Algorithms of Love

Source: pogonici/Shutterstock

Algorithms control much of what we do online. They tell us what to buy on Amazon, where to apply for jobs on LinkedIn, which friends to add on Facebook, and, increasingly, whom to date.

But is an algorithm really capable of judging something as elusive as human compatibility? When bad dates and failed marriages abound, could matching algorithms be the solution?

It can certainly help if you think so. There has been some skepticism in the scientific community about the compatibility matching process in online dating and how different sites make their recommendations. However, one consistent finding across both industry and academic research is that matching algorithms seem to work well when people expect them to.

The OkCupid Experiments

OkCupid co-founder Christian Rudder has spoken extensively about the site’s algorithm and the way it uses match percentages to estimate compatibility. In a now-infamous 2014 blog post (“We Experiment on Human Beings!”) for OkTrends, Rudder outlined the findings from a user experiment that tested the effectiveness of OkCupid’s recommendation system in eliciting increased messaging behavior on the site. He wrote: "In the back of our minds, there’s always been the possibility: Maybe it works just because we tell people it does. Maybe people just like each other because they think they’re supposed to?"

For the experiment, OkCupid told poor matches that they were highly compatible (and vice versa) and then observed what happened next. The data revealed that people were more likely to send an initial message to someone and to keep the conversation going when they thought they were compatible, even if they weren’t. However, it was still better if they really were a good match — that is, when beliefs and reality aligned. In other words, the OkCupid algorithm had an effect, but so did people’s perceptions. This raises an intriguing question about just how important beliefs are to the matching process.

Algorithms and First Date Success

In a new study, I was interested in exploring how people’s perceptions of matching algorithms shape the development of their online dating relationships over time. To do this, I surveyed online daters about their algorithmic beliefs (e.g., “Matching algorithms really work,” “I would trust a matching algorithm to find me a partner”) and followed up with them longitudinally after their first date with a prospective partner. The good news is that online dating produced mostly successful first dates, regardless of whether people used an algorithm or not. But the results also indicated that across sites, the first date was better when people believed in the effectiveness of the matching process.

Of course, none of this is to say that algorithms don’t also deliver high-quality matches. And some algorithms probably work better than others for identifying a mate. But it does point to people’s beliefs about matching algorithms as another potential factor in the success of their relationships.

As Rudder himself went on to note, in the world of digital dating, the best algorithm may still be you.