By Inga Ting
The Sydney Morning Herald
February 14, 2015
Click here to view original
In a world awash with data, algorithms aimed at “finding your perfect match” have reached a new plane of sophistication. The sales pitch of New York-based website OkCupid, founded by a group of Harvard mathematicians, says it all: “We use math to get you dates.”
Co-founder Christian Rudder spends his days crunching data from millions of users to come up with insights such as the single best predictor of sex on the first date is whether someone likes the taste of beer.
We like the idea of maths-based matching because it suggests we can set limits to a game of chance, says sociologist Lauren Rosewarne, of the University of Melbourne.
“We’re all control freaks. We like to think, ‘if I just work hard enough, if I sign up for a website, if I just apply the appropriate skills … I’m aiding luck finding me’,” Dr Rosewarne says.
Matching sites still start by asking users who they are and what they’re looking for, but their focus has expanded from measuring compatibility to also trying to predict “chemistry”.
Consider this. The biggest turn off for men on the eHarmony dating site, according to the company’s own research, is women “using the computer”.
We might laugh at the irony, but it illustrates a crucial point. “What we know about dating is that what people say they want, and what people really want, are very different in practice,” says Kari Taylor, marketing director of Fairfax-owned matchmaking website RSVP.
RSVP switched from profile-based matching to behaviour-based matching a few years back and saw an 80 per cent surge in users saying yes to conversation requests.
Behaviour-based matching is adaptive. It compares what you said you wanted with how you behave to work out things you might not even know about yourself.
For example, you said you wanted a partner with a steady income but you keep messaging “pro-bono computer game testers” and “freelance writers”, so the algorithm changes its recommendations.
Similarly, the algorithm knows that attraction is reactive – we’re more likely to be attracted to people who are attracted to us. So when someone likes you back, the algorithm analyses their behaviour and characteristics, and brings you more people just like them.
And it gets more sophisticated. Let’s say you had some success with Karen, Emma, Jane and Lilly. Through a process called triangulation, the algorithm identifies other people who hit it off with those women and introduces you to some of the other women they liked – even though you may have nothing in common other than your taste in women.
It’s a Venn diagram of attraction formed by millions of spheres of explicit and implicit desire. And for many couples, it works.
Janin Mayer and her husband Evi Bitran met on RSVP in 2011. They married 18 months later and now have a six-month-old son.
“From everything I could see, he was absolutely not what I was looking for,” Janin says.
“I’m an artist, he was studying a [Masters of Business Administration] … We definitely wouldn’t have met if it wasn’t for the internet. We are complete opposites.”
Evi seems to prove her point. “Immediately when I saw her profile picture, I wanted to talk to her … And no, I think even without the internet, the universe would have brought us together.”
eHarmony vice-president of matching Steve Carter says very little about our personalities, values or appearance tells us when sparks will fly.
“Predicting who someone is going to want to talk to is vastly more difficult than predicting who they would have a good marriage with,” he says.
“People choose who they’re attracted to based on a much more chaotic, much more subjective and very reactive set of characteristics.”
eHarmony’s predictive model for attraction has improved up to 400 per cent in the past couple of years, Dr Carter says. Which isn’t quite as exciting as it sounds. “We’ve gone from predicting a minuscule amount of the variance to predicting a very small amount.”
Carter believes an algorithm for attraction is possible given how much data is available. But the real question is whether that’s the direction the industry is heading.
“You talk about all this data being available,” he says. “But the biggest thing in the online dating universe … is an app that doesn’t use any information except a photograph.”
That app, of course, is Tinder. Launched in 2012, Tinder has taken the dating scene by storm by making a game out of judging people based on their photos. Search through users within a given radius and swipe right to like, left to reject. If someone “right swipes” you back, it’s game on.
There are no algorithms predicting attraction; no models measuring compatibility. Tinder relies on two of the oldest ingredients for romance – how someone looks and how near they are. The rest is serendipity.
“You could say it’s about judging people based on their appearance and you might say that’s shallow but that’s actually what happens in real life,” Dr Rosewarne says.
On the other hand, predictive algorithms do deliver surprises. Just look at the Match.com couple interviewed in Dan Slater’s book, Love in a Time of Algorithms. He was asthmatic. She was a smoker. Several (hundred thousand) triangulations later, they met and fell in love.