This reads like a stream of loosely connected thoughts. Maybe there's an essay in there but it hasn't really been written yet.
My best guess is that what they're trying to say is, most of the time that we talk about "likelihood", we're talking about things that aren't really determined by random variables, and so therefore probability theory is not a useful model. So, for example, "how likely am I to have a sandwich for lunch?" is not really a question of probability.
theres a lot of talk about bias skewing our expectations, which leads to discussions about why we "probably" expect certain outcomds, and if we "update our priors" we could better adjust our thinking to the available outcomes.
i think he might be confusing the analog vs digital of probability assessment. one may be able to catalog outcomes into bins of good and bad and neutral, then count up each bin and predict the type of outcome.
but those bins are still just filled with discrete expectations and because of that, what matters more is your ability to imagine various outcomes based on some expansive maze like pathfinding.
so one way to interpret it is that no one has a bayesian thought process (eg, this outcome is 80% probable) but the more useful conception is that one can learn to identify appropiate categories of important outcomes and sort out scenarions into these minimally differentiable outcomes and come up with a derivative probably of a good, bad or neutral outcome and not any discrete scensrio.
4 comments
[ 2.6 ms ] story [ 20.1 ms ] threadMy best guess is that what they're trying to say is, most of the time that we talk about "likelihood", we're talking about things that aren't really determined by random variables, and so therefore probability theory is not a useful model. So, for example, "how likely am I to have a sandwich for lunch?" is not really a question of probability.
Is there more to what he's saying than that?
i think he might be confusing the analog vs digital of probability assessment. one may be able to catalog outcomes into bins of good and bad and neutral, then count up each bin and predict the type of outcome.
but those bins are still just filled with discrete expectations and because of that, what matters more is your ability to imagine various outcomes based on some expansive maze like pathfinding.
so one way to interpret it is that no one has a bayesian thought process (eg, this outcome is 80% probable) but the more useful conception is that one can learn to identify appropiate categories of important outcomes and sort out scenarions into these minimally differentiable outcomes and come up with a derivative probably of a good, bad or neutral outcome and not any discrete scensrio.