The Week Link
Wednesday 7 February 2001

The Week Think

What's It All About?

Welcome to The Week Link. Each week this page will feature one or two longer analytical pieces (The Week Think), an occasional cartoon (The Week Joke), and a selection of quality links (The Week's Links), all of which may or may not be related in subject matter. There's no firm topic: The Week Link will basically be about whatever I want to write about that week. Serious, funny, technical, philosophical: The Week Link will, with any luck, be all of these.

What it won't be is a personal journal-style weblog; I've been there, done that for the time being. Nor will it be updated daily; this is my attempt to cut down on the amount of weblog-posting I do for a while. Who knows how long it will last, but for now, enjoy The Week Link—updated every, uh, week.

The Test Has Spoken

Meg Pickard has protested at the result returned by The Spark's gender test, which predicted that she was a man with eighty percent certainty. Personally, I think that makes her a more interesting human being than most (not because she's like a man, but because she goes against the norm, and that usually leads to interesting results). But she's annoyed, and objects that 'it all depends whether we assign traits of masculinity or femininity to specific criteria'.

Well, no. It doesn't. We don't have to assign anything to anything. It's all down to number-crunching and the awesome and dispassionate power of statistics. They could have asked any questions they liked—do you like roast beef? do you like none?—and over time, as more and more people answer and then identify themselves as men or women, the good ol' computer only has to calculate the proportions of yes-no answers to each question for each gender and perform a few statistical tests to predict the male or femaleness of any test-taker to plus or minus a certain level of accuracy. The tricky part would be assigning a weight to each particular question in the final calculation, but I imagine that could be determined automatically over time as the predictive power of each question becomes better known.

The main problem with a test like this is that some gender attitudes (at the very least) are culturally determined, and so an international (internet) test is going to make mistakes for people outside the mainstream American culture of those taking the test. This is best exemplified by the 'Canada sucks' question, which I and probably many other non-Americans left blank, but which was almost certainly supposed to be answered 'yes' or 'yeah' by any rah-rah-USA boy.

Meg further adds that she doesn't 'think anyone is necessarily male, just because they are rational and prefer logical reasoning to whimsical frippery mumble mumble mumble mumble mumble...'

Of course not; you only have to see the spread of pink dots over the 'male' side of their final graph to confirm that. But how exactly does one identify the 'rational' or 'logical' questions in this test, anyway? For example, did you choose 1 or 6 as your favourite number? I chose 1; does that make me a binary-obsessed logical computer nerd boy, or a whimsical holistic 'we're all one' girly-girl? God knows (or rather, god-in-the-machine knows). Anyone who has gone back and tried the test a second time will have seen that you can change quite a few of your answers and still end up with exactly the same result. All of which suggests that there must be a few key questions that push one into the female side or the male side.

And if you can pinpoint which ones those are, you're a cannier human with 100 percent certainty than I am.

The Week's Links

The Weeks Past