Hi Cip, thanks for finishing the book. I am glad you liked it! I am not sure how Popper and Deutsch would be antidotes (I know the former, not the latter). I know Taleb's critique which is interesting. In my view, the main problem with Bayesianism is the one I lay out in the epilogue. As Binmore says it is a bit of a mystery how we chang…
Hi Cip, thanks for finishing the book. I am glad you liked it! I am not sure how Popper and Deutsch would be antidotes (I know the former, not the latter). I know Taleb's critique which is interesting. In my view, the main problem with Bayesianism is the one I lay out in the epilogue. As Binmore says it is a bit of a mystery how we change our mind when we are surprised (well at least if you start from a Bayesian point of view).
I read too much into your critique of Bayes then. I have deeper problems with Bayes. The 2 main ones are its naive application to Extremistan or ludic fallacy (both Taleb terms which are pretty good) - Bayes is dangerous applies to fat tails; second is the epistemology implied by Bayes which is that one can incrementally get to the truth by updating their views based on evidence (in contrast with Popper's falsification); again this is more relevant in Extremistan.
David Deutsch is one of the brilliant minds outthere and his book "Beginning of Infinity" nothing short of exceptional. He is a Popperian but makes progress on Popper.
I think Taleb's critique is related to Binmore's. The fat-tail problem can be interpreted as reflecting surprises. Surprises arise from the fact that your model of the world does not account for all possible events (e.g., US intelligence did not consider the 9/11 scenario as a possibility). The solution may be to adopt a Bayesian approach but allocate some weight to the possibility of unknown events. This leads to different decision-making rules, such as the 'Precautionary Principle.' This recommendation aligns with Taleb's criticism. My colleague Quiggin has written on this: https://ageconsearch.umn.edu/record/149847/?v=pdf
This is likely a different discussion, perhaps you can write a post about the Bayes misapplication / ludic fallacy / precautionary principle & we can engage there. Taleb has a nuanced view on the precautionary principle which he calls the "non-naive" precautionary principle. For any innovation / "surprise", it distinguishes between thin and fat tails domains & local and systemic risks. It allows for local-bounded/thin tailed domain errors even if the problem is not well understood or complex.
Hi Cip, thanks for finishing the book. I am glad you liked it! I am not sure how Popper and Deutsch would be antidotes (I know the former, not the latter). I know Taleb's critique which is interesting. In my view, the main problem with Bayesianism is the one I lay out in the epilogue. As Binmore says it is a bit of a mystery how we change our mind when we are surprised (well at least if you start from a Bayesian point of view).
I read too much into your critique of Bayes then. I have deeper problems with Bayes. The 2 main ones are its naive application to Extremistan or ludic fallacy (both Taleb terms which are pretty good) - Bayes is dangerous applies to fat tails; second is the epistemology implied by Bayes which is that one can incrementally get to the truth by updating their views based on evidence (in contrast with Popper's falsification); again this is more relevant in Extremistan.
David Deutsch is one of the brilliant minds outthere and his book "Beginning of Infinity" nothing short of exceptional. He is a Popperian but makes progress on Popper.
I think Taleb's critique is related to Binmore's. The fat-tail problem can be interpreted as reflecting surprises. Surprises arise from the fact that your model of the world does not account for all possible events (e.g., US intelligence did not consider the 9/11 scenario as a possibility). The solution may be to adopt a Bayesian approach but allocate some weight to the possibility of unknown events. This leads to different decision-making rules, such as the 'Precautionary Principle.' This recommendation aligns with Taleb's criticism. My colleague Quiggin has written on this: https://ageconsearch.umn.edu/record/149847/?v=pdf
This is likely a different discussion, perhaps you can write a post about the Bayes misapplication / ludic fallacy / precautionary principle & we can engage there. Taleb has a nuanced view on the precautionary principle which he calls the "non-naive" precautionary principle. For any innovation / "surprise", it distinguishes between thin and fat tails domains & local and systemic risks. It allows for local-bounded/thin tailed domain errors even if the problem is not well understood or complex.
https://arxiv.org/abs/1410.5787
Thanks Cip, I read Taleb a long time ago. I'll put a note on possibly writing a post on that topic in the future and keep this as a reference for it.