Love in the Time of Analytics
It’s not for nothing that Harvard Business Review called data science the “sexiest job of the 21st century.” Online dating sites, “hookup” apps, and other data-driven relationship services have uncovered a variety of interesting insights about romance in the last decade. But even with all the technical advances, we still have a long way to go before machines understand our love lives as well as the sensitive operating systems in the 2013 film “Her.”
Consider the world of online dating. The most popular sites, such as OkCupid and eHarmony, have legions of number-crunchers working to find the best algorithms for matching similar users and the surest predictors of relationship success. And the insights these digital love doctors have gleaned are far from trivial; it might not be obvious to someone in a singles bar that older people tend to click with partners who have similar interests, while younger people are more likely to go for partners with whom they share mutual friends. Data also reveals the role seasonal and weekly rhythms play in dating, with new relationships on Facebook swelling on Valentine’s Day, Christmas, and after weekends. And be careful what kind of camera you use to take your profile picture: the make and brand of camera has a profound effect on how OkCupid users rate one another’s attractiveness. So far, though, the hyper-specific categories famous among Netflix users have not made it into online dating; if you’re only interested in tall redheads who make over $100,000 a year and play the trombone, you may have a problem an app can’t fix.
In many cases, online dating sites can help users avoid wasting their time with potential partners who have different values or tastes. But ultimately, whether or not two people will have chemistry still comes down to how well the actual in-person dates go. Once users log off and head to the restaurant or bar to meet their date, all algorithmic bets are off, and people who looked great online can turn out to have misrepresented their real characteristics. For example, even though OkCupid knows that, in aggregate, both male and female users both tend to exaggerate their height, it doesn’t police the liars individually or notify their potential matches. An analyst at Wired even sought to identify the most popular online daters’ traits, so that less popular users might “optimize” their profiles by listing those traits. On the Internet, no one knows you’re a dog, never mind if you’re a boorish, clingy, narcissistic one.
The younger set has embraced more casual dating apps, such as Tinder and Let’s Date, which often incorporate geolocation data to facilitate quicker meetings. Tinder’s data scientists have developed their own set of factors that are most likely to presage a hookup, including the length of its users’ in-app conversations, to help inform future recommendations. But again, these apps work best under typical circumstances, in big cities with large populations interested in casual dating. Gay dating app Grindr illustrates this best. If your nearest prospect is dozens of miles away—in Mississippi, the least gay-curious state in the union, for example—Grindr’s geolocation capabilities can only be so helpful. In these situations, it may be less helpful to recommend similar users than to recommend moving away from Mississippi at the first opportunity.
Soon, data-driven apps may be able to help with romance even after you’ve found a partner. Online dating site HowAboutWe has a Couples product that curates dates for two, and a recent job posting indicates that the company is looking to develop a recommendation algorithm associated with it. Such a product could be useful for couples exploring a new city or just looking for something different on a Saturday night. And since one of the keys to a good recommendation algorithm is the ability to predict what users will want before they even want it, expect a “Sit on the couch and watch TV” date to top the list consistently.
So, while data science has been crucial in creating the romance apps tens of millions of users swear by, not to mention fueling network-based relationship research, it’s helpful to take a step back from the hype. You may be 50 feet from your next hookup, but we’re still a long way away from “Her.”
Photo: Flickr user Shutterbugamar