How do you get past small talk?

Almost everyone has had the experience of a conversation that stayed on the surface for an hour when both people would rather it hadn't. It turns out that is not an accident of manners. It is a prediction error, and it has been measured about as thoroughly as anything in this guide.

The finding

Researchers ran a long series of experiments in which strangers were given either shallow questions to discuss — the weather, television, where they were from — or deep ones, along the lines of "what is one of the more embarrassing moments in your life?"

Before talking, people expected the deep conversations to be awkward. Afterwards, they were consistently less awkward than expected, and left people feeling more connected and happier than they had predicted. When people had both kinds of conversation, they reported feeling more connected to the person they had gone deep with than to the person they had made small talk with.

And when people were allowed to choose their own topics, they picked the shallow ones. They chose the conversation they enjoyed less, having mispredicted which was which.

Why the error happens

The explanation the researchers landed on is specific, and it's the part worth carrying around: people underestimate how much a stranger will care about what they say.

Telling somebody something real is only awkward if they are indifferent to it. A shallow conversation is safe precisely because nothing is riding on whether they're interested. So the calculation people run before opening up — they won't care, and then I'll have said it anyway — is reasonable given the input. The input is just wrong. Across these studies, listeners cared considerably more than speakers assumed they would.

That is why the barrier holds even among people who badly want the deeper version. It isn't reticence or lack of skill. It's a forecast about somebody else's interest, made privately, and never checked.

Doing it without it being strange

"Go deeper" is not an instruction anyone can act on, so more concretely — though this is the point where we stop reporting and start suggesting. The studies assigned their questions from a list; they tested nothing about how to raise one yourself.

What this doesn't claim

Not that deep conversations are always better. The finding is about the gap between what people expect and what happens, and some of these conversations were middling — the average moved, not every case.

Not that you should be doing this with everyone. The studies used strangers who had agreed to be in a study and would never see each other again, which is a specific and unusually safe setting. It resembles anonymous chat more than most research does, which is the reason this page exists — but "resembles" is doing real work in that sentence, because the participants were sitting in front of each other and had given their names to a university.

And on an anonymous site, there is one direction of depth that is not like the others: personal history is fine, identifying detail is not. Where you grew up, what you do, what your face looks like — those are not depth, they are the thing anonymity is protecting, and the conversation does not need them to get somewhere real.