Lately I’ve been trying to use Bayes’ Theorem in daily life — quick guesses, like someone’s nationality from a glance.
What I’ve noticed: my intuition does better when I don’t adjust for general priors. Corrections like “most people in Germany aren’t Russian” when someone looks vaguely Slavic often pull me further from the truth.
After five minutes of reflection, my best guess: explicit Bayes only really helps out-of-distribution, when we lack feedback loops — new domains, big decisions, reasoning about AI. That’s when 5 minutes of googling or reading a paper can give you better intuition than your System 1.
Lately I’ve been trying to use Bayes’ Theorem in daily life — quick guesses, like someone’s nationality from a glance.
What I’ve noticed: my intuition does better when I don’t adjust for general priors. Corrections like “most people in Germany aren’t Russian” when someone looks vaguely Slavic often pull me further from the truth.
After five minutes of reflection, my best guess: explicit Bayes only really helps out-of-distribution, when we lack feedback loops — new domains, big decisions, reasoning about AI. That’s when 5 minutes of googling or reading a paper can give you better intuition than your System 1.
Is this roughly in line with the Sequences?
You might be underestimating the strength of evidence that looking vaguely slavic gives.