In late July, the internet was shaken by a shocking revelation (even by internet standards): Christian Rudder, co-founder and president of OKCupid, announced that the dating site, one of the most popular on the web, had been collecting data to learn more about its users. Not too much to be surprised about there; unless you’re really naive, you probably realize that just about everything you do on a website may be contributing data that these sites can use – or even sell to interested third parties. But what was unusual about this situation is that OKCupid decided to go beyond passive data mining. Recently, a percentage of users were notified about good matches who were, unbeknownst to them, actually considered bad matches by the site’s algorithm. Some users were also notified that certain members were bad matches, when, according to the algorithm, they were quite compatible.
Why did the powers that be at OKCupid do this? There could be a bigger picture, but based on Rudder’s blog post about the topic, it seems like it came down to curiosity. I don’t know – while many of us dream of coming up with a website that becomes a huge success and makes us rich, maybe after it actually happens, you get bored. You could also blame a bit of it on human nature, too, I think. I mean, the experiment does raise some interesting questions. If you don’t already know, I bet at least some part of you is wondering how it played out. Well, it turns out that the people who were told bad matches were good ones, often ended up contacting members they otherwise probably wouldn’t, and frequently exchanging multiple messages with them, which OKCupid considers engaging in a conversation. Ultimately, it says a lot about attraction and the power of suggestion – and maybe also about just how useful internet dating sites are at playing matchmaker. Is it true that sometimes all it takes to bring two people together (at least for a little while) is positive reinforcement and reassurances?
Whatever you might think about the experiment, was it ethical? Rudder claims it was, saying pretty much what I did earlier in this post: “[G]uess what, everybody: if you use the Internet, you’re the subject of hundreds of experiments at any given time, on every site. That’s how websites work.” Some legal experts agree. But many other people in the internet world (including the FTC, who is, at the time I’m writing this, considering opening an inquiry into the issue) don’t.
Parallels are being made between this experiment, and one recently conducted by Facebook, in which users’ news feeds were manipulated in order to see how much this affected their moods. Interestingly, people seem to be making a distinct difference between the two studies. For one thing, OkCupid’s manipulation had a financial impact on some users – people spent money on dates that were bound to lead nowhere. And of course, worst of all, giving false hope to lonely hearts is a low blow. Reprehensible as the study might be, could it be a sign of things to come? Could data mining in general go from passive, to much more invasive? Not likely – if a site is looking for academic credibility; as this article points out, a study like OKCupid’s “cannot be described as particularly high-brow academia, because it is only taking into account one set of metrics — the only ones at its disposal. Aside from the personal details the customer has supplied at the time of signing up (details not verified for accuracy) little is known about the subject’s background, and no followup [sic] interviews allow for in-depth analysis.”
On the other hand, if a site’s owners are curious, or bored, or find some way to benefit financially or otherwise, why not? It will be interesting to see how things evolve. If nothing else, let’s take this as a reminder not to trust anyone or anything on the web too much.