Women are confusing, men are simple

[andrewducker]

Yesterday I posted a link to the page which tried to use your browser history to work out if you were male or female.

Today I found myself discussing the vagaries of research into female sexuality with marrog.

Strangely, these two things seem somehow linked together in my mind, along with the results of some of Simon Baron-Cohen's research into systemising/empathising functions in brains, and the male/female split therein. I exchanged an email or two with him after his research was covered in some newspapers, pointing out that while men did, statistically, seem to have a tendency to be focussed systemisers, lacking in empathy, women tended not towards empathy, but towards balance between the two functions. The papers were, of course, reporting it much more one-sidedly than that.

This linked into the poll yesterday, where it's obvious that the (very basic) algorithm can tell that a man is a man 2/3 of the time - but is no better at telling that a woman is a woman than a coin toss would be. Which would, again, tend to indicate that men are more likely lean over in one direction, making them easy to spot, while women are spread all over the place.

This tends to be picked up by reporters as "Men tend to be X, and women don't." and then reported as "Men are X, women are the opposite of X.", which is clearly nonsense.

(This then tends to be interpreted by a large chunk of people as "_all_ Men are X, _all_ Women are Y", which is beyond nonsense and into gibberish.)

Comments

[cangetmad.livejournal.com]

Well, no, only if you assume there's better-than-chance odds that your friends have similar tastes to you. They don't have to be the same sites, only sites in the same "gender category", and given the non-random nature of friendship, I'd say it has to skew the results in a way that provides at least as good an explanation for your results than assuming that the test always works that way.

[andrewducker]

Aaah, so it's skewing everyone on my friends list in a male direction? Which would mean that if it wasn't doing so then it'd be worse at predicting men, but better at predicting women?

[cangetmad.livejournal.com]

I don't think you can make that presumption from the data. Possibly only your female friends tend more male-typed, while your male friends are typical; possibly it skews everyone more male; possibly the test is a bit bollocks.

[andrewducker]

Well, it's clearly at least a _bit_ bollocks. It is, after all, just going from stats on what genders tend to use what sites, and thus open to all sorts of inconsistencies.

[terminalmalaise.livejournal.com]

Agree, I was coming here to point out the same possibility.

Also, it might have been interesting to have a better idea of how wrong the guesses were, since the results themselves gave an actual percentage likelihood. If, for example, the test were 100% certain on every correct guess, but only 51% certain on the wrong ones, the results suddenly might not look so bad.