VIDEO: The (Future) Friendship Machine

In the 21st century, we use machines to help us guess the future all the time, from forecasting the weather to diagnosing disease. But what if there was a machine that could predict friendships?

For the past five years, researchers at Columbia University have been slowly chipping away at an answer to precisely that question.

The work began in 2014, when a small team from the university's Interdisciplinary Center for Innovative Theory and Empirics gathered 20 young adults for a study about what happens inside the brain when we meet someone for the first time.

"Can we identify, in the neural response that we have to other people, who we're going to like many, many months later?" said sociologist Peter Bearman, who led the study with his colleague Noam Zerubavel, of the study's goal.

When participants arrived for the study, Bearman had them start by introducing themselves.

"We were all asked to stand up and share why they had decided to participate ... what brought them to that moment," said Kendra Cornejo, who was there that day.

It was basic, getting-to-know-you-type conversation, said another participant, Brian Kundinger. "Kind of the typical, 'Hi, my name is and I'm from wherever.' "

Even though Cornejo and Kundinger had actually gone to the same college, they both said that when they met on that first day, the chemistry they felt as friends wasn't exactly staggering.

"I didn't feel any immediate connection really," said Kundinger.

Instead, they both picked out other people in the group they felt they'd end up befriending. It's a pretty normal experience: You walk into a room and there are people you like and people who leave you cold.

After the introductions, Bearman sat everyone down with a bunch of questions about how they relate to other people.

"So, for example, we asked them for each person in the group: To what extent do you like them? How similar are you to the other person? To what extent does that other person like you? To what extent do you and the other person have shared values?" he said.

Then came the fMRI machine. One by one, participants were inserted into the brain imaging machine.

"We would show them pictures of themselves and everybody in their group 10 times. Over and over again in a random order," said Bearman.

The goal was to capture the neural responses that participants were having toward one another.

Specifically, the researchers looked at an area of the brain called the "valuation region" or the "reward region." It's an area that includes the ventromedial prefrontal cortex and the ventral striatum of the brain. As Bearman explained, "It's that region of the brain that gets activated when people see things that are relevant for goals that they have."

"So, you might show someone an image of a $20 bill and that region may get activated," he said. "[But] nobody ever has studied how these regions relate to our capacity to interact with others in a real-life context."

At the end of the study, the results were both striking and surprising.

According to Bearman, "What we end up seeing in this study is that there is a gap or an asymmetry sometimes between [someone's] neural response to somebody who they see in the scanner and who they say they like. ... It's like looking at two different datasets, like frozen yogurt and fish!"

And as it turns out, when it comes to predicting future friends, our gut instincts are nowhere near as reliable as our brains.

"Our neural response to other people is very strongly predictive of who we're going to like months later," said Bearman.

In other words, the researchers realized they could tell more about friendships that would form within the group by looking at the participants' brains than by asking them directly. One reason this might be, said Bearman, is that our brains may actually have more access to our inner goals than our conscious mind.

"Our brain response to someone is a neural response that we know is activated when things we see are valuable to us ... are relevant for our goals. So, we might not know precisely what are our goals at the moment. We may not have conscious appreciation for [that]."

It's a finding that struck a chord with Kundinger and Cornejo. Even though neither felt great affection for the other initially, their brains saw things differently. Sure enough, after a couple of weeks, they were so enmeshed, they actually had to police their desire to spend time together.

"I remember like wanting to hang out with her in this personal capacity but knowing that we kinda needed to split up for our professional capacity," said Kundinger. Cornejo agreed, remembering that it was a struggle "everyday" to spend time with the other people in the program. Even today, more than five years after that initial meeting, the two still keep in touch.

Now, this is a small study. It's also the first of its kind, which makes it hard to draw broad conclusions. Still, it's a strange thing to think about how one day in the summer of 2014, a machine looked into the brains of people, and in that moment, could see more about the feelings that would flower than the person lying inside it.