The conversation you're dreading will go better than you think

There is a specific kind of dread that has nothing to do with public speaking. It's the lift with a colleague you half-know. The queue where someone catches your eye. The party where the person you came with has wandered off. Nothing is at stake, and yet the whole thing feels like a test you'd rather not sit.

What makes this interesting is that the dread is measurable, and so is its accuracy. It turns out we are not slightly wrong about how these conversations will go. We are wrong in a reliable direction, by a large margin, and we stay wrong even after being shown otherwise.

The commuter experiments

In 2014, Nicholas Epley and Juliana Schroeder published a set of studies in the Journal of Experimental Psychology: General with a design simple enough to explain in a sentence. They approached people commuting into Chicago and randomly assigned them to one of three instructions: talk to a stranger on the train, sit alone, or just commute as they normally would.

Then they asked how it went.

The people told to talk to a stranger had a better commute than the people told to sit alone. That's the finding, and it replicated on buses with a separate sample of 75 people.

But the more useful half of the study is what a different group of commuters predicted would happen. Asked in advance, 105 people forecast the opposite: that sitting alone would be the more pleasant option, and that striking up a conversation would cost them both enjoyment and productivity. Neither cost showed up. The conversation group lost no productivity at all.

So people aren't just mildly uncertain about how these interactions will go. They confidently predict the wrong outcome.

The part that matters more: you also misjudge how it went

If the first finding is about forecasting, the second is about memory — and it's the one that keeps people stuck.

In 2018, Erica Boothby, Gus Cooney, Gillian Sandstrom and Margaret Clark published work in Psychological Science on what they named the liking gap: after a conversation, people consistently believe their partner liked them less than that partner actually reported liking them.

The mechanism is the interesting bit. In one study the researchers looked at the thoughts people had after the conversation and found that the more negative someone's own post-conversation thoughts were, the larger their liking gap. People's internal commentary about themselves was reliably harsher than their thoughts about their partner — and that self-criticism crowded out the actual evidence sitting right in front of them, in the form of a person who had been visibly enjoying the conversation.

And it doesn't evaporate after one awkward introduction. In a study that followed university dormmates across an academic year, measuring at five points from September to May, the liking gap persisted at nearly every stage. People spent most of a year systematically underestimating how much the people they lived with liked them.

What to actually do with this

The unhelpful version of this advice is "so don't be nervous." Nobody has ever talked themselves out of nerves by being told the nerves are statistically unwarranted.

The useful version is narrower. These findings identify two specific moments where your judgement is unreliable, which means those are the two moments to stop trusting it.

• Before: your forecast is not information. The feeling that this will be awkward is not data about the conversation. It's a prediction with a documented bias, and the bias has a direction. When you notice it, you're allowed to treat it as noise and go anyway.
• After: your replay is not a verdict. The post-mortem where you catalogue everything you said badly is exactly the process the liking-gap research found causing the misjudgement. The more thorough your replay, the more wrong you're likely to end up.

That second one is worth sitting with, because it inverts how most people think about it. Replaying a conversation feels like diligence — like you're learning something. The evidence suggests that past a very short window it does the opposite: it manufactures a negative account of an event that other people remember perfectly fondly.

The ten-minute rule

A concrete version: after a conversation you're unhappy with, give yourself ten minutes to think about it. Set an actual timer. When it goes, write down one line about what you'd do differently, and stop.

Ten minutes is enough to extract anything genuinely useful. Everything past it is the process the research describes — building a case against yourself out of evidence you're not qualified to assess, about a person whose actual opinion you don't have access to.

The uncomfortable implication

If your read on your own conversations is unreliable, then so is your sense of whether you're improving. You could get meaningfully better at this and not notice, because the thing doing the noticing is the same faculty that's already been shown to be miscalibrated.

Which is the argument for measuring something. Not because a number tells you whether someone enjoyed talking to you — it emphatically doesn't — but because how fast you spoke, how often you hedged, and how many times you said "um" are facts that don't care how the conversation felt to you afterwards.