Wikipedia:Reference desk/Archives/Science/2022 February 15

From Wikipedia, the free encyclopedia
Science desk
< February 14 << Jan | February | Mar >> February 16 >
Welcome to the Wikipedia Science Reference Desk Archives
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.


February 15[edit]

Artificial Intelligence, any ways to guess gender from names?[edit]

Greetings,

@ WT:WOMRED#Growth Team Features; we need mentors! I came across one interesting point being discussed by User:Ipigott.

Personally when I visit multiple language Wikipedia where in I do not know the respective language and culture at all, even without taking into account the articles they are editing, I can guess at least few female user accounts, request some help on women's rights related article and do get such help. Even I can make a guess to some extent what likely percentage of women are editing on the wiki project looking at recent changes. Of course I am too rational to claim any supernatural abilities in me, How I have ability to guess gender may be this question pertains to multiple disciplines like cultural, linguistic, anthropology, psychology etc.

But giving such task of guessing gender from user names for software would sound next to impossible. Still what possibilities would exist for AI researcher?

Are there any scientific studies on this topic in above discussed various scientific fields?

Bookku, 'Encyclopedias = expanding information & knowledge' (talk) 12:29, 15 February 2022 (UTC)[reply]

There are at least two separate questions here. When editing Wikipedia and independent of their User Name, editors can set Special:GlobalPreferences#mw-prefsection-personal to the gender which they prefer to be addressed, or deliberately leave this unspecified, which is the default for new accounts. Then any other user can use the correct terms when referring to them, if that has been specified. The Navigation popup gadget (active it on your own page here) will show the gender and other information when you hover over the name of another editor in talk pages and elsewhere. As to the more general question, then "guess" is the best that can be done, although many names, if not Usernames, do indeed mostly get applied to one preferred gender. Mike Turnbull (talk) 13:10, 15 February 2022 (UTC)[reply]
English, at least, has quite a few unisex names, and it would be interesting to see how, or if, AI could somehow figure out the gender based on nothing but the name. --←Baseball Bugs What's up, Doc? carrots→ 14:50, 15 February 2022 (UTC)[reply]
Reminds me of the Turing test and Douglas R. Hofstadter's wonderful piece from Scientific American in 1981 here, where male?/female? Chris, Pat and Sandy are discussing what Turing called the 'imitation game'. --T*U (talk) 18:31, 15 February 2022 (UTC)[reply]
Yes, there is lots of such software.https://duckduckgo.com/?q=name+gender+software] 2602:24A:DE47:B8E0:1B43:29FD:A863:33CA (talk) 22:27, 15 February 2022 (UTC)[reply]
And they are all based on guesswork. If you have a name such as "Chris Smith", and no other information, it is impossible to know what the gender is. --←Baseball Bugs What's up, Doc? carrots→ 15:15, 16 February 2022 (UTC)[reply]
Well put Baseball Bugs. Many years ago I knew a family where Chris and Chris were married, and named their first child Chris. Now sort that one out! Martin of Sheffield (talk) 15:24, 16 February 2022 (UTC)[reply]
Evelyn Waugh was married to someone named... Evelyn Gardner. They didn't have children however. --Jayron32 15:58, 16 February 2022 (UTC)[reply]
Even with a given name such as "Andrea" or "Anne" you can't be sure: Andrea Denver, Anne Mulder.  --Lambiam 15:49, 16 February 2022 (UTC)[reply]
Michael Learned would also give such an AI trouble as well... --Jayron32 15:59, 16 February 2022 (UTC)[reply]
And Christopher Norris. --←Baseball Bugs What's up, Doc? carrots→ 16:20, 16 February 2022 (UTC)[reply]
On top of all that, now we have the minefield of not knowing whether a random person prefers to be associated with a gender that does not match their outward appearance, or no particular gender, or all genders, or something that changes sometimes; and having to choose one's pronouns accordingly. By the way, where could I get hold of a litre of that "gender fluid"? -- Jack of Oz [pleasantries] 20:53, 16 February 2022 (UTC) [reply]
Yes, by switching language names can change gender, but maybe that's cheating. In English Andrea is female, in Italian it's male. Andrea Denver is an Italian living in the US. In Dutch Anne is a female name, but in Frisian it's male, the female equivalent being Antje. Have a look at the map on [1] (sorry, Dutch only). Once you know the language, the number of unisex names rapidly drops. And in languages with marking of grammatical gender on nouns, it would nearly always be possible to correctly guess someone's gender from the name. PiusImpavidus (talk) 09:47, 17 February 2022 (UTC)[reply]
All AI approaches require training data; a set of inputs with corresponding correct answers. Given that those data are quite simply not available over the set of Wikipedia editors, the best that you could expect would be for the AI to repeat the skew attributable to the initial assumptions. 37.166.208.180 (talk) 19:56, 17 February 2022 (UTC)[reply]

I think some respondents here have missed the point. The question was about guessing gender, and I'd expect guessing in this context to mean choosing an answer with a reasonably high probability of being right, not determining a right answer with certainty. There is never certainty anyway (people in this modern era can switch gender at any time). Names could be misleading even in earlier times, e.g. Waltons' actress Michael Learned. I do have the impression that those software models do reasonably well at guessing genders from names, but I would never claim them to be perfect. 2602:24A:DE47:B8E0:1B43:29FD:A863:33CA (talk) 22:08, 17 February 2022 (UTC)[reply]

One more: Shirley Crabtree. Alansplodge (talk) 11:34, 22 February 2022 (UTC)[reply]
How about Christian Ramsay? —Kusma (talk) 12:07, 22 February 2022 (UTC)[reply]
And Shirley Povich. --←Baseball Bugs What's up, Doc? carrots→ 16:57, 22 February 2022 (UTC)[reply]