Two things: First, OkCupid’s OkTrends blog is back. This is incredibly exciting, as OkTrends was one of the best parts of the online dating site in its earlier days (and for the curious, no, it wasn't shut down by Match.com when Match acquired OkCupid). And second, Facebook isn’t the only website that’s been experimenting on its users; in its first post in over three years, the blog revealed that OkCupid itself had performed not one, not two, but a whole bunch of similar experiments over the ten years the site has been around. In the blog post, they walk us through three different experiments they’ve run; the first one is a little more transparent — it involves how usage of OkCupid launch day celebration for a new blind date app compared with normal, everyday usage of the site — so I’ll let you head on over there to read it yourself. The second two, though, fall into a similar realm as Facebook’s much-debated experiment, so let’s take a look at those. Ready to fall down a deep and bizarre rabbit hole?
The first examined how much a profile photo tends to be worth to users in comparison to the actual text of the profile. Originally, OkCupid gave users two scales on which to rank matches: Personality and looks. After some time, it became apparent that “personality” and “look” were pretty much the same thing to people. They know this because the user with this profile picture:
Had no actual text in her profile picture — and yet was still ranked in the 99th percentile for personality.
So they got rid of the two scale system and replaced it with just one. They still suspected that users were still rating matches primarily on photos, though, so they performed a little experiment: For a space of time, they took a small sample of users — they don’t specify how small — and, half the time they displayed their profiles, they hid the text. This way, they gathered two sets of scores, one for both the picture and the text, and one for just the picture. They found that the text accounts for less than 10 percent of what people think of you in the online dating scene.
Yikes.
The second of these two notable experiments delved deep into the question as to whether their match algorithm actually works. According to their internal measures, it does — but they often wondered whether people reported liking their matches because their match percentage suggested to them that they were supposed to like each other. So, they took a whole bunch of people who only had a match score of 30 percent and told them they were actually a 90 percent match. When they were told they were more compatible with someone, users ended up sending more first messages.
But it didn’t just stop there. The odds that the wrong-for-each-other pair exchanged at least four messages — a number OkCupid considers to be the point when a real conversation develops — increased a huge amount when they were told they were right for each other: They only hit the four-message threshold 9.7 percent of the time when they thought they were a 30 percent match, but hit it 17.4 percent of the time when they thought they were a 90 percent match.
And then they went even deeper, Inception-style, and took a look at a whole matrix of possibilities: What happens when you tell bad matches they’re bad ones, bad matches they’re good ones, and good matches they’re good ones. The ideal situation, they found, occurs when good matches are told they’re good matches — then the chance of a real conversation developing becomes 20 percent. Conclusion: The algorithm works — but it doesn’t exist in a vacuum.
The question, of course, is whether all these online experiments are ethical; after all, none of the participants expressly gave their permission to be used as test subjects. OkCupid makes the remark, however, that “if you use the Internet, you’re the subjects of hundreds of experiments at any given time, on every site. That’s how websites work.” And you know what? They’re kind of right. We see it in everything from how sites personalize the content they display for you to how advertisers target your interests and show you stuff you’re more likely to click on.
Is this “right?” I don’t have an answer to that; I do know, though, that if it bothers you, the solution is easy: Just stay off the Internet altogether. After all, we may love funny cat videos, but we can definitely survive without them.
