Gender analysis

[lnr]

Today on LJ I discovered this tool for identifying what gender you appear to be, based on your most recent livejournal posts:

My journal says I'm 55% masculine.

What does your LJ writing style say about your gender? UsernameCommunity

LJ Gender Tool by hutta

I tried it on myself and it thought I was just about male, so I started trying it on my friends, and it thought j4 was male too. So I started trying it on all the women I could think of, and nearly all of them were male too. So I reckon I'll have a go with all the women on my friendslist, plus a couple of others I know:

Male
aendr
ailbhe
aldabra
ali_anarres
angua
antinomy
atreic
beckyc
bellinghwoman
bopeepsheep
ceb
claroscuro
dmwcarol
dreamingchristi
feanelwa
florence90
geekette8
ghoti
gnimmel
j4
julietk
jvvw
k425
karen2205
keris
lark_ascending
lnr
marnameow
mst3kgirl
nassus
nilasae
redcountess
rillaith
rmc28
soubrette
sphyg
tamsinj
uitlander
vyvyan

Female
_lois_
ejde
hollyp
lanfykins
lostprophet
mofette
naath
q_skud_
taimatsu
teleute

Of course this is just based on public posts. Wonder if I can be bothered to do the same with the other half of my acquaintances and see what the results are there.

Comments

[arkady.livejournal.com]

You missed me from your list, but it said I'm probably male too. However that's based on only one public post, so hardly accurate....

[lnr]

It didn't give you any percentages, so I kind of figured with so little data it was better to leave it out. There was someone else similar, but I forget who.

[bopeepsheep.livejournal.com]

smallclanger, however, is 55% feminine. Given the authorship of both journals I feel there's probably some flaw...

[glitterboy1.livejournal.com]

My memory from when I first saw it a few months ago is that it greatly over-reports 'male' (as LNR has just shown!). I fed it a load of my locked posts, as well as public, and I think even I came out as male on a lot of samples. :-) It seems to use some odd things as indicators.


Given the authorship of both journals...

What, do you mean that SC writes *your* journal, too? ;-)

[vyvyan.livejournal.com]

Mm, all the male friends I've tried it with come out male too! Maybe there's an average difference in percentages, though e.g. men might tend to get higher "masculinity" scores than women, even though both groups may tend to get above 50% masculinity (and hence be classed as "probably male") by this tool. I have no idea whether the features of writing they're identifying as "masculine" or "feminine" have any basis in research, though - it might be entirely speculative.

[lnr]

It's true that many of these people had a score of only 50-55% masculine, but quite a few were well up in the 70s and 80s too. And I was mildly surprised that Skud, who is one of the women I'm least likely to think of as stereotypically feminine, was in the female group.

[yonmei.livejournal.com]

If you check out my poll (http://www.livejournal.com/users/yonmei/144693.html) it does seem to get women wrong more often than right. Much more...

[pseudomonas.livejournal.com]

It might be the case that people who are reported wrongly are much more inclined to leave feedback.

[yonmei.livejournal.com]

True, but even so, women appear to be reported wrongly much more often than men.

[lnr]

I'd spotted the poll and voted in it, but hadn't had time to look at the results yet.

I find myself wondering how much of it is due to the fact that many of the women I know here are geeks, or hang around with geeks, or work in geek jobs.

Or perhaps something in the general style of writing livejournal posts is different from other blogs, and hence livejournal postings are rated incorrectly as a result. I do think the way LJ is typically read in the context of lots of other journals (ie via friends pages) rather than as a standalone blog could affect the way people word things quite strongly.

And then there's the question of which posts are public. Perhaps if we could see the results with locked posts we might find them more feminine? Though in my case it would make no difference.

[vyvyan.livejournal.com]

The results of the poll suggest that the tool is more likely to define a piece of writing as by a man than by a woman (over 60% of respondents were counted as men), and since about 80% of the respondents were actually female, this gave it a worse than 50% success rate. It would do better if it guessed completely randomly! But to be honest, I'm really not sure that its criteria are meaningful anyway: I looked through the list of "masculine" and "feminine" cues it looks for, and they bear no relation to sociolinguistic literature I've read on differences between men's and women's language use.

[pseudomonas.livejournal.com]

It looks like it ought to be fairly easy to make the thing learn from its mistakes as it goes, taking feedback on the output page. It might be fun to have a go at it if I can find some CGI space and some round tuits.

[imc.livejournal.com]

I did my friends (of which there are rather fewer than yours <g>) and myself, and arranged them in order of percentage maleness in this table:

(Green users are male infants whose journals are (at least predominantly) authored by their mothers.)

There's a very slight concentration of females at the top of the table, but apart from that the spread seems very even. Obviously this detector produces far too many `males', but even if you took (say) 10% off everyone's rating you'd end up with just as many wrong answers as right ones.