August 15, 2022

On Noble Lies

Original Twitter thread

Many institutions have been accused of playing 4D chess with public health advice around covid. Usually people point out that this is a bad thing because if you do it badly, it erodes trust in institutions - I agree with this

But I have been plagued by a nagging thought: What if institutions weren’t so bad at it though? Isn’t it possible that in theory, Noble Lies could lead to better outcomes? Can they therefore be justified?”

In the set of possible things an institution can say, there is: A (the truth), B (a partial truth), C (another partial truth), D (an outright lie), and so on. Of all of these, why should we assume A will generally lead to the best outcome?

Sure, if the pubic is reliably very smart and reasonable, we would expect that giving them accurate information would lead to better outcomes (setting aside weird game theory dynamics like hoarding goods). But most people don’t think of the public as reliably smart and reasonable.

But clearly institutions are bad at guessing how people will behave, and bad at 4D chess. It does seem like whenever they attempt this, it would have been better if they’d just been honest.

It feels like a Noble Lie involves messing with a complex, delicate system and if you try to interfere with one part it has all these second order effects and unintended consequences (kind of like messing with prices in a market).

If we assume that we’re very bad at predicting the ways in which the public will react unreasonably to our message, then we might as well model their unreasonableness purely as a noise term we have no control over, then the reasonable part of their reaction is the only part we can know how we’re affecting.

Given this model of public reaction as reasonableness + noise, accurate information will overall lead to better outcomes.

So while in theory, a well-chosen Noble Lie could have a better outcome, the chance that you will choose one that outperforms the truth out of all the options is very slim, so you would be much better off telling the truth and taking advantage of the small amount of reasonableness” in the system that tends towards better decisions.

In sum, Noble Lies are bad because:

  1. It’s almost impossible to predict which lie will work, whereas if you tell the truth you have the reasonableness of the system working in your favor
  2. In the very likely case a Noble Lie doesn’t work out, you damage institutional trust.