Jacked up dollar bills

by Collin Lysford

From a friend of mine who works at a hot dog stand:

I don’t mind at all if you pay with jacked up $1 bills. Don’t bother straightening them up at all. I set those aside and give them as change to customers who don’t look likely to tip and they quite often toss them back right away in my tip jar. At which point I fish them out again. Profit, repeat.

A debate is raging on between Scott Alexander and a literal banana ¹ about “automaticity”: how much you can understand human behavior through rules that are unconscious to the people following them. Which way does this anecdote point? Is this evidence for or against automaticity? If you asked someone just leaving the hot dog stand “Why did you tip there?” would they implicate the jacked up $1 as part of their reasoning? Or would they answer with something broader about tipping: “Hey, I like my local hot dog vendor, I can spare it, why wouldn’t I?”

What’s interesting about this example is that taking it step by step, it’s all instantly legible without needing any psychological leaps of faith. Who doesn’t prefer a crisp $1 in their wallet vs a crumpled up one? Who’s surprised to learn that the contents of the tip jar aren’t an entirely honest record of customer contributions? Okay, maybe if you haven’t worked service industry, you can’t recognize “those who don’t look likely to tip”, but they’re out there.

But does that mean you ’re considering the salience of these factors? We all recognize when a bill is crumpled to hell, but most people don’t walk around with a mental index of the crumple factor of each of their bills. The factors that seem to be relevant to whether you’ll tip are the appearance or taste of the hot dog, the conviviality of the hot dog man, and the specific amount of money in your wallet. Dollars are fungible in most of the ways you interact with them, so it’s easy to elide over the difference “good dollars” and “bad dollars” have on your decision making, even if you recognize it when you’re directly asked to compare them.

Tipping isn’t fully adversarial but what springs to me here is the distinction between information and state in anti-inductive games: two people may be looking at the same thing, but not resolving it the same way. Perhaps automaticity is relevant not as a bunch of secret triggers that require minute experimentation to tease apart, but as factors we’ll notice when pointed to but often don’t consider “in frame” for a particular interaction. You see it as a detail, but do you see it as a chunk?