29 January 2008

Finding a mate: the algorithmic approach

This is one of the more visible applications of artificial intelligence.

New York Times, "Hitting It Off, Thanks to Algorithms of Love":

eHarmony [...] pioneered the don’t-try-this-yourself approach eight years ago by refusing to let its online customers browse for their own dates. It requires them to answer a 258-question personality test and then picks potential partners. The company estimates, based on a national Harris survey it commissioned, that its matchmaking was responsible for about 2 percent of the marriages in America last year, nearly 120 weddings a day.

[Academics] are skeptical, because the algorithms and the results have not been published for peer review. But they also realize that these online companies give scientists a remarkable opportunity to gather enormous amounts of data and test their theories in the field. EHarmony says more than 19 million people have filled out its questionnaire.

... [One] thing is already clear. People aren’t so good at picking their own mates online. Researchers who studied online dating found that the customers typically ended up going out with fewer than 1 percent of the people whose profiles they studied, and that those dates often ended up being huge letdowns.

The last bit is unsurprising. How much information could you get about a person from a blurb, and how much of that is likely to be be useful information?

No comments:

Post a Comment