In our house you could nail it a lot less scientifically. Just scan the domain names and add 1 point for male with a /porn/ match and 1 female point for /baby/. I dare say the Google seaerches would be an immediate giveaway too!
Yeah, not only are they probably using two-year-old data, but if the confidence given is 97% for it being the wrong gender for me, I doubt that it's actually an effective tool.<p>There are all kinds of fuzzy behavioral ways to analyze gender. I remember when the Gender Genie (<a href="http://bookblog.net/gender/genie.php" rel="nofollow">http://bookblog.net/gender/genie.php</a>) made a similar splash. The issue is that it ignores the finer-grained sociological contexts in how certain people behave.<p>I can't find their name, but there is an African tribe where, as a courtship ritual, men dress up and wear makeup to attract women. To other people, this would sound like gender roles are reversed, but in reality, it's just their cultural definition of gender roles at play.<p>Mainstream culture subsumes a variety of subcultures and idioms, meaning that using statistical methods to determine the characteristics of specific behaviors within a coarse-grained social group may overgeneralize. For a more nuanced approach, it would be a good idea to use ensembles, or add subculture membership as a feature to a linear regression classifier.
God.. CSS Visited Hack again. WHY does nobody come up with something? I know it's a fundamental issue, but hell.. this is a massive privacy exploit. Please Google? W3C? Someone?