Nice review (particularly the end), but I’m still looking for a more in-depth, nerdy takedown of the Bayesian fetish. https://thepointmag.com/politics/the-bookmaker/
-
@UlrikeHahn While not a great description of Bayesianism, I think that they do assign probabilities to arbitrary propositions. Am I missing something? https://scholar.social/@ehud/113017242130000833
-
@ehud 2/3 if, instead, “they” and “Bayesians” simply refers to people for whom the Bayesian framework somehow features centrally in their work (thousands of people across many disciplines) or, alternatively, the large set of people who prefer Bayesian over frequentist statistics (even more people) then it feels like saying
“Bayesians wear purple socks”
because Nate Silver (or Silver and some of his friends) do (if, in fact, Silver actually subscribes to that claim)
-
@ehud the sentence “Bayesians believe we can quantify our certainty about any statement we can imagine”
strikes me as wrong in a technical sense for two reasons: 1. “Bayesianism” describes an idealised formal framework not an account of what people can or cannot do 2. among those working on that framework, the possibility of statements that are assigned no specific value, or are assigned ranges, is a matter of explicit debate, see 3.1 here, for example https://plato.stanford.edu/entries/epistemology-bayesian/ 1/3
-
@ehud … if you had written papers that had the word “Bayesian” in the title and felt about Nate Silver the way I do, you would want people to be precise about the meaning of “Bayesian” too ….
3/3
-
@UlrikeHahn it’s more about what I like to call the Bayesian fetish than actual Bayesianism.
-
@ehud ah, could you elaborate?
-
@UlrikeHahn It would require an essay... But fundamentally, it is people who claim that they are "rational" because they never say something is absolutely true or false (probability of zero or one), and like to believe that they update their credences slowly, based on the quality and degree of evidence (as if those are simple to judge).
-
@ehud assigning probability zero or one (by definition) means you can never change your mind, there is no amount of evidence whatsoever that could alter your belief.
isn’t that something we would want to typically avoid?
-
@UlrikeHahn , indeed that’s what they argue. I don’t have time now to write a lot on this, but for the record I do object to this as a good rule of thought to follow in practice (or pretend to).
-
Bayesian fetishism, like game theory, ir, and nuclear deterrence fetishism, lack of:
' Socratic humility:
we should try to get to know how much we do not know, as filling these gaps with heuristics and misplaced certainty will yield flawed and potentially disastrous predictions and recommendations.' -
@teixi
interesting paper, and definitely agree on the point about epistemic utility.still unclear who the Bayesian fetishists are, though