It’s true, “GenderAnalyzer” told me so.
I even took a pic of the form results so there is proof. Next time anyone accuses Shameless of being say… a harpies nest of left-wing feminist reactionaries, we can easily fire back with “yeah but we write like men, so whatever, talk to the pen!”
Apparently I also write like a man, or pick manly topics, or use manly sentence structures, or have a masculine vocabulary? Who really knows (sigh). From a quick scroll of the last few entry topics on Shameless we’ve written posts on:
- The US election(4)
- Fairies vs. Princess (1) ahem - ever so manly
- Sci Fi and Occult TV shows
Oddly both Feministing, and Blogher, were correctly attributed to the ladiezzz. So the theory that this weeks emphasis on politics may have gotten us onto the other team fails, as a quick glance reveals that those two US-based blogs are even more heavily weighted with political content.
(And yes of course it’s silly that ‘political’ writing is considered more masculine, I don’t make the stereotypes I just write about them.)
Now before we get all upset, it appears the algorithm that is sniffing our gender via writing isn’t actually doing all that great, with a success rate of only 56%. The makers of the site claim that they set up a filter for the analyzer using uclassify a free web-based filtering service. What that means is they picked approximately 2000 blogs and gave each blog an attribute male or female based on authorship. The text-filter that came out of the analysis of 2000 blogs is now applied to any blog entered onto the GenderAnalyzer site, and it spits out a result based on that filter.
The filter does not change the more people add blogs to the analysis, which partly explains why it is so darn wrong. If there are only 2000 blogs out of however many millions that exist being used to develop a measure for gender, there is just no way that a reasonable rate of variance (error) can be achieved. The sample size is too small and it is not being adjusted with new data each time it produces an error.
See how fun metrics can be :)
What else have we learned? (and no, this does not involve math). I’ve have learned (and not for the first time) that online girl/boy tests are really only useful for pointing out how inadequate gender is as an analytical framework.
Whenever I “fail” at online gender tests, I feel a certain satisfaction. And let me tell you, I fail at pretty much every, “tell if you are a boy or girl” quiz the internet has to offer. When I began writing/hanging out on the web I didn’t think that my inherent tomboyishness/ queerness would translate well onto the webs. I figured that since technology is composed of binaries itself (ha ha geek humour) most online analysis of gender would not read the complexities of my everyday self. This is true in one respect, since I clearly fail at being a girl according to the internet. But there’s another way to read unintentional online transgenderism.
When I fail at a gender test online, it means I’ve gamed the system. I’ve proven that a programmer can’t take my blog, compare it to two thousand others and fit me tidily into my gender. When things like Genderanalyzer are correct a mere 56% of the time they accidentally demonstrate that gender is not a binary, and online tests that rely on the opposition of boys vs girls, are kinda bound to fail.
Thank goodness the genderanalyzer didn’t identify Shameless as “woman’s writing”, we must be doing something right !