Re: Outed

1

I'm not going to read the paper, because it seems too much like work, but:

Given a single facial image, a classifier could correctly distinguish between gay and heterosexual men in 81% of cases, and in 74% of cases for women.

doesn't seem very hard to do. Below, is complete code for my AI bot that I suspect can achieve about 90% accuracy. Programmed in SAS for ease of use.

data aibot;
input prediction $7.;
datalines;
Not gay
;run;


Posted by: Moby Hick | Link to this comment | 09- 8-17 6:20 AM
horizontal rule
2

It's easier to be better than people because people are shit at understanding base rates.


Posted by: Moby Hick | Link to this comment | 09- 8-17 6:22 AM
horizontal rule
3
Prediction models aimed at gender alone allowed for detecting gay males with 57% accuracy and gay females with 58% accuracy.

I don't understand this line. Can anyone explain it better?

Consistent with the prenatal hormone theory of sexual orientation, gay men and women tended to have gender-atypical facial morphology, expression, and grooming styles.

The last thing on that list doesn't seem to belong. I mean, I don't know the details of the prenatal hormone theory of sexual orientation, but assuming a plain English meaning, it doesn't seem to have much to do with grooming styles.


Posted by: Cyrus | Link to this comment | 09- 8-17 6:22 AM
horizontal rule
4

Just try drinking a cup of estrogen and wearing an Ed Hardy t-shirt and you'll see.


Posted by: Moby Hick | Link to this comment | 09- 8-17 6:30 AM
horizontal rule
5

Moby, I love you completely in whatever way is appropriatrly non-creepy I hope and 1 is perfection.

I don't know how masculine my face is without grooming markers but my finger length conforms to the prenatal hormone theory. And I'm not perfectly gay, so to ogged's too I guess.


Posted by: Thorn | Link to this comment | 09- 8-17 6:35 AM
horizontal rule
6

Do heterosexual men worry about creepy? Let's ask a computer.


Posted by: Moby Hick | Link to this comment | 09- 8-17 6:37 AM
horizontal rule
7

IANAComputer but can assure you not enough of them and not enough.


Posted by: Thorn | Link to this comment | 09- 8-17 6:38 AM
horizontal rule
8

I'm not perfectly gay

Don't be too hard on yourself. We know you're being the best gay you can be.


Posted by: ajay | Link to this comment | 09- 8-17 6:52 AM
horizontal rule
9

I wonder about the finger-measuring thing. As near as I can tell, my index and ring fingers are the same length, but that depends on how you pick the bottom of the finger.


Posted by: Moby Hick | Link to this comment | 09- 8-17 6:59 AM
horizontal rule
10

How does the finger-measuring thing work?


Posted by: Mossy Character | Link to this comment | 09- 8-17 7:06 AM
horizontal rule
11

The link doesn't seem to provide their gaydar test, so I found one here.

(It's a pretty long test, and doesn't give you any indication of how far along you are as you take it, which makes it seem even longer, I think.)

I scored 56.7%, which was a little worse than the 58.5% average. Here's how they say that breaks down by orientation:

Straight 58.0%
Bisexual 58.6%
Gay/Lesbian 59.7%
Asexual 55.6%


Posted by: politicalfootball | Link to this comment | 09- 8-17 7:07 AM
horizontal rule
12

You measure your fingers and if the ring finger is shorter than the index finger, you're supposed to become fond of your mother.


Posted by: Moby Hick | Link to this comment | 09- 8-17 7:08 AM
horizontal rule
13

12 to 10.


Posted by: Moby Hick | Link to this comment | 09- 8-17 7:09 AM
horizontal rule
14

Huh. My ring fingers are substantially longer, such that there's no need to measure (the first knuckles are pretty much in a straight line, so that's not an issue). I didn't know there was so much variation in hand shape.


Posted by: Mossy Character | Link to this comment | 09- 8-17 7:15 AM
horizontal rule
15

Congrats, Mossy. You're a straight boy! Though so am I by that standard.


Posted by: Thorn | Link to this comment | 09- 8-17 7:19 AM
horizontal rule
16

The boringly obvious response to 1--probably too boring, but alas that is my way--is that the set they were testing it on has equal numbers of straight and gay men, and straight and gay women. However, for larger numbers of pictures per subject it's skewed slightly towards straight men and lesbians.

Consistent with the PHT, gay faces tended to be gender atypical. Average landmark locations revealed that gay men had narrower jaws and longer noses, while lesbians had larger jaws. Composite faces suggest that gay men had larger foreheads than heterosexual men, while lesbians had smaller foreheads than heterosexual women.

I can't say my gaydar is very good, but I would never have guessed nose length is one of the factors visually distinguishing gay men.


Posted by: dalriata | Link to this comment | 09- 8-17 7:24 AM
horizontal rule
17

I would never have guessed nose length is one of the factors visually distinguishing gay men.

IT'S SUBTEXT.


Posted by: Cyrano de Bergerac | Link to this comment | 09- 8-17 7:37 AM
horizontal rule
18

I got a 66.7 on the test in 11, which does and doesn't surprise me. (I think I have very good gaydar IRL, but the images test was hard.)

I have super a femme face and build, and I think it's why even when i get a pixie cut and even though I prefer to dress fairly androgynously, people still see me as a pretty femmy woman. My index finger is also noticeably longer than my ring finger, which makes me a straight woman.


Posted by: Buttercup | Link to this comment | 09- 8-17 8:40 AM
horizontal rule
19

The longer nose thing for gay men doesn't surprise me, when I think about it now. And thinking about the pictures, stubbier noses read as more hetero to me.


Posted by: Buttercup | Link to this comment | 09- 8-17 8:47 AM
horizontal rule
20

Lorenzo Medici, Mossadegh, and Suleiman all revealed by the ways of science.


Posted by: lw | Link to this comment | 09- 8-17 8:56 AM
horizontal rule
21

I, like pf, also got 56.7%. It's a hard test.


Posted by: dalriata | Link to this comment | 09- 8-17 8:58 AM
horizontal rule
22

65% on the test. Ring and index fingers on my right hand are the same length. (So far as I can tell - I'm with Moby in not being sure what counts as the base of the finger.) On the left, the ring finger is about 0.5 cm longer. Cue Whitman quote...


Posted by: One of Many | Link to this comment | 09- 8-17 9:30 AM
horizontal rule
23

Just visually, looking at my hand the ring finger looks much longer, but the base of the index finger (looking on the palm side) is lower.


Posted by: Moby Hick | Link to this comment | 09- 8-17 9:35 AM
horizontal rule
24

60%, but I got bored and started clicking at random.

I wonder whether people would do better if they watched a short video clip rather than just looking at a photo.


Posted by: ajay | Link to this comment | 09- 8-17 9:46 AM
horizontal rule
25

gif-dar?


Posted by: Moby Hick | Link to this comment | 09- 8-17 9:48 AM
horizontal rule
26

I find the facial structure claims really super hard to believe.


Posted by: LizardBreath | Link to this comment | 09- 8-17 9:53 AM
horizontal rule
27

A gaydar png.


Posted by: Ginger Yellow | Link to this comment | 09- 8-17 9:54 AM
horizontal rule
28

That test is too goddamned long. This is the internet, ffs, and allofasudden I'm supposed to have an attention span.

I scored dead-on average, although I ran out of patience about 2/3 way through and just started clicking every picture on the right. But if that's a representative score for me, maybe that explains something about my dating record... Although seriously, it's a bit silly to understand this in terms of IRL gaydar. Nobody in practices makes the call based on what a picture of somebody's face looks like.


Posted by: Swope FM | Link to this comment | 09- 8-17 10:02 AM
horizontal rule
29

Where's the link to the test? I just see the link to the paper.


Posted by: Barry Freed | Link to this comment | 09- 8-17 10:02 AM
horizontal rule
30

The test is the paper.


Posted by: Opinionated Litmus Test | Link to this comment | 09- 8-17 10:07 AM
horizontal rule
31

I did worse than average, which doesn't surprise me. Anyone who's not displaying explicit signifiers, I make guesses but they're usually wrong.


Posted by: LizardBreath | Link to this comment | 09- 8-17 10:31 AM
horizontal rule
32

I did worse than average, which doesn't surprise me. Anyone who's not displaying explicit signifiers, I make guesses but they're usually wrong.

I was almost exactly average (Your score was 58.3%. / The average score is 58.5%) which surprised me because I was just randomly guessing for a bunch of them.


Posted by: NickS | Link to this comment | 09- 8-17 10:38 AM
horizontal rule
33

It's easier to be better than people because people are shit at understanding base rates.

Moby gets an A


Posted by: Sifu Tweety | Link to this comment | 09- 8-17 10:50 AM
horizontal rule
34

I wonder whether people would do better if they watched a short video clip rather than just looking at a photo


Posted by: Sifu Tweety | Link to this comment | 09- 8-17 10:51 AM
horizontal rule
35

26: me too because my forehead is huuuuuge.


Posted by: | Link to this comment | 09- 8-17 10:53 AM
horizontal rule
36

On the other hand, I don't even SEE "remember personal info."


Posted by: Thorn | Link to this comment | 09- 8-17 10:54 AM
horizontal rule
37

although I ran out of patience about 2/3 way through and just started clicking every picture on the right.

Twinsies!


Posted by: heebie-geebie | Link to this comment | 09- 8-17 11:05 AM
horizontal rule
38

I have trouble dealing with anything that involves looking at lots of faces. This probably explains my dating record.


Posted by: fake accent | Link to this comment | 09- 8-17 11:48 AM
horizontal rule
39

There's always butts.


Posted by: Moby Hick | Link to this comment | 09- 8-17 12:02 PM
horizontal rule
40

Until the rise of the carnivorous yoga pants.


Posted by: Moby Hick | Link to this comment | 09- 8-17 12:10 PM
horizontal rule
41

39

You can tell if a person is gay by his/her butt?


Posted by: Buttercup | Link to this comment | 09- 8-17 12:16 PM
horizontal rule
42

You can't?


Posted by: NickS | Link to this comment | 09- 8-17 12:21 PM
horizontal rule
43

The general question is tough to answer, but generally you can tell after a while if they're interested right now.


Posted by: lw | Link to this comment | 09- 8-17 12:21 PM
horizontal rule
44

When their butt cozies right up to you.


Posted by: heebie-geebie | Link to this comment | 09- 8-17 12:22 PM
horizontal rule
45

You can tell if you are gay by whose butts you stare at.


Posted by: Moby Hick | Link to this comment | 09- 8-17 12:22 PM
horizontal rule
46

I stare at my own butt.


Posted by: heebie-geebie | Link to this comment | 09- 8-17 12:33 PM
horizontal rule
47

Who stares at the buttstarers?


Posted by: dalriata | Link to this comment | 09- 8-17 12:44 PM
horizontal rule
48

46

Autosexual!!


Posted by: Buttercup | Link to this comment | 09- 8-17 1:12 PM
horizontal rule
49

Or would you prefer egosexual?


Posted by: Buttercup | Link to this comment | 09- 8-17 1:12 PM
horizontal rule
50

Autoregressive.


Posted by: Moby Hick | Link to this comment | 09- 8-17 1:43 PM
horizontal rule
51

heebie gives a new meaning to "heteroflexible."


Posted by: teofilo | Link to this comment | 09- 8-17 3:01 PM
horizontal rule
52

heebie gives a new meaning to "heteroflexible."

What's really difficult is staring at the butts of all those who don't stare at their own butts.


Posted by: One of Many | Link to this comment | 09- 8-17 5:11 PM
horizontal rule
53

And here's a fight between the authors of the thing in the OP and HRC and Glaad.


Posted by: chris y | Link to this comment | 09- 9-17 9:48 AM
horizontal rule
54

I got 66.7% on that test. Guessing the men seemed easier than guessing the women.


Posted by: Frostbite | Link to this comment | 09-10-17 11:42 AM
horizontal rule
55

Anyone who's not displaying explicit signifiers, I make guesses but they're usually wrong.

If someone of the same sex as you doesn't want to display their explicit signifiers to you, they're probably not gay.


Posted by: Todd | Link to this comment | 09-10-17 1:05 PM
horizontal rule
56

To the OP, this is interesting: https://theoutline.com/post/2228/that-study-on-artificially-intelligent-gaydar-is-now-under-ethical-review-michal-kosinski

I still can't find the link to the test though, is it on Standpipe's blog?


Posted by: Barry Freed | Link to this comment | 09-13-17 2:58 AM
horizontal rule
57

I don't think it was ever linked in the original article.


Posted by: heebie-geebie | Link to this comment | 09-13-17 7:36 AM
horizontal rule