Thursday, December 17, 2009

Bayesian 'Reasoning' vs. Judgment

Right, so what is the prior probability that a baby dying is worse than an adult woman dying?

How would you even start to collect such a statistic?

The purpose of Bayesian Reasoning is, ultimately, to make decisions. Beyond even its issues coming from being derived from an existing framework, the problem is that, even if priors can be calculated, all reasoning begins and ends at human judgment and human values. Bayesian calculations are just something that can make the middle part easier. I strongly suspect Reverend Bayes would have been the first one to tell the modern Bayesian this. (Would have replaced the concept 'human values' with 'divinity,' though.)

No comments: