OkCupid Sometimes Messes A Bit With Love, In The Name Of Science
July 29, 2014 — 2:10 PM
Listen to the audio:
OkCupid, the online dating site, disclosed Monday that they sometimes manipulate their users' profiles for experiments. Christian Rudder, co-founder and president of OkCupid, tells Audie Cornish that these experiments help the site improve how it works.
Source: NPR [http://www.npr.org/2014/07/29/336356931/okcupid-sometimes-messes-a-bit-with-love-in-the-name-of-science?ft=3&f=1003,1004,1007,1013,1014,1017,1019,1128]
AUDIE CORNISH, HOST:
The online dating website OkCupid claims it uses, quote, "math," to get you dates. You go to the site, answer a bunch of personal questions, and you're matched up with people based on your answers. Here's how OkCupid describes their system. It's extremely accurate as long as, A - you are honest, and B - you know what you want. Well, you may know what you want, but what users probably didn't know until yesterday is that the company sometimes manipulates dating profiles. OkCupid's president, Christian Rudder, disclosed this in a company blog post. He says the company has been running experiments with the data in the profiles. Christian Rudder joins me now from our New York bureau to talk more about this. Christian, welcome to the program.
CHRISTIAN RUDDER: Hi, I'm glad to be here.
CORNISH: So your blog post reveals the results of several OkCupid experiments. One, you strip the photos from people's profiles - right? - so that they had to decide...
RUDDER: That's right.
CORNISH: ...whether to pursue someone without the benefit of knowing what they look like - kind of virtual blind date. But in another instance, you tried to test the power of suggestion. How did you do it?
RUDDER: Well, we did in two ways. We took people who we thought weren't compatible, and we told them that they were. And then we did kind of the opposite where we took people who we thought were compatible and told them that we weren't. And then we basically monitored the conversations that these people had under these auspices.
CORNISH: How did they act?
RUDDER: The power of suggestion was powerful. It turns out that our own internal prediction, our match percentage which is our guess at your compatibility, was kind of half the story and then just the mere label was the other half.
CORNISH: So the mere suggestion that you're compatible is actually as good as really being compatible?
RUDDER: It seems like it, at least, you know, from our experiment. Of course the best thing - that it's doubly powerful to both be a good match and be told that you're a good match, which of course is what we do almost all the time.
CORNISH: Now, you've essentially sent people on bad dates in the name of science. What's been the response to you revealing this?
RUDDER: Well, my last interview - this British dude called me a sociopath, so there's that.
RUDDER: But, I mean, just to clarify, we did tell people after the fact - after we had collected the messaging data that they had been in an experiment and we gave them the correct match percentage.
CORNISH: You know, we asked some of our listeners to weigh in on twitter and several noted that in academia...
RUDDER: (Laughing) A fair and balanced medium if there ever was one.
CORNISH: Yes, a fair and balanced medium. Research involving human subjects is limited. It's monitored. It's run by an institutional review board. How is what you're doing - what other companies are doing ethical?
RUDDER: I'll just sidestep the fact that you at least have the charade of consent with these terms and conditions.
CORNISH: You call it a charade of consent. That's not a good sign.
RUDDER: Yeah, I mean - yeah, sure. But at the same time, look, when I was in college, you get 20 bucks to go sit down in the - you know I went to Harvard, and William James Hall is where they do all these psychology experiments. You need 20 bucks, you're like sure, I'll go sign up for something, and you sign a form. You have no idea what the experiment is. And so you - 'cause if there is information, it ruins the experiment. So this kind of good old days of past research standards - I don't even think - it doesn't pertain even there, you know? I think, also, experiment is such a loaded word. You know, I mean we test things all the time. I honestly cannot remember the last time we launched a significant feature on OkCupid that we didn't at least roll out to ten percent of the user base beforehand just to make sure that it behaved equivalently with the kind of status quo.
CORNISH: You know, another question that people asked us to ask you was did you give any consideration to the people who are genuinely looking for a relationship and feel betrayed?
RUDDER: Yes. We have a tremendous amount of empathy for our users. People are at that office because they like what we do. We create love in the world. We create sex - whatever people want, you know? Yeah, we do take that very seriously. Now, the betrayed part - I feel like that is the misunderstanding part that I'm kind of here to address or that I kind of poked in the eye in the blog post is that it's not betrayal. It's part of how we got to serving that first goal. Just consider the alternative for running the experiment that we ran. What if our algorithm had been garbage, you know? Then, if we hadn't run this kind of experiment and all these other kinds of experiments to test it, then instead of running garbage at a thousand users or whatever it was in this experiment, we're throwing garbage at four million people instead. And we don't - you've got to ask the question in order make sure you're doing a good job.
CORNISH: Christian Rudder, he's president of OkCupid and author of the forthcoming book "Dataclysm." Thanks so much for speaking with us.
RUDDER: It was my pleasure, Audie. Transcript provided by NPR, Copyright NPR.