Wonder if the context makes a difference. Would people condemn police detectives using Bayesian reasoning to decide which lead to follow in a murder investigation, for instance?
People morally condemn and punish others for applying sound "Bayesian" statistical reasoning to group members - which they blithely adopt themselves. https://journals.sagepub.com/doi/abs/10.1177/0956797618805750 …pic.twitter.com/FcM9yBGHCz
-
-
-
Not only would they condemn it, they’d hold a protest
-
Quite the hypothesis.
-
It's ironic don't you think, that the snippet shows a strong lack of understanding just what bayesian probability is. First, "bayes rule" is regular conditional probability. Second, while being a surgeon would be higher, having same skills as surgeon would be exactly the same
-
That is, once we've observed the act, even though probabilities of being a surgeon continue to gate on sex, the probability of being as skilled as a surgeon gates solely on succesfully completing the act. Thus the probability of having the same skill is the same regardless sex
-
Second, there is no such thing as "Bayesian rationality". There is "consistent updating of a set of weights assigned to some set of propositions after observations". It is impossible to achieve exactly in reality due to computational intractability. Even approximate is too hard.
-
Here is what I mean, written in a probabilistic programming DSL I work on. observe (act && g =X); return (issurg) vs observe (act && g =X); return (sameskills), the latter will not depend on what g is equal to while the prior will.pic.twitter.com/jc1ontIEf6
-
Oh, to show information flows properly (is surgeon? when the act isn't observed, conditional probability is: (also, model correctly captures that P(surg | failed) < P(surg)))pic.twitter.com/SXaDP45LH6
End of conversation
New conversation -
-
-
I have trouble with the "Forbidden base rate" theory on the ground, because just like a judge instructing a jury to disregard a prejudicial outburst, it's hard to get people living in the real world to disregard other variables influencing their perception of data 1/
-
... Including their suspicions as to the intent of the questioner (what is the real question here and how will my answer be framed) and their hypotheses about third factors influencing the observed data 2/2
End of conversation
New conversation -
-
-
The first 2 studies in this paper have a strange structure that seems to introduce an asymmetry in likelihoods: the likelihood of *doctor *given *performed surgery* is much higher than firefighter| extinguished fire and butcher|butchered pig. 1/
-
This differential in likelihoods would influence respondents' view of the Bayesian reasoner more than social desirability bias. I'm not sure using a Likert scale on fairness accounts for the way people heuristically estimate likelihoods 2/
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.