LanguageHat reports that a computer program has been developed that can tell with 80% accuracy whether a given text is written by a man or a woman. LanguageHat (is that a male or female name, and could the program tell) is impressed by the results, though s/he is rightly very suspicious of the stereotypes the programmers used in building the machine.

If the results are good, shouldn’t that be enough? Maybe not. I’m always reminded in these cases of Daniel Hausman’s little refutation of Friedman’s instrumentalism: Why Look Under the Hood?, just about my all-time favourite philosophy paper. Here’s a bad way to judge the quality of a used car: drive it around the block a few times and see if anything goes wrong. That’s not a useless test. After all, you’re testing whether the car does what you eventually want it to do. But we know in practice it’s much better to look under the hood, and see how it’s doing what it does. (If you disagree, contact Hausman – he’s got some great cheap cars to sell you.) Hausman argues that the lesson generalises. Some theories do well for a while by luck, or because they have only been tested in areas for which they were specifically designed. Looking forward, and outward, it’s more important to know how they get it right than that they get it right.