing heuristic genders, they have a very high rate of correctness, why not do the research to prove them right/wrong? If a given name is used by both genders, the formula does not assign a gender to that given name. You can also look for a primary/secondary source that mentions gender.” [2]. Once they discovered that the consensus appeared to lean towards reverting the changes, the same user stated, “We should probably look at error rates before making a decision, the error rate is most likely less than 1 in 1,000 based on the heuristic”. [3] This reasoning seems to be saying that since most affected people are unlikely to be harmed by misgendering, it’s ok for a minority of people to be harmed. It also seems to suggest that quality control and time-consuming research to verify the accuracy of huge batches of edits should not be the responsibility of the editor in charge of the batch, which seems irresponsible and unsustainable to me.

It is not guaranteed (and is arguably unlikely) that corrections to bad data from such batches will ever happen given the enormity and difficulty of the task of verifying gender identities of hundreds of thousands of people. Furthermore, harm from misgendering cannot be entirely eliminated by reverting bad edits. Wikidata is quickly ingested into the Semantic web (a strength that draws people to Wikidata, but also a weakness where bad data is concerned). Additionally, bots and other Quickstatements users make edits based on data as it exists at the time of the edit. These secondary edits aren’t reverted when the data they were based on changes. The longer an editor “waits for consensus” before reverting, the higher the risk of harm being multiplied in this way.

Myself and others have identified two proposals to limit swift, large-scale harm based on P21 moving forward:

First, I would like to propose that bots and Quickstatements activity affecting more than ten items at a time be disallowed for P21. For properties that may violate privacy, the “edit first, ask questions later” approach is inappropriate and has already done a lot of harm. Where privacy may be violated or people may be harmed, “ask questions first” is a better approach. Is this something the Wikidata community can get behind?

Second, I would like to propose that labels and given names not be accepted as references for P21 statements. I believe that this practice is unethical, harmful, and ought to be discouraged. Crystal Yragui, University of Washington Libraries (talk) 22:16, 29 March 2023 (UTC)Reply

•   Support I support both of these proposals, for the reasons given. We need to be more careful with statements that could do harm or violate privacy than the bots/batch edits allow, and these proposals would help with that.

Emwille (talk) 13:57, 30 March 2023 (UTC)Reply

•   Support I also support both of these proposals for the reasons Crystal stated. We need to be more mindful of the ramifications of including sex/gender in general in my opinion.

Sandrileine (talk) 18:51, 30 March 2023 (UTC)Reply

•   Support I support both of these proposals. Batch edits should not be used for properties that may violate privacy and/or cause harm. Also, in many languages and cultures, names are not strictly gendered, so attempting to use them to infer gender doesn't even make sense and is likely to lead to misgendering. Regarding P21 in general, I think it is problematic that it conflates sex and gender in the first place. - Ɛs Huvər (talk) 21:28, 30 March 2023 (UTC)Reply

  Support -- Egon Willighagen (talk) 09:34, 1 April 2023 (UTC)Reply

  Support I agree that this should be a policy not to batch add sex or gender (P21) based solely on a given name. AdamSeattle (talk) 21:58, 5 April 2023 (UTC)Reply

  Oppose "that bots and Quickstatements activity affecting more than ten items at a time be disallowed for P21" - this needs to be dealt with on a case-by-case basis. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:39, 5 September 2023 (UTC)Reply

  Support I support a policy not to batch add sex or gender (P21) based solely on a given name Jerimee (talk) 22:07, 3 May 2024 (UTC)Reply

  Support Proposal 2: I agree that adding gender statements based on name is logically and ethnically flawed. Claims of accuracy are doubtful -- has there been an analysis to determine how often each of these people's gender identity matches with the bot's gender assignments? Sonoet2 (talk) 14:04, 13 May 2024 (UTC)Reply

  Support Proposal 1 with some reservation: Agreed: We can slow down to be intentional. As we strive to highlight more women, let's please be respectful of identities and the complexity of gender because we may be missing the chance to highlight gender non-conforming folks as well. Batching isn't going to always be the best solution. I'd just suggest revisiting if enacted to see if there are unintentional consequences on the Quickstatements end. Thanks for putting forth these proposals. Sonoet2 (talk) 14:04, 13 May 2024 (UTC)Reply

Hi,

at which place are the suggestions of P21 managed? Since a short period demigender (Q63715923) and gender unknown (Q116741172) are in front of male and female.

It is illogical that the most common sex comes only at 4th place.--U. M. Owen (talk) 17:56, 21 August 2023 (UTC)Reply

@Infovarius: Since you've just added curated Wikidata property (Q121928698): May I ask where the curation list is located?--U. M. Owen (talk) 05:07, 9 September 2023 (UTC)Reply

I see the list at

⟨ P21 ⟩ property constraint (P2302)   ⟨ one-of constraint (Q21510859)      ⟩

. I don't know what makes an order as present. --Infovarius (talk) 19:45, 10 September 2023 (UTC)Reply

I've removed the Pokemon gender because it doesn't fit to the others. Let's see how this changes the suggestion list.--U. M. Owen (talk) 01:40, 16 September 2023 (UTC)Reply

Now, multiple Pokémon-related items have a problem with this property. Maybe remove demigender, since no item currently has it as a gender/sex. MikutoH (talk) 01:23, 17 September 2023 (UTC)Reply

It’s currently ordered by frequency for me. My display language is set to German right now, but I doubt that matters. Karl Oblique (talk) 12:40, 25 September 2023 (UTC)Reply

Frequency of use probably shouldn't be the only consideration, since we want people to see undisclosed gender (Q113124952) and agender (Q505371) near the top of the list before they go to set the special 'no value' or 'unknown value' toggle. Arlo Barnes (talk) 21:26, 12 October 2023 (UTC)Reply

The situation is still very unsatisfying. Some last added fictional porn genders like futanari appear now close to the top ahead of male and female. This must be changed. Male and female or vice versa must be on the top. Currently in English one hast to scroll down 9 lines to find female. Especially on small monitors and mobiles the scrolling is very time consuming if adding a lot of human beings. --Florentyna (talk) 12:22, 3 December 2023 (UTC)Reply

I've reported the problem at Wikidata:Report a technical problem#sex or gender (P21) and suggestion list. Korg (talk) 14:18, 3 December 2023 (UTC)Reply

As the person who added those fictional genders i am strongly against them being located at the near top. They should preferably be at the bottom Trade (talk) 12:22, 8 December 2023 (UTC)Reply

Seems improved now... Arlo Barnes (talk) 05:38, 9 December 2023 (UTC)Reply

Not really

https://i.ibb.co/zG0xd7V/Screenshot-20231209-141434.png Trade (talk) 13:16, 9 December 2023 (UTC)Reply

Weird, today it reverted for me also. But yesterday it was in the order shown in the one-of constraint... Arlo Barnes (talk) 20:20, 9 December 2023 (UTC)Reply

Still crazy:   --Florentyna (talk) 19:55, 5 January 2024 (UTC)Reply

Is there some magical creature behind the scene that controls the order of items suggested by properties? This is the only explanation i can think of since nobody on this site seems to have any control over it Trade (talk) 18:21, 17 January 2024 (UTC)Reply

Even though they are allowed as items in statements of this property, when putting them they are classified as "Unexpected value for gender" and "Could not save due to an error. (The save has failed. (Warning: The property sex or gender uses only some values, and some of those cannot be used in all items. The full documentation of the property is on Property talk:P21.))" (so that we should click in publish/enter again). Is there somewhere else where we can allow these items? Especially gendered intersex items. LIrala (talk) 20:42, 9 April 2024 (UTC)Reply