Can dating-site algorithms actually calculate true love?

I recently made my first online dating profile.

I uploaded a profile photo, filled out a roster of stats about my body type and drinking habits, and started answering questions like “Do you recycle?” and “How much do you usually tip a server who did a fine job?”

Then I asked my boyfriend of eight years to create a profile of his own.

No, we’re not breaking up or trying out an open relationship. I was conducting an experiment to see how online dating works—specifically, how accurate the matchmaking features on dating sites are. Millions of people use these sites every day to try to find something I already have—a loving long-term relationship—and I wanted to know if they were going about it all wrong.

Let me back up. I met my boyfriend on our first day of high school. I fell hard for him—he was smart and funny, could play multiple musical instruments, and already had a beard—but we didn’t start dating until college. As it turned out, high school-freshman me somehow had extremely good taste. My boyfriend is mature and honest, never picks a fight or raises his voice, and sees men and women as equals. And he’s still a talented musician with a beard, to boot. We’ve been together for eight years now, and it's hard to imagine two people more compatible for long-term coupledom. We’re both introverted atheists who are cautious with money, and don’t want kids. We make each other laugh, and love spending time together. We would've gotten married a long time ago except neither of us believes in marriage. What we have, in short, is True Love.

When my friends start dating someone, I wish for them what I have. I want their partners to support them, and say they love them every day. I want my friends to tweet all the cute things their partners do just because they’re so delighted. Unfortunately, a lot of my friends are still searching.

And because I’m a millennial, many of my friends are searching online. I never had a reason to do any online dating, so naturally, I’m curious about it. Are dating sites actually effective in matching people up? Do they connect people who seem like a good match on paper, but don’t get along in person? Is data-based dating any better than its analog counterpart at helping people find their soul mate? 

Would my boyfriend and I still have ended up together if we met online?

To find out, I ran a subjective and extremely unscientific experiment. I created profiles on several dating sites, and had my boyfriend do the same. Then I set out to answer the question: Can dating-site algorithms calculate true love?

For most sites, the answer seems to be a clear-cut no. Match.com asked me to provide only the barest information, and paired me with people because neither of us smoked. To answer more in-depth questions, I had to link my account to Chemistry.com, which said my password was invalid and never actually sent me a password-reset email. Meanwhile, eHarmony kept telling me to relax my standards, and date much older or younger men. Other less well-known sites and apps were similarly disappointing. Their software didn't seem capable of engineering a decent match, let alone love.

The only one that seemed to have an algorithm that could hypothetically work was OKCupid. If you don’t already know, the OKC system works like this: You answer multiple-choice “match questions” such as “How important is religion/God in your life?” and “Other things being equal, would you be more attracted to an artist or a scientist?” You also specify how you want your date to answer, and rate how important the question is to you. OKC then uses this data to create a “match percentage," which measures how much you and another user have in common. When you look at a potential partner’s profile, you see this number at the top. There’s also a second figure, the “enemy percentage,” which measures conflicts between two users’ information.

After my boyfriend and I created profiles, and answered 100 questions each, OKC determined that we're an 85% match and 4% enemies:

As you can see, we agree on most things. (He’s the sloth, on top. I’m the human, below.)

We did, however, differ on the issue of scary movies. (My answer is in red because my boyfriend accidentally didn’t include it as an “acceptable” answer.)

OKC’s number-crunching indicated that we're pretty compatible, but I could see plenty of other guys who had an even higher match percentage with me. So, my boyfriend and I descended back into the mines, and chipped away at another 100 questions.

Then here's what we got:

An even higher match and lower enemy percentage—but we still hadn’t broken 90%. Did that mean OKC’s math was flawed? Were these men actually a better match for me than my boyfriend? Or were there simply elements of a happy relationship that the algorithm couldn’t measure?

As I clicked around people’s profiles, it quickly became clear that a high match percentage was not necessarily an accurate predictor of lifelong romantic bliss. Sometimes, another user felt the same way I did about the importance of art, or agreed that dedication is more important than passion in a relationship—making us seem super simpatico—but then it turned out he wanted kids or believed creationism should be taught in schools. Even men who had very high match percentages with me gave answers that I considered deal-breakers. For example, they said they didn’t consider themselves adults yet, or marked it as unacceptable that I had a master’s degree when they had a bachelor’s.

I knew from the start that I’d never be compatible with someone who wanted kids or was religious, but some of these answers brought up issues I’d never considered (and never had to). It didn't occurred to me that a man might dislike that I went to school for a few years more than he did—but even a 99% match can't make up for that kind of sexism.

OKCupid emphasizes the match percentage over the enemy percentage, displaying it more prominently and in more places. But maybe it’s actually the enemy percentage that’s more significant.

Take a look at “Jason," who an automatically generated email told me was “a really good match.” Jason and I have the same match percentage as my boyfriend and I got (88%), but with Jason, I have a much higher enemy percentage (20%).

Jason and I are both atheists and pro-choice. Neither of us follows sports, or would date someone who kept a gun in the house.

But then there’s this:

And this:

And these:

Jason and I would not make a good couple. He says he couldn't respect a woman who had sex on the first date, and that eye-rollingly misogynist opinion ensures that I never would with him. We agree on things more often than we disagree, but the disagreements utterly invalidate the agreements.

I saw this pattern play out again and again in the profiles I looked at: The enemy percentage could void the match percentage, but not vice-versa. Not only does this point to something larger about the way OKC's algorithm works, but also about successful relationships in general.

To be in a mature, equal partnership, you need to know what you refuse to put up with it (I’m not talking about abuse, though it goes without saying no one should have to endure that). Personally, I refuse to be with a man who doesn’t see women as fully equal, or who holds conservative political or religious beliefs I consider harmful or lacking in empathy. Or, in OKCupid vernacular, I don’t want to waste my time with someone who has a high enemy percentage.

To be in a mature, egalitarian partnership, you need to know what you refuse to put up with it.

But the match percentage is a little different. Of course, certain commonalities help build a healthy relationship, mostly big amorphous things like respect and communication. But two people don’t need to both like documentaries or spicy food to fall in love. In fact, if I had filled out my OKC profile before I started dating my boyfriend, I would’ve said I liked neither. And my boyfriend wouldn’t have listed Fiona Apple or St. Vincent among his favorite artists. We would have had a lower match percentage, but we would still have been compatible, just like we’re still compatible now—even though our match percentage remains in the 80s.

Perhaps it would break into the 90s if I answered yes to all the questions about playing video games, and he answered yes to all the questions about being active on social media. But instead, we answered no, and marked as “acceptable” all possible answers that a potential date could give. I enjoy wasting time on Twitter, but I really don’t care if the person I date does the same. When you mark all answers acceptable like this, OKC says you’ve rendered the question “irrelevant,” and warns you against doing it too often. But those questions are irrelevant. Or, rather, they refer to the kinds of things you get to learn about from each other, and explore together. They’re part of the joy of discovering all the surprising nooks and crannies of this wonderful brain you’re connecting with. But they say nothing about whether you’re ultimately right for each other.

“Okay,” you’re saying. “So you figured out that the ineffable mystery of true love is difficult to quantify via multiple-choice questions. Big whoop.” But actually, I think OKCupid did a stellar job. Those numbers were accurate!

It’s just that, like any technology, humans can choose how to use it. And, like any technology, lots of humans will choose to use it in stupid ways. Like the guys who kept messaging me even though my profile said I was only on the site for a writing project, had a boyfriend, and was not looking to date anyone. Many of them had low match scores and high enemy scores. They seemed to be messaging me solely based on the fact that I was blond, which is the same trait that men often mention when they catcall me on the street. Way to take a sophisticated algorithm, and use it like the crudest and least-effective pick-up method of all time!

Way to take a sophisticated algorithm and use it like the crudest and least effective pick-up method of all time, dudes!

But you can also use technology intelligently, and dating sites give you tools beyond algorithms. They let you read people’s profiles to see how they talk about themselves in their own words, which helps you feel a spark (or lack thereof) in a way that numerical values could never provide. When men shared my political views, but wrote preachy, humorless paragraphs about them, I closed the browser tab. When they had funny, interesting things to say, I noted them as people I would’ve been interested in messaging. (Of course, messaging and then meeting someone gives you even more information than a dating profile, but my experiment could only go so far.) 

My boyfriend writes funny, interesting things every time he texts me. We have a decent match percentage, an exceedingly low enemy percentage, and did I mention he’s a musician with a beard? All of that makes me believe that if we had met online, we would've definitely started chatting with each other.

So maybe dating sites haven’t yet perfected the super-powered artificial intelligence that will one day distribute marriage assignments to the general population. And you can’t stop people from using these sites in less-than-exemplary ways. But from what I can tell, online dating gives you as good a chance at true love as meeting someone at a bar, going on a blind date, or (in my case) attending your first day of high school.