Bad Social Equilibria

2022-03-14 6 min read Wiresong

Table of Contents

Imperfect information, distrust, and unrealized gains from trade


You’re a rational college student, pursuing a degree in hypothetical scenarios.

You’re a bit of a socialite and enjoy going out with your friends every evening. Because you’re the socialite, though, it usually falls upon you to organize your ragtag group of friends and decide on where to go and what to do.

Ten weekends in, you decide to go out yet again, and start figuring out all the logistical details. Opening your messaging app of choice, though, and looking at the 150 “Hey 😊 you wanna go out? <3” messages you’ve sent in the last ten weeks, you realize the following: no one else cares about going out, it’s just you who always takes the initiative. If your friends actually cared about going out with you, they would have chosen to message you at least once, but they didn’t, so they don’t care.

Feeling hurt by the fact that your friends don’t want to hang out with you (for they only care when you organize something), you stop organizing things.

Because all of your fellow socialites are also rational, they all eventually come to the same conclusion, and no one from your college goes out. In the farewell party, Everyone wonders why this is the first time they’ve seen each other outside classes.


You have a crush on someone, and now you must engage in the age-old tradition of expressing your love by strategically-placed hints, hoping your partner will offer some strategically-placed hints of their own.

Things are not going so well, though. She always seems to be busy; you only catch up with her in lunchbreaks and such, but otherwise you don’t really talk. You’ve tried to make her realize how you feel, by revealing things that you’ve never told anyone else. She never seems to get the hint though.

One month in, and everything is the same. At this point, you sadly realize that she just isn’t interested in you. Slowly you begin to withdraw; spending lunchbreaks together every day turns into spending lunchbreaks every other day, then once a week. Even when you spend time together, you talk about nothing of substance; conversations are light and superficial. After another month, you barely talk at all.

It’s strange though. Whenever you run into her, and your eyes meet, she seems to be somewhat sad and melancholy, just like you…


It is exam season, and things aren’t going well. Your other responsibilities have made you leave your studies by the wayside, and now you must engage in yet another age-old tradition of memorizing as much as possible as fast as possible.

this, understandably, is very stressful. Generally, whenever you’re stressed, you like to talk to your friends; this typically works.

However, this time, you reason as follows: my friends, having similar responsibilities as me, must also be trying to study in the same circumstances as I am. If I talk to them, I would just be another source of stress. It’s better to study by myself, rather than bothering my friends.

You and all your friends do this, and no one talks to each other. In the exams, you and your friends do only mildly well. Someone wonders why they didn’t make some sort of study group, but no one has an answer.

the Problem

Of course, the above three cases aren’t completely realistic, at least in the way that they’re portrayed. However, they’re not completely unfamiliar either, and it’s worth thinking about why they occur.

The fundamental problem in all of these scenarios is that, to measure an underlying unobservable value, some sort of observable alternative is used; then, forgetting that the observable alternative is error-prone, it is taken as raw truth, and decisions are made based on it.

For instance, in the first case, caring is measured by willingness-to-socialize (whereas all your friends might be introverts); in the second case, ‘being liked’ is characterized by time spent together (whereas from her perspective, she might only have lunchtime free, which she makes sure to spend with you; also see The Five Love Languages); in the third case, willingness-to-talk is characterized by the amount of work that the others have (whereas people might in fact get satisfaction from helping, or just talking in general).

Phrased somewhat differently, the problem here is that there are ‘unrealized gains from trade’: if the socialite could know the actual value of how much his/her friends care, he/she would be willing to socialize more; both the people in the second example would be in a relationship, if both actually knew how they feel about each other; everyone in the third example would benefit from knowing the actual willingness-to-talk that everyone has. (they might, finally, even be able to get a study group.)

this has already been expressed in different ways before (see Fundamental Attribution Error, people in general not being very agentic), but really all those things seem to be reifications of what I talked about before: an error-filled signal is taken as a true signal, and everyone ignores the error.

The obvious solution (and why it doesn’t work)

The obvious solution to this, of course, is people talk honestly and openly to each other. Everyone expresses their actual preferences, and people take actions based on accurate information thus expressed.

this never happens in practice.

first off, using error signals, in the manner that we’ve seen in the examples above, only really happens when there is some amount of distrust. This is especially prevalent in the grey area when people almost-but-not-quite know each other, as in the second example.

But even when there is some underlying trust, I suspect our social structure pushes us to not actually talk. I’m still thinking through this, more on it in another post (but as a sneak peak, there seem to be two extreme cases in terms of the social spectrum: ‘politeness’, where we are almost forced to send error-filled signals, and being a Utility Monster, where it’s valuable to be as histrionic as possible (see ‘stage 3’ in this blog post)).


It’s tempting to model peoples’ internal thought processes by some sort of revealed preference argument: ‘if someone really wanted to do x, they would do x. If they’re not doing x, then it means they must not want to do x’. For many reasons, though, especially when looking at nebulous things like emotions, this model doesn’t seem to hold up in practice.

The other potential way to model peoples’ internal thought processes is to straight up ask them. But due to distrust, or because of social norms, it’s hard to be confident that what they externally express is what they internally feel.

This leads to ‘bad social equilibria’: a state where there could have been potential gains from trade, but either because of misinformation or mistrust, no one is willing to engage in the trade in question (even though, if they did, they would have benefitted).

The only reason that this isn’t widely prevalent is because of two reasons: first, for people who trust each other sufficiently, it’s easier to believe the information elicited from each other; second, because everyone isn’t rational, and people are willing to act even with imperfect information.