The researchers brought them in expecting a fight.

Two groups of people, opposite sides of a topic so polarizing I don’t need to name it for you to picture the room. Each side walked in believing the other side was the enemy. Each side believed they had nothing in common with the people across the table. Each side believed the conversation would go nowhere.

Then the researchers did something most of us never do. They trained the participants. Not in debate. In listening. In how to make the person across from you feel heard — not agreed with, not validated, heard. Then they paired them off.

What happened in those rooms wasn’t agreement. It was something rarer. One of the participants said it was the most understood he had felt in any conversation in his adult life. The person who made him feel that way held beliefs and values that were the opposite of his own.

That’s the experiment. That’s the result.

Then the researchers tried to scale it.

They built a Facebook group. They invited the participants who had been through the training. And they invited a hundred more people who hadn’t.

It took 45 minutes for the whole thing to fall apart.

Not because the trained participants forgot what they had learned. Because a hundred untrained voices walked in with their default attack methodology, and the medium itself rewarded them for it. You can’t make someone feel heard through a comment thread when the people next to you are sharpening knives. The training couldn’t survive contact with the platform.

I read this in Charles Duhigg’s Supercommunicators this week and I haven’t stopped thinking about it.

Here is what the experiment proved that the researchers didn’t set out to prove. The reciprocity you get from listening to another human being is not available through a keyboard. Not because keyboards are evil. Because writing is a high-skill medium, and most of us are not skilled at it.

Most people are not good at writing. I don’t mean that as an insult. I mean it as a fact about the moment we live in. Writing carries no face. No tone. No pause. No softening of the eyes when something hard needs to be said. The whole apparatus humans evolved to communicate meaning is stripped away, and what’s left has to do all the work alone. That’s a hard medium even for skilled writers. For most people, it’s an impossible one.

What’s left is venting. You sit behind a screen, you type how you feel, you hit post, and for a moment you feel better because you got it out. Then it lands in front of someone who can’t see your face, can’t hear your tone, can’t extend you the benefit of the doubt because the medium has trained them not to. They read venom you didn’t mean to send. They send venom back. The cycle compounds.

Both sides do this. I am not picking a team.

We are alike on most things. Family. Food. The basic dignity of being known. Where we differ is sometimes a question of method — how to organize, how to solve — and sometimes a question of values, of who gets protected and who pays the cost. The deeper disagreements are real and they don’t dissolve just because two people sat across from each other. But the experiment showed something the participants didn’t expect either: even when the values were opposite, the experience of being heard didn’t require agreement. It only required the room.

The Facebook group failed in 45 minutes because the people who hadn’t been through the training brought the old habits in with them, and writing — for most of us — can’t do the work that face-to-face does.

So we have to go back.

Not as a preference. As a necessity. The technology is going to keep accelerating. The amount of text in the world is going to keep multiplying past anyone’s ability to read it carefully. If we don’t relearn how to sit across from each other and listen, the fracture that social media started will widen until there’s nothing left to bridge.

This doesn’t get fixed by politicians. It doesn’t get fixed by platforms. It gets fixed by two people, in a room, doing the work the experiment proved was possible — and the Facebook group proved was fragile.

One conversation. Then another.