I’ve occasionally been guilty of that, but I see it less as a magical talisman of +1 to prediction skills and more as a means of improving my calibration after the fact: if I discover I’ve been systematically overweighting (or, much less likely, underweighting) the probability of some category of events, that’s a good clue that my thinking about them is flawed in some way. Can’t do that if I don’t quantify my uncertainty in the first place, although of course the flaw might come in at the quantification step.
This only works if you actually go back and collect that data, though, and I haven’t generally been very good about using PredictionBook or any similar tools.
I mostly approach this as a set of jargon words that express finer gradations of confidence than the conventional language.
That is, in normal speech I use tags like “I suspect that X,” “I expect that X,” “I’m fairly confident that X,” “I doubt X”, etc. On LW I use “I’m N confident that X” instead, where N is typically expressed to one significant figure (except I use ”.99+” to denote virtual certainty).
I endorse that, although I also endorse remembering that what I’m talking about is my intuitions, not reality. That is, when I say I’m .7 confident that it’s going to rain this afternoon, I have said something about my mind, not about rain.
I do find that the exercise of thinking more precisely about what my intuition actually is is helpful in encouraging me to pay more attention. That is, trying to decide whether I’m .6 or .8 confident in X (or whether all I can really say is that I’m .6-.8 confident) is a meaningful exercise in clarifying my own thoughts about X that I’m not as encouraged to do if my lexical habit is to say “probably X.”
I do this in real life quite often. But I always try to explain that I’m talking about my state of mind. I occasionally get good reactions to this along the lines of whomever I’m talking to not having ever thought about the distinction between rain and what your mind thinks about rain.
I express completely or nearly-completely subjective degrees of belief as numbers (and hear people doing the same) so often that I sometimes forget that not everybody does that, and take a while to realize why people hearing us talk look startled. (I guess that’s what happens when most of the people I hang around with are physicists, and many of those who aren’t are into sports bets and/or poker.) I don’t think I “feel so much more Bayesian” for that (except to the extent that I know frequentists would feel an even more painful kick), though; I mostly take that to be a figure of speech, as TheOtherDave says.
Possibly related… I developed the mannerism with my husband of answering rhetorical questions with oddly precise arbitrary numbers a while ago, mostly as an expression of dislike of rhetorical questions. (As in “How weird is that?” “Twelve.”) It amuses him, lets me blow off steam, and really bewilders listeners.
Possibly related… I developed the mannerism with my husband of answering rhetorical questions with oddly precise arbitrary numbers a while ago, mostly as an expression of dislike of rhetorical questions. (As in “How weird is that?” “Twelve.”) It amuses him, lets me blow off steam, and really bewilders listeners.
That’s a really good idea. I’m going to make it my new policy!
This only really irritates me when the person stating the estimate feels obligated to put a “p” somewhere, e.g., “I believe blah with p = 0.04″ or “Blah has probability (0.05 < p < 0.1)”.
This just signals a confusion of frequentist p-values and subjective-ish Bayesian prior probabilities, and indicates to me no real understanding of either of them.
Some of us do so because we have been or expect to be asked to do so, or both. Some others maybe just want to fit in, as they entirely knowingly affect.
Oh, I got another one, mostly confined to this forum: people making up numbers for probabilities of certain events and feeling so much more Bayesian.
I’ve occasionally been guilty of that, but I see it less as a magical talisman of +1 to prediction skills and more as a means of improving my calibration after the fact: if I discover I’ve been systematically overweighting (or, much less likely, underweighting) the probability of some category of events, that’s a good clue that my thinking about them is flawed in some way. Can’t do that if I don’t quantify my uncertainty in the first place, although of course the flaw might come in at the quantification step.
This only works if you actually go back and collect that data, though, and I haven’t generally been very good about using PredictionBook or any similar tools.
How are we supposed to get better at quantifying our degree of belief without practice?
You’re not, which is why not keeping track of the results is a way of doing it wrong.
(Not that I do it, but then I don’t assign spurious numbers to my feelings either. Possibly I should, but if I do then I need to keep track.)
I mostly approach this as a set of jargon words that express finer gradations of confidence than the conventional language.
That is, in normal speech I use tags like “I suspect that X,” “I expect that X,” “I’m fairly confident that X,” “I doubt X”, etc. On LW I use “I’m N confident that X” instead, where N is typically expressed to one significant figure (except I use ”.99+” to denote virtual certainty).
I endorse that, although I also endorse remembering that what I’m talking about is my intuitions, not reality. That is, when I say I’m .7 confident that it’s going to rain this afternoon, I have said something about my mind, not about rain.
I do find that the exercise of thinking more precisely about what my intuition actually is is helpful in encouraging me to pay more attention. That is, trying to decide whether I’m .6 or .8 confident in X (or whether all I can really say is that I’m .6-.8 confident) is a meaningful exercise in clarifying my own thoughts about X that I’m not as encouraged to do if my lexical habit is to say “probably X.”
I do this in real life quite often. But I always try to explain that I’m talking about my state of mind. I occasionally get good reactions to this along the lines of whomever I’m talking to not having ever thought about the distinction between rain and what your mind thinks about rain.
I express completely or nearly-completely subjective degrees of belief as numbers (and hear people doing the same) so often that I sometimes forget that not everybody does that, and take a while to realize why people hearing us talk look startled. (I guess that’s what happens when most of the people I hang around with are physicists, and many of those who aren’t are into sports bets and/or poker.) I don’t think I “feel so much more Bayesian” for that (except to the extent that I know frequentists would feel an even more painful kick), though; I mostly take that to be a figure of speech, as TheOtherDave says.
Possibly related… I developed the mannerism with my husband of answering rhetorical questions with oddly precise arbitrary numbers a while ago, mostly as an expression of dislike of rhetorical questions. (As in “How weird is that?” “Twelve.”) It amuses him, lets me blow off steam, and really bewilders listeners.
I’ve done this for years, and “Twelve” is my go-to number. So much so, that my wife often preempts my retort with “And don’t you say ‘twelve’!”.
Sometimes I throw a “Seven” in there, just because I’m a wild and crazy guy.
I’m amused; my husband and I have precisely that dynamic. And I use seventeen as my alternate meaningless number, in almost exactly the same way.
Is that because of the humbug?
That’s a really good idea. I’m going to make it my new policy!
This only really irritates me when the person stating the estimate feels obligated to put a “p” somewhere, e.g., “I believe blah with p = 0.04″ or “Blah has probability (0.05 < p < 0.1)”.
This just signals a confusion of frequentist p-values and subjective-ish Bayesian prior probabilities, and indicates to me no real understanding of either of them.
Some of us do so because we have been or expect to be asked to do so, or both. Some others maybe just want to fit in, as they entirely knowingly affect.
Products of the environment, you know.
That really gets me.