Do dating apps measure love — or create it?
And if the latter, what would that mean for *hiring* apps?
In early 2013, my roommate Matt and I had just arrived in Washington, DC. Along with many of our neighbors, we started using a newly launched app called Hinge, which analyzed our respective networks of Facebook friends. The app used a simple algorithm, introducing strangers who had a friend in common. It was a great idea. Other dating apps, then and now, emphasized “swiping” through lots of pictures and seemed designed to set users up for one night stands, but Hinge showed different priorities, offering a few matches each day and encouraging conversation. (Today it calls itself “the dating app designed to be deleted.”)
As it happened, one of these early Hinge users was named Anna, and a couple of her friends from college had later gone to law school with Matt and I. Matt was her first Hinge date, in June, and by that fall, they were engaged — the first such couple to meet through the app. A year after they met, I found myself standing on their apartment roof — empowered, as a “civil celebrant,” to pronounce them man and wife.
The algorithm that chose to introduce Matt and Anna — rather than any of the hundreds of other friends-of-friends it might have shown to either of them — was an adaptation of old human wisdom. Meeting through mutual friends has a long and largely low-tech history. In this case, they’d spent time getting different degrees in the same college town, and shared not one but two Facebook friends, and they seemed to have many of the same preferences, all of which the algorithm may have calculated. In human terms, out of all the possible bridges that could be drawn across the social graph, only some of the possibly suitable matches will come to mind, and only some of the human links between possibly well-suited strangers will take the initiative to suggest an introduction. Likewise, this algorithm had many options, and for whatever reason, put these two in touch.
Was Hinge’s prediction of a successful match really true before it was made? Or did the robotic reassurance of a computer’s recommendation help motivate the two of them to get to know each other, and thus sort of make itself true?
Interestingly, the dating site OKCupid did test this idea: One of its founders described an experiment in which the company “took pairs of bad matches . . . and told them they were exceptionally good for each other.” They found that “[w]hen we tell people they are a good match, they act as if they are. Even when they should be wrong for each other.”
For Matt and Anna, this question may be moot. But in the larger scheme of things, it’s huge. If we think about tomorrow’s dating apps, or tomorrow’s job applications, tomorrow’s courtroom risk scores or tomorrow’s teacher performance scores, then it does matter.
Was Hinge only measuring something about Matt and Anna? Or was it actually creating something? It’s a question that comes up whenever software describes a human being in some way that’s meant to change what happens.
What these algorithms are really doing depends, I think, on a deep question about human nature, a question scientists, philosophers and many others have argued over for centuries.
One perspective about people is that our actions and desires tend to reflect deep inner truths about who we really are. For instance, two people who date are either right for each other, or they’re not. If they date each other with an eye to settling down, they’re basically both trying to figure out what’s always been true, about whether or not they make a good couple. Criminals are dangerous people — people disposed to be dangerous. Students succeed or fail based largely on innate intelligence, a factor we flatter ourselves to think we can change. And the job candidate who makes the manager is the one who brings the right personality into the job, who has always been right for the job.
If that’s how people are, then people-judging algorithms are basically measurement tools. They detect who is compatible with whom, who’s dangerous, or who’s trustworthy.
This inner-essence perspective on human identity is very popular. A mountain of social science evidence demonstrates that people tend to believe in it. And we have for a long time: The western idea of an immortal soul, for instance, fits right in to this school of thought.
Social psychologists have probed this in the lab.1 For instance, in one experiment students were assigned to write persuasive essays for assigned points of view — in favor of the Communist regime in Cuba, say, or against legalizing marijuana. Other students, reading these essays, were told the true set-up (that is, they were told that each author had been assigned what position to argue for). But the readers still reported a belief that the essays reflected the true dispositions of the authors, rather than the orders the authors had been given.2 In another example, student “volunteers” who were paid for their service signed up for a service activity more often than those who were not offered any payment. But even after knowing about this compensation, the students’ peers still judged that these well-compensated volunteers were more innately generous, and would be more likely to volunteer in a future, unpaid setting.3 My favorite study of this type comes from a simulated office where people were assigned either to management or menial jobs — via a process that all could see was random, and left in those jobs for a couple of hours.4 At the end, participants were asked to rate each other on a variety of personal qualities. Participants still ended up thinking that the “managers” were better leaders, smarter, more assertive, and more supportive of their coworkers. The “clerks,” on the other hand, were rated as more hardworking than the managers.
Maybe it’s just me, but at a human level, these findings ring true.
If I’m standing in line at the grocery store, and the person in front of me snaps at the clerk, my first thought is likely to be, “what a jerk!” — as though the person in front of me were somehow disposed, in general, to treat grocery store clerks badly. But if I’m the one who snaps at the clerk, I’m apt to have a circumstantial explanation. I was hungry and cranky. I hadn’t slept. Surely, in any case, my momentary anger wasn’t a revealing window onto the kind of person that I am.
That’s the pattern: my own behavior may be explained by circumstances, but others — especially strangers — act in ways that reveal who they really are. It’s like there is an interpretive undertow, pulling our attention away from the circumstances that surround a person’s actions and zooming in instead on their innermost nature. That, we seem naturally disposed to think, is where true understanding really lives.
But there has also been, for thousands of years, a dissenting view about human nature and human life — and if the dissenters are right, then the role of person-judging algorithms in our lives is very different from what it seems.
Nest week, we’ll meet some of those dissenters.
See Ross and Nisbett’s Person and the Situation, a classic summary and synthesis of this work.
Jones & Harris, The Attribution of Attitudes (1967)
Nisbett et al., Behavior as Seen by the Actor and as Seen by the Observer (1973)
Ronald Humphrey, How Work Roles Influence Perception: Structural-Cognitive Processes and Organizational Behavior (1985)