OKCupid: Who believes compatibility ratings on dating websites?

The dating website OKCupid has pretended to users with "bad" levels of computer-assessed compatibility that they're well-suited. Can belief in algorithms make people like each other, asks Justin Parkinson.

Christian Rudder has publicly questioned whether the algorithms used by his dating website to recommend potential partners are "garbage". He experimented by telling users of OKCupid with a 30% "compatibility" rating that they were in fact much more likely to be suitable - either 60% or 90%.

Doing this caused a higher percentage to make contact by sending an initial message. But Rudder, a Harvard maths graduate, went further, looking at what proportion of each group got on so well that they sent four messages to each other. It almost doubled for those who had been deceived.

Rudder concluded that "the mere myth of compatibility works just as well as the truth". Does this mean faith in algorithms - a series of calculations based on data provided by users - is influencing whether people "like" someone?

"Priming" people in this way creates a "disinhibition effect", according to cyberpsychologist Berni Good. "It's interesting that people feel so comfortable with almost removing their sense of self, telling themselves they're compatible even when all the evidence - hobbies listed and that kind of thing - tells them they're not," she says. "It's probably going to end in tears when they actually meet."

Traditional notions of romance rely on indefinable spontaneity or "spark". Algorithms - series of calculations widely used by companies to predict consumer tastes - do not. OKCupid asks users about 350 questions to gauge their interests and personalities. But many are unconvinced.

"I strongly disagree with the need for algorithms," says TV presenter Sarah Beeny, founder of the mysinglefriend.com dating website. "Only the people themselves can see whether the magic's there, not a computer."

Rudder admits that "OKCupid doesn't really know what it's doing", but adds: "Neither does any other website."

Beeny acknowledges that there is a power of "suggestion", as exposed by the experiment. "But it's pretty amazing that he's come out and said all this," she says. "It seems like a bit of a Gerald Ratner moment. How can anyone trust his algorithms after this?"

Subscribe to the BBC News Magazine's email newsletter to get articles sent to your inbox.