Twice A Stranger, Once A Friend

YYou meet someone at a party where you know almost nobody. For the first ten minutes you consider leaving, then somebody asks you a question, which becomes another question, and forty minutes later you're standing beside a bowl of melting ice talking to a woman named Nina about the summer she worked at a funeral home.

On the way home, you think: That was actually fun.

This should count for something. You were reluctant to go, expected the conversations to be awkward, went anyway and acquired evidence that your prediction was wrong. Next time, presumably, your brain can consult the record.

Researchers recently tested something close to this assumption and found that the record has an oddly short shelf life.

Across three longitudinal experiments, Stav Atir and Nicholas Epley found that people tended to expect conversations with strangers to go worse than they actually did. Immediately after having one, though, participants learned from the experience: their expectations about another conversation became more positive and better matched what had just happened. Then the researchers waited.

Within one or two weeks, much of the correction was gone.

Participants who had actually talked with someone were again approaching another conversation with expectations similar to people who hadn't received that earlier evidence at all. In one experiment, this happened even when they knew the next conversation would be with the same person. The first conversation had already occurred. It had gone better than expected. For a while this information mattered.

Then apparently Nina's clearance expired.

We have been over this

There is something especially irritating about a fear that survives being wrong.

Most of us have at least one small version of this. You dread the party and enjoy the party. You don't want to go to dinner because you won't know anyone, then spend two hours talking to someone you like. You worry that calling an old friend after too much time has passed will be awkward, finally call, and within three minutes the gap has disappeared.

Afterward, the lesson seems obvious: I always think this will be worse than it is.

That sentence should be useful because it contains both the prediction and the correction. You have caught yourself making the error. Yet somehow, two weeks later, another invitation arrives and the dread presents itself with all the authority of new information.

The experiments suggest this isn't simply a failure to notice that the conversation went well. Participants noticed. Immediately afterward, their expectations changed in the sensible direction. The strange part is that the change didn't hold.

We usually imagine learning as cumulative, especially when the lesson concerns our own experience. Touch the hot stove, remember the stove. Take the shortcut, discover it isn't shorter, stop taking it. But social expectations may be capable of absorbing contradictory evidence without permanently surrendering the original forecast. The experience wins the argument on Tuesday and, by next Friday, the argument has somehow been reopened.

It was nice, probably

Memory appears to be part of the problem.

In one experiment, researchers asked participants two weeks later to remember how the earlier conversation had felt. By then, they remembered it as less positive than they had reported immediately after actually having it. The deterioration in memory partly explained why their expectations had become pessimistic again, although it didn't explain the whole effect.

This is a particularly efficient arrangement if your goal is to remain nervous around strangers forever. First, underestimate how pleasant the conversation will be. Then have a pleasant conversation. Correct the underestimate. Wait fourteen days. Downgrade the evidence.

Nothing has to be forgotten completely. The conversation can remain in memory as perfectly fine while losing the force it briefly had as an argument against your expectations. That went better than I thought becomes that was nice, which becomes one pleasant evening among many things you know about yourself, including the apparently indestructible fact that you don't really like talking to people you don't know.

The experience survives. Its jurisdiction shrinks.

This may be why social confidence can feel strangely non-cumulative. You can possess years of evidence that people are generally nicer than you expect, that parties usually become tolerable once you're inside them, that asking the question is easier than standing nearby considering whether to ask it, and still encounter each new situation as though the previous cases were heard in another court.

Same person, new emergency

The experiment involving repeat partners makes this harder to explain away. Immediately after a conversation, people were somewhat more optimistic about talking again with the same person than with someone new, which makes sense: perhaps the first conversation taught them something narrow rather than something about strangers in general.

But when the second conversation was separated from the first by a week, expectations became more pessimistic again. Even knowing who the person was did not fully protect the lesson.

This feels familiar in a way that extends beyond strangers. There are people you've seen twenty times and still briefly worry you won't have anything to talk about before seeing them for the twenty-first. There are friends you routinely enjoy but occasionally dread making plans with because leaving the house has somehow become confused with the experience you'll have once you've left it. There are gatherings whose anticipated awkwardness keeps regenerating despite a suspicious shortage of actual disasters.

The research here is specifically about conversational expectations, so it can't establish that every recurring social reluctance works this way. But it does expose an odd possibility inside a narrow case: experience may be perfectly capable of correcting a belief without replacing it.

For a while, you know better. Then you know the old thing again.

Tuesday's evidence

There is an appealing story we tell about courage in which each successful experience makes the next one easier. You do the uncomfortable thing, discover you survived, and gradually become a person who no longer needs courage for it.

Sometimes that happens. But it may depend on what your mind does with the evidence afterward.

The stranger-conversation experiments suggest that at least some social learning needs maintenance. The useful experience isn't necessarily installed permanently just because it happened. Its emotional accuracy can fade; the prediction it briefly corrected can recover; eventually you may find yourself standing outside another room full of people, equipped with considerable evidence and behaving as though the case remains undecided.

Source

Atir, S., & Epley, N. (2026). “Fleeting Generalization: How Unstable Belief Updating Keeps People Overly Pessimistic About Talking to Strangers.” Journal of Personality and Social Psychology, 131(3), 422–443. https://doi.org/10.1037/pspa0000483