THE WEB CONSTRUCTION OF GENDER….Via Unfogged, I came across The Gender Genie, a web version of an algorithm that is alleged to have 80% accuracy in detecting whether a piece of text was written by a man or a woman. (The algorithm itself is described here.)
So I fed in some of my blog posts. It said I was mostly female.
Then I fed in a few of Jeanne d’Arc’s blog posts. She’s female too.
Next I tried Matt Yglesias: female. Steven Den Beste: female. Max Sawicky: mixed. Glenn Reynolds: mixed. Jane Galt: male.
I can think of two possible explanations. The first is that blogging, by its very nature, is personal and interactive, and thus fundamentally feminine in nature. This strikes me as horseshit unsupported by any evidence, however.
The other is that the algorithm is actually no better than flipping a coin. As you can see from the screen shot above, which shows that its predictions are successful 49.83% of the time, this seems like a much better explanation. But at least it has a sense of humor.
(By the way, this post that you’re reading right now is, according to the algorithm….strongly male! Go figure.)
POSTSCRIPT: For what it’s worth, I note that the algorithm is allegedly for use with fiction. I don’t know why, but perhaps it just doesn’t work well for nonfiction. I also note that it’s extremely sensitive to the use of the word “the” (strongly male) and the word “with” (strongly female). Keep that in mind if you feel the urge to change the tone of your writing someday.
UPDATE: In comments, one of the authors of the Gender Genie warns not to take the 49% figure too seriously:
Those stats, by the way, are not to be trusted. They’re entirely user-dependent, and users have been much more motivated to let it know when its wrong.
That sounds reasonable, although I’m curious if it’s just a guess or if they actually have some evidence for thinking that.
