You don’t want Big Data to pick your Valentine

We may earn a commission from links on this page.

The dating sites in which so many of us invest our hopes of finding love are based on algorithms. These algorithms promise to find us long-term relationships with partners that match us perfectly. But their process is fundamentally flawed from the outset.

As I explain in my new book, Data, A Love Story: How I Gamed Online Dating To Meet My Match, the process of creating a successful dating site happens in many steps. Developing a set of algorithms is the start. But equally important is the data itself. It turns out that the design of a dating website and how it manages data collection is significantly more important than the algorithms alone in determining successful matches.

Dating sites require a steady stream of user data in order to function. They’re hungry beasts that need constant feeding. How we enter our information and create our profiles is what differentiates each one of the dating services.

To start, dating sites ask users the wrong questions. What’s your favorite book? How do you practice your faith? Are you a Republican? These aren’t good data points for matching algorithms, because most of us answer the questions on dating sites aspirationally rather than honestly. We think about idealized versions of ourselves and paint a skewed profile, often not on purpose, but because the signup process on dating sites is designed to make us feel great about ourselves. After all, if we don’t enjoy the experience of entering our own user data, then the system will have less information to parse and ultimately too little content to push through its algorithms.

But if we want a stable, happy long-term relationship, we can’t answer questions as the people we hope to be five years from now, but instead must answer them as the people we are right now, regardless of how overweight / flat chested / not well traveled / whatever we are in the present.

Think about the way you’ve set up your Facebook profile. And if you don’t use Facebook, instead think about how you’ve described yourself to new people you’ve met recently. You list your favorite foods, bands, books. You talk about cities you want to visit. These aren’t meaningful data points; they’re stylized nuggets of information meant to personify ourselves in a formulaic way to others. A Facebook profile is in many ways an outfit we wear and the accessories and cologne we put with it: We’re hoping to project a particular image in order to socialize with (or avoid, in some cases) a particular group of people.

Dating sites and the algorithms they advertise purport to sort through our personalities, wants, and desires in order to connect us with our best possible matches. Which means that we’ve outsourced not just an introduction, but the consideration of whether or not that man or woman is really our ideal. We’re putting our blind trust in a system that’s meant to do the heavy lifting of figuring out what it is that we really want out of a mate, and what will truly make us happy.

And this job is being processed using the information that we, ourselves, have entered into a computer system. Bad data in equals bad data out. Algorithms that dating sites have spent millions of dollars to refine aren’t necessarily bad. Many of them work exactly as they were intended. It’s just that the algorithms can’t possibly be as good as we want them to be, because they’re computing our half-truths and aspirational wishes.

A smarter approach would be to force users to create a big list, as I show in the first part of my book. Ask daters what, exactly, they want in a mate. Create a user experience where they will unabashedly enter all of the various granular qualities they need to be happy; then, a dating site should parse that list both with machine learning and human input. As I argue, great matching can’t happen via code alone within the limitations of our current digital ecosystem.

On the other hand, it’s possible that creating perfect algorithms and ideal matches isn’t actually the goal. Think about how these sites earn their revenue. OkCupid uses an advertising model, so the more times you visit the site and click around, the more money is generated. JDate,, and eHarmony rely solely on subscriptions to make money, and they all offer three, six, and even twelve-month memberships, with a hefty discount for paying in advance.

The profit model for dating sites relies on retention, even though our desire as members is exactly the opposite. We want to find true love so we can be finished with dating altogether. Once we’re in relationships, we’re theoretically off the market. We cancel our memberships and spend our money elsewhere. Fresh new crops of daters should cycle through the various dating sites just as current members are leaving, but that’s not always the case. And besides, the more members or page views a site can count, the more money there is to be made. Algorithms may try to connect people on a maximum number of compatible data points; but bean counters are likely tasked with ensuring that we go on good dates—just not great ones.