In late July, Christian Rudder, a co-founder of the online dating site OkCupid, plunged himself into the middle of an Internet maelstrom when he published a post with a classic poke-the-anthill headline: "We Experiment on Human Beings!"
The provocation came in the middle of a storm of commentary sparked by the revelations that Facebook had been purposefully manipulating its users' emotions by tinkering with its news feed. Rudder contended that such tweaking was commonplace and normal. In OkCupid's case, the company had temporarily adjusted its matching algorithm so that some people ended up with recommendations that the algorithm would normally have considered bad matches -- and vice versa, some people whom the algorithm should have concluded were good matches were told they were a bad fit. There was no ill will involved; from Rudder's perspective, it was just an experiment designed to serve the larger goal of improving the overall OkCupid user experience.
The Internet reacted harshly. But in an unplanned twist, the post turned out to be good publicity for Rudder's new book, "Dataclysm: Who We Are When We Think No One's Looking." Case in point: I had an advance review copy of the book sitting on my desk, but it was only after the hoopla over Rudder's blog post that I took a closer look and decided it was a must-read.
And indeed it is. "Dataclysm" is a well-written and funny look at what the numbers reveal about human behavior in the age of social media. It's both profound and a bit disturbing, because, sad to say, we're generally not the kind of people we like to think -- or say -- we are.
Rudder begins his book with a distressing opening salvo: two charts that reveal what age groups men and women generally find attractive. From age 20 to 50, women are consistent -- they're drawn to men who are in roughly the same age cohort. Men are equally consistent: From age 20-50, they are attracted to 20-year-olds. The discussion is over: Men are dogs.
Rudder's data on race leads to similar implications -- prejudice is alive and well on online dating states, and what we say -- and don't say -- in our profiles offers impressive support for cultural stereotyping. Rudder does the math on what different groups are most or least likely to say in their profiles: Black men, for example, hardly ever mention Belle and Sebastian, snorkeling or "Dr Horrible's Sing Along Blog." White women don't talk about slow jams, j-pop or Malcolm X. White guys, however, are really into mentioning their "blue eyes," brewing beer, and Robert Heinlein. Asian men frequently say "tall for an Asian," "gangnam style" and "noodle soup.")
Rudder treats these insights into the human condition with bemused -- and very useful -- intelligence. We're only just beginning to understand how much we can learn about ourselves and others from the data that is constantly being harvested from us. The more we know, the better armed we are to navigate the future.
Rudder spoke by phone to Salon from OkCupid's offices in New York.
So men are sexists, and we're all racist?
The more you look at the data, the more it does confirm the cynics' intuition about humanity. People online are free to act out their worst impulses with very little incentive to act out their best. I guess it just goes to show how politeness or propriety keeps us decent human beings. Offline, society actually has a very good effect on behavior in a very large sense.
That raises an uncomfortable question: Does our wholesale move online undermine how society traditionally keeps us in line?
I'm not qualified to give a real opinion on where society as a whole is headed, but I think when you look at stuff like rage storms on Twitter, or even the thing that happened yesterday -- the celebrity nude photos being leaked -- you see that there are definitely some disgusting impulses that the Internet can gratify instantaneously. In the same way Cool Ranch Doritos gratify certain taste receptors that are probably not very good for my digestive tract, things like Twitter or Reddit or even OkCupid gratify our tastes in ways that should probably best be left unsated.
How does that make you feel as a researcher? Have you become more cynical as a result of what you've learned by watching how people behave on OkCupid?
I definitely have a certain amount of ambivalence about the Internet generally and what we do at OkCupid. OkCupid does a lot of great things. We do find people love, we do create marriage and children and happiness in a pure sense, in a way that, say, Amazon does not. But there is a downside: In the process of finding that love or sex or whatever they're looking for, people are able to be more judgmental. It's a fraught thing. I can see the good and the bad in all this, but where it all comes out in the end, I'm not sure. I think the existence of the Internet is a good thing, but I do wish people exercised more humanity in using these tools.
I'd like to break the format of the typical Q&A a bit, and just read some lines from your book that jumped out at me, and see if I can prompt you to elaborate on them. For example, you wrote that "the Internet will democratize our fundamental narrative." What does that mean?
What I meant was that the Internet will enable, on a mass scale, something like what Howard Zinn was doing in his "People's History of the United States." Zinn's trying to reach for what the common person thought about World War I or the Civil War, or go back and find out what a housewife in 1970 was thinking about her life. But by and large he had to put it all together from a few diaries and a ton of leg work and obviously there's a lot of selection bias involved.
But in the future, as people continue to live out their lives through these technologies, all of our lives are almost by definition going to be captured. The computer that is crunching all that stuff pulls us all together. In a very real sense, we are all given the same weight in any of these calculations.
I guess that connects directly to another sentence that caught my eye: "With data, history can become deeper, it can become more."
That's exactly what I'm talking about.
How about, "It's when people don't understand their own hearts I get interested"?
I like it when you are able to look at a behavior in two ways. One: what people think they are doing or wish they were doing, and two: what they actually do. At OkCupid we have a great mechanism for looking at that: We have all these match questions where we ask people what they believe or what they think, and then we can go in and measure exactly what they are actually doing. I just think that the space between self-image and action is very interesting.
What data points jumped out at you the most?
Well, the most obvious thing is racial messaging patterns. We asked people about race and everybody is like, yeah, interracial marriage is totally great. Something like 96 percent are totally fine with it, or support it. We also asked people questions like "would you ever date someone who told a racist joke" and the answers are very strongly liberal in the way you would expect. Everybody is fine with it, blah blah blah. But then you go out and look at what people do or who they choose for themselves, and you see that this is just not the case. Race is a huge factor and certain types of interracial relationships -- I wouldn't say are taboo, but certainly in the aggregate they are less desirable.
Again this gets back to what we were talking about at the very beginning. If that's what I want why don't I just put that into the form? It would work better, if I was just honest with OkCupid and myself about what I wanted.
You mention Naomi Wolf's "The Beauty Myth" and you wrote, "for the beauty myth, social media signifies judgment day." Is this just a reflection of the fact that women who are considered highly attractive get by far the most messages from men?
I was having a little bit of fun. There's just so much judgment that goes on in social media. If most myths are built around some kind of cataclysm or apocalypse, then for the beauty myth, Ragnarok is social media. Men who are free to judge photos without conforming to social norms go crazy clicking girls in bikinis.
Maybe the most discomfiting point you make in your book is your acknowledgment that the kind of people who work for the NSA crunching our data are much smarter than you are and have access to far more information. Eventually, the sophistication of the algorithms will become so great that pretty much everything important about us will be inferred from just a few data points. That's scarily determinist. Do we even have free will when our data trail tells employers or the government or prospective mates exactly who we are?
That is a great question, and I don't think I can give an answer that is both hopeful and honest. The tech industry side of me wants to say that this isn't just a problem of social media -- the same thing happens with your credit score, for example. But you are right. It is scary. There will always be highly motivated, powerful entities using this data for their own good, which often implies an adversarial relationship against you. I will say one thing: If we consider Facebook as stand-in for all this stuff, I think people have generally approached these social media networks with a level of naiveté that is changing. We're beginning to understand the pitfalls of volunteering all this data about ourselves.
That's why a book like "Dataclysm" is important. The more we know about what you guys are finding out, the easier it will be to set societal guidelines for how this information can be used, and to become masters of our information.
Exactly right. It's a strange time for me and I'm sure for you too and anybody else working in this milieu. The technologies are pervasive but comprehension of them is not.
Which leads me to my final question. Let's revisit that experiment in which you tweaked the matching algorithm. I think for a lot of people that smacked of manipulation that crossed over the line. It seemed different than just changing the layout of a page to see what works better. It seemed like you were messing with people's minds. Why did you do it?
Let me just step back and add a little more context. So, we tweaked an algorithm. Now, some algorithms can be considered as a sort of fact. If you are trying to pull a record out of a database there is a canonical or fastest way or best way to do it and to deviate from that would be silly or would be wrong in a real sense. But when we describe people as good or bad matches -- the truth is for any two people on OkCupid, we just don't know. We're making a guess; our algorithm is a version of a guess. It's not a fact.
There are tons of different ways to bring people together. We often use common interests, like how well you and I satisfy each other. But there are other potentially workable heuristics, like, for example, "opposites attract." The test I wrote about in that blog post was on a continuum of those kinds of tests: We were really genuinely trying to figure out what works best, how to improve the user experience.
What we were doing was different, to me, than "lying." Lying would be distorting matters of fact, rather than opinion. I have no idea what your sexual orientation is, but just imagine if you were gay, and I go and tell people that you are straight. That's very clearly false, and possibly harmful. We would never do that because that is altering a fact about people ... But with any algorithm that is about how to recommend something -- there is no canonical perfect way to do it. So we treat it sort of like an opinion.
But doesn't that enter a fuzzy area? A selling point of OkCupid is supposed to be that it actually works, which implies that your "opinions" as to who is a good match are actually facts ...
For sure. For sure. But part of what makes us sure that we can give people the best match, and that we can make good guesses about what two people are going to get along, is that we are constantly working on refining our methods.
Look, I definitely understand the feelings about what we did. Especially given the way that I first laid it out, and then later, in the way I reacted to the media. Both my presentation and reaction were flawed. But we did not do it to mess with people. Everything we do at OkCupid is done with discretion, and, I hope, some level of emotional intelligence.