The Washington PostDemocracy Dies in Darkness

The one thing about ‘matching’ algorithms that dating sites don’t want you to know

Tinder released an updated version of its matching algorithm today, a “big change” that CEO Sean Rad has been hyping for the past week. In a blog post, Tinder offered few details on the new algorithm — but basically promised that it would revolutionize the quantity and quality of matches each user receives.

“Just open Tinder to check it out,” they encourage. “We’re sure these updates will make swiping even better and will lead to more meaningful matches.”

But here’s a little factoid about that new algorithm that Tinder presumably will not be trumpeting: Dating site algorithms are meaningless. They really don’t do anything. In fact, the research suggests that so-called “matching algorithms” are only negligibly better at matching people than random chance.

[Absolutely everything science knows about how Tinder’s changing the way we date]

The strongest evidence for this comes from a 2012 paper published by Northwestern University’s Eli Finkel and four co-authors in the journal “Psychological Science in the Public Interest,” which not only eviscerated the very concept of matching algorithms, but called on the Federal Trade Commission to regulate claims about their effectiveness.

To understand why these authors found these claims so troubling, you first have to understand some basic things about how relationships work. Leave aside, for a minute, your Disneyland notions of soulmates or true love: In reality, most people could happily pair off with a large number of potential partners, and the factors that determine whom they do pair with have as much to do with circumstance as anything else.

Relationship success basically depends on three things, Finkel et al. explain: individual characteristics, like whether you’re smart or what kinds of hang-ups you have around relationships; quality of interaction, or how you hit it off in-person; and surrounding circumstances — stuff like your race or health or financial status.

Right off the bat, this proves a major obstacle for matching algorithms. They simply can’t account for your future circumstances or the way you’ll jibe with another person, particularly before you’ve met; they might attempt to model those things, but there’s not enough input data to account for the diversity of possible outcomes.

[What you don’t know about Internet algorithms is hurting you]

Given that, matching algorithms tend to focus on personality alone — matching you with someone who’s similar to you, or similar enough that you won’t instantaneously swipe them off your phone. But that presents its own problems: like the fact that major, large-scale studies of married couples have shown that the similarity of partners’ personalities accounts for only half a percent of how happy they are. (Tinder, curiously, has just begun adding job and education data to its profiles, too, presumably so you can pick people who have similar backgrounds to you.)

On top of that, there are so many questions about compatibility that psychologists haven’t answered. Like: Given a mishmash of data points on which you and a partner do and don’t match, which ones should the algorithm privilege when calculating your “match percentage”? (OkCupid does this, for instance, by letting users rate how important these data points are to them; but people are notoriously bad at rating that kind of thing, and mathematically speaking, it’s a blunt instrument.) And what if you lie about what you’re like, or what you find attractive in someone else? Or what if your beliefs and personality change between the time you began using a site and the present moment?

Worse, how can the algorithm account for a basic, well-documented quirk of human nature: that people are actually pretty whimsical about whom they’re attracted to?

“It is virtually impossible to succeed at the task many matching sites have set for themselves,” Finkel et al. conclude. “… Despite grand claims to the contrary, it is unlikely that any matching algorithm based upon data collected before people have encountered each other can be effective at identifying partners who are compatible for a long-term relationship.”

In other words, Tinder’s claim that it can algorithmically make matches more “meaningful” is … basically bull.

That said, of course, Tinder is no or eHarmony; if you’re swiping through your Tinder matches on a Friday night, you are presumably not on a quest to find your one-and-only. That’s ideal, because — while apps are really bad at predicting relationships long-term — they’re very good, the research suggests, at helping you meet more people.

So if your goal’s a quick drink or a short-term hook-up that potentially leads to something more, then by all means — swipe on! If it’s “meaning” you’re after, though, then both you and Tinder may be going about the whole thing wrong.

Solo-ish editor Lisa Bonos let her parents try out the popular dating app, Tinder. (Video: Pamela Kirkland/The Washington Post)

Liked that? Try these!