News and views about writing, eating, gaming and watching

RSS

Can a dating site tell if you're a secret racist?

How the OkCupid website, started by four Harvard geeks, used statistics to unearth its users’ secret

True love: A neon sign by artist Chris Bracey. Photograph: Getty Images

When it comes to love and sex, how do you find out what people really feel, rather than what they say they do? Well, there are easier ways, but you could always start a dating website. Every time a user responds to a message, or clicks on a profile, she is telling you who she finds attractive, and who she's interested in starting a relationship with. Multiply that by a million or more and you have one hell of a database to plunder for insights.

That's what OkCupid did. Four Harvard graduates - Chris Coyne, Max Krohn, Christian Rudder and Sam Yagan - started the dating website in 2004. They'd previously run thespark.com, which offered personality quizzes such as the Purity Test (sample questions: "Have you ever fantasised about a family member? Have you ever fantasised about your own member?"), and decided to take the same slant with their dating venture. OkCupid users answer some, lots or occasionally all of the 4,000 questions written by the site or submitted by others about what they are looking for in a partner. They also give feedback on how useful the question was and how much weight they would give to the answer when setting up a date. That allows every candidate to create their own algorithm - the mathematical "secret sauce" that finds you matches.

While the site users were clicking away, trying to find someone to go for a drink with on Friday night, the founders were busy crunching their data. In June 2009 the first blog post popped up on OkTrends. "Since we went online in 2004, we've collected an enormous amount of data on human interactions," it said. "This blog was started as a way to share some of the things we've learned about people."

And boy, did they discover some interesting things. Some findings were quirky: that users of both genders added two inches on average to their height - even though shorter women got more messages. Or that using ur instead of you're or your in a first approach shrank the reply rate from 32 per cent to around 6 per cent.

In October 2009 a post titled "How Your Race Affects the Messages You Get" appeared. "We've processed the messaging habits of over a million people and are about to basically prove that, despite what you might've heard from the Obama campaign and organic cereal commercials, racism is alive and well," Rudder wrote. "It would be awesome if the other major online dating players would go out on a limb and release their own race data, too. I can't imagine they will: multimillion-dollar enterprises rarely like to admit the people paying them those millions act like turds."

Rudder showed that the percentages of matches were roughly even across all races. But white men got the most responses from almost all ethnic groups; white, Asian and Hispanic women preferred them to the exclusion of everyone else. Black women, on the other hand, get a bum deal - even though they reply more often than any other group to messages from every race, including their own, their messages get by far the fewest replies. "Essentially every race - including other blacks - singles them out for the cold shoulder," Rudder wrote.

He contrasted users' actions with their words: only 6 per cent overall said that interracial marriage was a bad idea, and 38 per cent that they would "strongly prefer" to date someone of their own racial background. (Among white users it was 45 per cent and among non-whites, 20.)

After that, the blog tackled such taboos as rape fantasies (these are deemed much more acceptable in Nevada than in New England, and in Lithuania than actual England) and even questioned the extent to which bisexuality exists. Noting that 80 per cent of self-identified bisexuals were interested in only one gender, Rudder concluded: "This suggests that bisexuality is often either a hedge for gay people or a label adopted by straights to appear more sexually adventurous to their (straight) matches."

Strange combinations

Last April, however, postings on the blog abruptly stopped. What happened? As the OkCupid number-crunchers would say, correlation does not imply causation, but it's hard to feel it wasn't anything to do with the site being acquired for $50m in February by its paid-for rival match.com. Was it intending to suffocate its free competitor? When I asked Yagan, the OkCupid chief executive, he said the blog would return but refused to be drawn further.

Then again, running a dating site exposes you to bits of humanity that maybe are best hidden. OkCupid was rare in making its data public, but our hidden prejudices and preferences are clearly well known to those in the industry. When I emailed Markus Frind, founder of Plenty of Fish, to ask about his rivals at OkCupid he said his matching system was better, because "we look at trends or patterns in couples . . . A female doctor is never going to date a carpenter. There are many, many combinations of relationships that will never happen or are very unstable." And he had the data to prove it.

What OkCupid showed was that, when it comes to choosing our partners, none of us is as progressive as we think. Perhaps the world is a better place for not revealing our deepest, darkest secrets?