Most people can’t do bayesian inference properly. in politics bayesian-type inference is

grotesquely misused. Therefore, I think the standard for public discourse should

discourage this inference. I may be completely wrong about my interpretation.

But im specifically talking about things like trusting of sources, corrupting influences etc.

Sides are able to simply dismiss fully legitimate arguments from their rivals by identifying

the specific arguments with the most disreputable and/or disliked institution/person

making that argument, eg. identifying free trade with Heritage or AEI, anti-torture arguments

with ACLU (bad in some people’s worldviews).

We should simply strive to answer the best arguments on the other side, and once we’re fully

and completely confident that we have made a good-faith effort to address the best argument,

and all we get in return is ad-hominem and/or bad faith…should we START to make bayesian

inferences about sources, corrupting influences, follow-the-money thinking, etc.

# Reverend Bayes Eschews Politics

Advertisements

I find the standard way that the Monty Hall problem is explained is harder to understand than the following:

1. What ever you choose, you have a two thirds probability of being wrong. That is, the car is more likely behind one of the other two doors.

2. When the host reveals that it is not behind one of those doors, you have a two thirds probability that it is behind the remaining door (the one you did not choose), so you should switch. The likelihood that you are wrong is only 33% (the original probability that you had chosen the correct door).

However, I still Bayesian stuff hard.

Yes I was also confused by the standard explanation for Monty Hall. Bayesian reasoning on complex topics requires even more care, which is my point – it is notoriously easy to misuse. $X million dollars donated by oil industry to climate change skeptics => AHA! Discount to 0! Oh but I didn’t realize that BILLIONS of dollars go to climate scientists to support AGW. An easy example.