My heuristics for interacting with humans

On game trees and reasonableness

I don’t sense that I’m viewed as particularly skilled at human interaction. Still, some poor fools sometimes ask me for advice, and I find myself repeating the same little speech.

For context, in life as a human you will often face difficulties with the other humans. Like:

• Alice keeps stealing all the credit at work. She never lies, exactly. But somehow she always manages to be the one sending the final email to the CEO and “forgets” to invite you to important meetings. You grind your teeth when you hear people complimenting her for your work. But if you talk to her maybe she’ll just brush you off? And if you tell other people “No, I did that!” maybe you’ll just look petty?

• You’re about to get married. You have an uncle that lives far away and complains of ambiguous health issues. Before picking a date, you call him, but he only says, “Oh ho ho, don’t worry about me!” But you sense ill-intent. He’s waiting for you to choose a date so he can say it doesn’t work for him and badmouth you to the rest of your family.

• You submit a paper to a journal. Reviewer #1 is positive, Reviewer #2 is confused, and Reviewer #3 is hysterical that you didn’t mention the seminal works of Lai (1997) and Lai and Cummings (1998) and Lai (2002) and Lai et al. (2004). These are all irrelevant, but if you say that, maybe you’ll anger Reviewer #3, whoever that could possibly be? But if you don’t say it, then maybe the other reviewers will think the complaint is valid?

• You write things on the internet. Every time you do, a few people on forums say, “trololololol, stopped reading after [misinterpretation of irrelevant point]”. Slowly, these people seem to burrow into your mind. As you write, you find yourself defensively adding clarifications everywhere. But this makes your writing long and confusing and boring.

What should you do?

To win at a game like chess, you need to anticipate all possible reactions of your opponent: “If I do this, then they will do that, after which I can do this.” It’s helpful to think in terms of a “game tree”. This is a graph of all the board states connected with arrows for each possible move. Here’s what this looks like for tic-tac toe:

To play such a game well, you need to do “game tree search”: Work backwards from the end and find a branching strategy that guarantees you always win.

That’s how you play tic-tac-toe. But go apply this thinking to some social dilemma. Usually, you’ll find it doesn’t seem to work. No matter what you do, other people could respond in many ways, and then other people will respond to them. It’s all so uncertain!

While we may not think of it as “game tree search” we often get trapped in similar thought patterns in difficult social situations. We try to figure out how to control the outcome and get demoralized when that seems impossible.

That brings us to the heuristics:

1. If in doubt, act assuming that other people won’t be unreasonable.

2. Avoid applying game tree search to social dynamics.

Here’s how you apply these: If you’re tortured that social dynamics are so complex, then take your game tree and delete all the edges where people do unreasonable things. Usually, this makes your best action pretty obvious. In fact, usually it renders the whole game-tree framing unnecessary. Take that action as your “null hypothesis”. Deviate if you must! But only with a good reason. And if you deviate a lot, that’s a warning sign.

Why these heuristics?

Unreasonable reactions are arguably not your problem

If you’re interviewing someone for a job and they have something stuck in their teeth, should you worry that mentioning it might enrage them or give them a panic attack?

If that’s too deontological, rephrase it this way: Worrying about unreasonable reactions will sabotage your interactions with everyone else. Think of everyone who wanted to know about the stuff in their teeth.

More generally, if you assume everyone else is a sneaky snake, then you might come off that way to everyone else. Ultimately, people are smart and they will figure you out. (If there’s one thing human intelligence is optimized for, it’s that.)

The branching factor is too damn high

Unreasonable people are unpredictable. This means that you’re usually better off “betting on reasonableness”. If you’re right, great. If you’re wrong, the situation is often nigh-on-uncontrollable anyway.

Consider internet writing. Most people read something in the hopes of getting some kind of value. They want to like it. Yes, a small minority can get triggered by weird things and another small minority just enjoys picking things apart. But there are an infinite number of things for these people to get upset about. If you try to satisfy everyone, you’ll add lots of unnecessary clarifications that will bore and annoy the people who want to play along. (You’ll also drive yourself insane.)

It’s good to cultivate this way of being

If you’re “scheming” at work, it might be hard to drop the act at home.

More generally, in life you probably want to surround yourself with people who are reasonable. By assuming that as a default, you avoid getting stuck in a mutual assumption of bad faith with other reasonable people. And for unreasonable people, you more quickly discover who they are. (Enabling you to take evasive action.)

Remember Alice stealing your credit at work. What would these heuristic say to do? Well, what are your options?

If you do nothing, nothing will happen.

If you ask her to let you send the emails, and she’s reasonable, she’ll agree. If she’s unreasonable, maybe she’ll do it anyway. Or maybe she’ll tell everyone you’re a crybaby.

If you try to outmaneuver her and she’s reasonable, it might work, or might not. Either way, she’ll probably notice what you’re doing and stop trusting you. If she’s unreasonable, who knows?

If you complain to your boss, maybe your boss will try to force Alice to do what you want. Maybe that will work, or maybe it won’t. Or maybe your boss won’t do anything. Or maybe your boss will be annoyed you didn’t try to resolve the problem yourself first.

In the world where Alice is reasonable, the clear winner is talking to her. If she’s unreasonable, it probably won’t work. But in that world, nothing is guaranteed to work.

Note something else here: If you look at the plausible outcomes when you assume other people are reasonable, they are usually… not that bad? In this case, what would that be? Maybe Alice is mean to you? Maybe she talks about you behind your back? That’s… not that bad?

Of course, all of this is just a heuristic. If you fly into Singapore and the immigration officer smells cannabis on your breath, do not assume they will be reasonable! If you’ve interacted with Alice many times and she’s proven to be a sneaky snake who can’t be trusted not to do insane things, then you have no choice but to use that information.

But in general, if you find yourself playing out lots of game trees of social interactions in your head, you should probably see that as a warning sign that either you’re underestimating people, you’re trying too hard to make everyone happy all the time, or you might want to switch your social context, lest you become a sneaky snake yourself.