highfemme: (Default)
[personal profile] highfemme
Generally I'm with the anti-quiz Nazis but I thought this was worth posting:

My journal says I'm 60% masculine.
What does your LJ writing style say about your gender?
LJ Gender Tool by [livejournal.com profile] hutta

Because that, of course, is pretty much my point. I may wear a lot of pink and look good in micro-skirts. But don't think that you know shit about my gender identity.

Update: Apologies if more community-minded LJers know this. However I ran that test on my small-but-very-select group of LJ friends, all of whom identify as female I believe, and all but one of them came out as male. The Koppel-Argamon test, which claims 80% accuracy, is less than 10% accurate in identifying the gender of my LJ peer group blogs. Whether this is a problem solely with the test, or particularly with its bookblog implementation I'm not sure. Either way it suggests that the algorithm can't cope with smart, often queer, women with attitude. What a surprise. Gender constructionism, anyone?

But my initial childish pleasure at subverting 'the rules' was diminished when I read the Koppel-Argamon paper properly. The algorithm was trained on some decent literature. There should be no gender bias in the type of material selected (though it is a little hard to tell from their description of their methodology, which doesn't properly examine this point). If we give Koppel and Argamon the benefit of the doubt on that, though, somewhere along the line we, with the honourable exception of [livejournal.com profile] tragicmulatta, all started writing in a way which has more in common with the style generally adopted by men. Is that really something to be proud of? I don't think [livejournal.com profile] shakinghell's Any Fran├žois would think so. Judiths Butler and Halberstam might be proud of us; but Audre Lorde probably wouldn't be.

Update: Yes, there are some men on my friends list now. Fortunately. My LJ was showing a disturbing separatist tendency.

Date: 2004-04-02 12:38 pm (UTC)
From: [identity profile] ericaceous.livejournal.com
It does sound like there may be a some material bias at work, maybe.

The paper you linked to above talks about how the test is not accurate for non-fiction when trained on fiction because the gender signifiers are different. Perhaps journal writing has yet other signifiers that make the test inaccurate.

The performance was far worse than random if you have correctly identified the gender composition of your friends list. Interestingly, the paper itself notes (see page 9) that "Examination of the six misclassified non-fiction documents reveals that they are all biographical, or diary-like works." Interesting!

One thing that struck me about the methodology is that the most recent writing included was from just over 10 years ago (1993). I think that most of the people on your reading list are on the young side, and possibly the writing style we have developed in this age of word processing fits the "masculine" signifiers more than the feminine ones. For example, certain prepositions are a feminine signifier. Maybe people write short, simpler sentences using fewer prepositions in the 21st century.

Now I have to think about the possible reasons diary-like and biographical works may have different language patterns, and what it may mean if my hypothesis about everyone's writing is becoming more masculine is correct.
(deleted comment)

Date: 2004-04-02 01:06 pm (UTC)
From: [identity profile] ericaceous.livejournal.com
I agree that the evolution of language thing might make Audre Lourde's point again (if it is even valid!).

I'm no linguist, but I really do believe that we are so heavily verbal in our mental processing at this point, that our language really can influence the way we process information and analyze trends.

I'm really glad that you wrote this entry that started me thinking!

Date: 2004-04-02 06:04 pm (UTC)
From: [identity profile] ericaceous.livejournal.com
Oh gosh, I revisited this and I have the need to make it clear that it is m,y hypothesis about the evolution of language, and not Audre Lourde's point that has the questionable validity. I am absolutely there with Lourde.

Date: 2004-04-02 01:04 pm (UTC)
From: [identity profile] highfemme.livejournal.com
You've really given this some thought! Thank you...

The remark about misclassification of non-fiction was referring to the 98% accurate test for distinguishing between the two types of writing. Indeed, diary-writing resembles fiction.

The 98% accuracy of the genre test reflects the fact that many features which show a difference between gender show a much more marked difference between genre.

The algorithm has been specially reconstituted for blogs, but I haven't found the details of how they did this. If they used a fixed weighting that would obviously be unreliable, since my friends' blogs are more analytical than average, and women's non-fiction is not so far from men's fiction in using those linguistic elements. In other words, if the algorithm was expecting fiction/memoir, or a fixed proportion of fiction, then it's not surprising that my friends come out as male. The presence of [livejournal.com profile] tragicmulatta as the sole algorithm-woman adds some weight to that hypothesis as she is also the most autobiographical. So if the blog implementation algorithm has a fixed weighting for genre, then it is the implementation which is at fault.

However some features differ more than others between genre, so any sensibly-designed blog algorithm would determine a genre-weighting rather than adopting a fixed one. It is at least possible that the implementation used here does do this; in which case the worse-than-random performance of the test in this case does reflect a divergence in non-fiction writing styles between my sample and that used in the study.

I was struck by your point about the evolution of language. Certainly true. Though I wonder if that just makes Audre Lorde's point again? We are writing our criticism in a way which our mothers would not recognise. We can fight about whether that is an inevitable aspect of claiming and acquiring power, or whether it is a sign that we have forgotten what we should have been fighting for. That is the argument at the end of the beginning of every revolution. But here it is (apparently), in the language we use to post our anti-patriarchal tales.

Date: 2004-04-02 03:13 pm (UTC)
From: [identity profile] tragicmulatta.livejournal.com
Oh my. [livejournal.com profile] highfemme, this is fascinating! Thanks to you for sharing about the algorithm, and to you and [livejournal.com profile] ericaceous for further thoughts on it. I'm feeling really lucky to have such thinkers on my friends list, and will commit to letting this one marinate in the old brain juices for a bit. More to come.

Date: 2004-04-06 03:51 pm (UTC)
From: [identity profile] highfemme.livejournal.com
Looking forward to the further thoughts... But in the meantime, we're thinking of you, girl.

wow, surprising that we scored almost the same

Date: 2004-04-03 02:11 am (UTC)
From: [identity profile] slantedtruth.livejournal.com
My journal says I'm 59% masculine.
What does your LJ writing style say about your gender?
LJ Gender Tool by [livejournal.com profile] hutta

Date: 2004-04-06 03:38 pm (UTC)
From: [identity profile] beelavender.livejournal.com
The primary criticism of my writing is that my perspective is too masculine - I was surprised that my score on this quiz was only 62%.

Date: 2004-04-06 03:45 pm (UTC)
From: [identity profile] highfemme.livejournal.com
Hiya! How ya doing?
Too masculine? I worry about essentialist definitions of femininity. Especially if they're used to question the value of your experience of motherhood.

Date: 2004-04-06 06:58 pm (UTC)
From: [identity profile] leaving-home.livejournal.com
My email is broken and my wrist is sprained, but other than that I'm well! We had a lovely holiday.

I agree about the issue of definitions. Unfortunately, I take a perverse pleasure in making people wonder about .... everything. When I decided to have a child at age eighteen my faculty advisor told me that my choice was anti-feminist and that I should drop out of school. I took that as a major challenge not only to persevere in my education but also to work to change the game for anyone following me. Nobody defines my feminism except me - with due respect to the women who raised me.

In terms of the quiz, and the ranking of masculine/feminine in writing, I really have to wonder what it comes down to on an essential level - adjective choice? Length of sentences? I'm an editor and evaluate the work of others all the time, and there is definitely a pattern to what I read. But the good stuff, the extraordinary - that never fits the schematic. It might be conveying the same message but the word choice is markedly different.

Date: 2004-04-07 04:27 am (UTC)
From: [identity profile] highfemme.livejournal.com
Indeed. The discussion with ericaceous covers the main points on this but the algorithm seems not to be based on sentence length but primarily on pronouns and conjunctions reflecting agency and logical connection. Such words are hard to substitute for. The correlation between words and gender in this study could, on my superficial reading, be seen as reflecting greater abstraction and greater attribution of direct as opposed to associative causation by men.

This is why the question of whether the algorithm assumes or tests for the proportion of fiction/narrative vs. analysis is crucial. If it uses a fixed ratio, this has shown nothing except that my friends write a lot of analysis. If it uses an adaptive ratio, then my friends are more prone to abstraction and direct causal attribution than was typical among published women writers of analytical texts as measured ten years ago. Which is interesting.

September 2005


Style Credit

Expand Cut Tags

No cut tags
Page generated Sep. 20th, 2017 09:37 pm
Powered by Dreamwidth Studios