Jane, John, Leslie? A Historical Method for Algorithmic Gender Prediction | Hacker News Reader

Jane, John, Leslie? A Historical Method for Algorithmic Gender Prediction

digitalhumanities.org

26 pointsbenbreen40 comments

Surprising to see such a long article that doesn't touch on the modern issues surrounding gender.

You're much better off just letting the user tell you which gender they prefer to be identified by. This is such a personal issue, and so key to one's identity, that having the computer guess is just a bad idea.

edit: As icebraining pointed out, the article does address this point.

I'm very good friends with one of the authors (Cameron Blevins) and some context here is helpful: He is an expert at applying computational tools to historical records. For example, his thesis at Stanford is about using information about the expansion of the postal network to map the westward expansion in the US. For him, the interest here is mapping names to gender conditional on a specific era, not about doing so today when you have the option of asking the person.

Direct input is helpful, but if you're integrating with a user's existing data, they may not have noted it (I don't have a gender field in my phone's contacts, and I probably wouldn't fill it out anyway), but a good guess on gender may help sentences feel more natural. Eg: Brian sent you a message, call him back? You could put 'them' or 'Brian' again, or maybe just 'call back', but I have high confidence that you prefer male pronouns.

But what do you think you're getting in return for the risk of being wrong? What's the reward?

Market segmentation data at the risk of offending an imperceptibly small amount of people.

More natural sentence structure. Writing to avoid gendered pronouns can be awkward.

Not really. Just use they/them. It's perfectly ok in modern English

"Brian has updated their status" is really awkward.

Not really. It can get slightly more awkward when you use the gender-neutral pronoun repeatedly in one or two sentences - but only in the sense that it draws attention to the use of it.

One of my social networks (a popular network for a niche community, not queer-specific) uses a similar sentence - "X updated their about me section". Nobody's particularly fussed.

I don't think it's awkward at all, especially since I know some trans people who didn't change their names after coming out, but requested that people refer to them by different pronouns.

Having heard and said things like "Brian has updated her status" enough times, the "their" is hardly awkward at all :)

I don't find that awkward.

"Brian thanked Jamie for updating their status."

Exactly. I don't understand the hesitation for using they or their. We will never be able to remember all the different pronouns people want to be called (without some sort of system to look it up constantly at least), I believe it's perfectly acceptable.

But it's unclear whose status has been updated.

"Brian thanked Jamie for updating his status" is also unclear and needs context. Did Jamie update Brian's status? That's the first thing I parsed from it. If that sentence appeared as a machine-generated message on a social feed, I'd expect to see the actual status duplicated underneath.

Thing is... there's already plenty of people going primarily by singular "they" for a pronoun, and it generally works pretty well.

I agree with vertex, even with the male pronoun, it could still be confused( like a lot of things in English, you have to make an educated guess as to the meaning).

I hear that a lot from students.

"they/them" works wherever "he/him" does, and is gender-neutral.

I'd ask what's the risk? A chuckle with a friend that my phone thinks they are a different gender? I can't really imagine people getting upset with a mistake like this.

Has it been updated? I see a paragraph that does just that: http://www.digitalhumanities.org/dhq/vol/9/3/000223/000223.h...

Looks like I accidentally an entire paragraph. Yikes.

Well, I feel pretty stupid now.

Skipping things a lot these days?

Edit: apparently I am an idiot as well.

http://knowyourmeme.com/memes/i-accidentally

>having the computer guess is just a bad idea

Not necessarily. If you approach this strictly from a business perspective as a commercial software developer, statistical guesses (not just about gender) that, for example, delight 70% of your users at the price of upsetting 3% can be worth it.

Well, assuming that the amount of "delight" on success is comparable to the amount of annoyance on failure. I suspect that most users won't notice or care if the system guesses their gender correctly instead of using gender-neutral terms; but when you misgender a user, they will definitely notice, and they may be extremely annoyed, they may think it's funny and then tell all their friends about how stupid your service is, etc. It doesn't seem like a wise trade-off.

Relatedly, deliberately choosing to piss off a subset of users at no major benefit to yourself because you've decided there's not enough of them to affect your bottom line is kind of a dick move, even if it doesn't actually impact your financials directly. That's not a good habit to get into.

Although I think your post is also wrong on an ethical level, which would supercede any business arguments, I'll engage the one you've presented anyway.

Upsetting users by misgendering them won't just upset those users, it also has the potential to upset their friends and family. You're not just upsetting them, you're upsetting their network. That 3% starts to look a lot bigger.

And if we consider market trends, facebook actually expanded their gender identifying process, rather than simplifying it. Facebook has probably more data than any other entity for the purposes of guessing, and yet they chose to allow users to self-identify anyway. That would suggest it's not a business-correct decision.

There doesn't necessarily have to be any offence, and this doesn't necessarily have to be about pronouns. Suppose A/B testing on your site shows that female users are more likely to buy your product with a given display alternative over another, while males prefer the opposite. If you guess wrong, you would not be offending users by showing their least likely preferred alternative; you would just be reducing your chances of making a sale.

>I think your post is also wrong on an ethical level

Do you actually mean the argument in the post or the post itself for presenting it?

>You're not just upsetting them, you're upsetting their network.

Right, this would be a valid concern to an amoral profit maximizer. It would have to be balanced against the potential profits from bad publicity and, more deviously, the benefits of a good "redemption story" of changing the functionality at the upset users' request.

As for Facebook, it is an activist business with the power to shape social trends, so it is hard draw conclusions from its particular case.

I'm curious if Facebook cares about the long list gender box for advertising and demographic purposes. I cynically expect them to not trust what you fill in there.

Do people really care if you use "they" instead of (s)he? I suspect most don't even notice.

If you have the ability to use he or she, like with this algorithm it's a nice touch.

Or a constant annoyance.

http://www.robot-hugs.com/cognitive-load/

Oh but when I have trouble calling your company because they changed the gender of my record that's ok then.

Edit: I get passed around between call center people because they think I'm fraudulent because of software guessing this bullshit

I agree. However, some people change their given name to match the gender they identify with. In such cases, if the data about their gender preference were missing, this paper can also be useful.

I completely agree. Pronouns, just like names, are simply references. It's how you refer to someone.

You wouldn't assume you could look at someone and guess their name. Why assume you can guess their pronoun (from their appearance or name)?

NOTE: name+pronoun references != identity. Gender identity is a spectrum, and can't map 1-to-1 to low-bit pronouns.

Pronouns that are just like names are worthless, since they are a placeholder for a name. Also, if I could look at 99.7% of someones and guess their name I'd call it an amazing heuristic.

They're not just placeholders though, because as you've pointed out tautologically (and perhaps by accident) that would be worthless. They're also an abbreviation, a label, etc. If you're saying the value in pronouns is that you can guess someone's pronoun by looking at them, you're misguided. Alternatively, you can just use they/them/theirs all the time, and you'll always be right.

Interesting. It would be useful I think if somebody implemented a gender predictor for Chinese first names, which is difficult: http://chinese.stackexchange.com/questions/316/is-there-a-ru.... Also the set of Chinese first names is much larger than for any given Western country.

My favorite is "Anne" -- unambiguously female, but originally male, even so a couple of hundred years ago.

I would be far more interested in an explanation of why women are stealing all male names [1] in the first place.

Hypothesis 1: toxic patriarchy devalues everything associated with women, who are forced to resort to male names once all female names are considered worthless.

Hypothesis 2: rapacious women feel entitled to everything that is men's, so taking over their names fits an established pattern.

(Just to get the stupid ideas out of the way.)

[1: http://everything2.com/title/Women+are+gradually+stealing+al...]

Madison - the authors seem keenly unaware of Tom Hanks' mermaid fetish.

Reply on news.ycombinator.com