November 28, 2007
Love plus one
I neglected to alert y’all to Monday’s More or Less - which covered quants and the credit crunch, extreme weather and climate change, a 20-year old from Birmingham who solved an international mathematical puzzle - but you can download a podcast if you like.
Particularly recommended was Ruth Alexander’s experience in trying to find love using a dating algorithm:
First of all, the website invites me to answer a long questionnaire. How much reassurance do I need from a partner? How much do I give to charity? How tidy is my home?
"The majority of questions represent statements commonly heard by therapists in couples counselling," says True.com psychologist, Dr Garth Bellah.
The website then uses what statisticians call a regression equation to determine what sort of person I would be best matched with, according to my character and how that fits with historical data about other people’s relationships.
The company says it’s identified 99 distinct factors found in successful relationships. Another dating site says there are 29 - its mathematical match-making is based on research it says it’s done on 10,000 married couples.
Looking at the profiles of the men the computer highlighted for me, I was highly sceptical, bordering on horrified. But I gritted my teeth and sent a flurry of e-mails: would you like to go on a date with me and my microphone?
It turns out the compatibility test doesn’t yet measure aversion to journalists. Only two people e-mailed back - and one of those dumped me for someone else before we’d even met.
So I was left with one chance of love. And he was wearing a burgundy bow-tie and waistcoat. But our compatibility score was 94%. The computer said yes.
This was terrific radio. A fair test of the dating algorithm? No.
We were put onto the scent by Ian Ayres, author of Supercrunchers. He thinks that algorithms can do a better job than, say, friends, in finding a good love match. I am persuadable, but not convinced that the current products - which are confidential black boxes - actually do the job.
Here’s an alternative take on the maths of dating, in which I point out that we tend to adjust quickly to "market conditions" in accepting potential dates. More on that in The Logic of Life.
I find the old fashioned way - where you meet people through people you know and trust - to be the best. If the research is wrong, then no matter what algorithm you write, it’ll always bring up the wrong results. Better I feel to use a learning algorithm (such as bayesian) where it starts to learn what type of people you are interested in
Posted by: John cage | November 28th, 2007 at 2:06 pm | Report this commentI wonder how the dating algorithm defines a “successful” relationship? Perhaps they asked 10,000 married couples and then only used the data from the 5,000 who eventually died without having broken up. Otherwise the criteria for “success” seem to involve nothing more than happening to be married at the time of the survey, which I suppose is success of a sort but - given the ease and swiftness with which people can marry and divorce - hardly anything I’d like to base a compatibility test upon.
Posted by: Seamus McCauley | November 28th, 2007 at 3:27 pm | Report this comment