Â鶹´«Ã½

From Lawrence D'Oliveiro

Robert Matthews talks of the connection between belief and Bayes’s theorem (13 March, p 38). He seems to think that, because you have to start with some arbitrary level of belief in order for Bayesian reasoning to work, it follows that different people starting from different points, and looking at the same evidence, might never come to the same conclusions.

In fact this point was addressed in your previous feature on Bayes’s theorem (22 November, p 36). There it was pointed out that the system was largely insensitive to the initial arbitrary starting probability. No matter where you start from, you will still converge on the truth.

Alternatively, consider that Bayesian reasoning can be applied to any kind of belief in the face of evidence, including belief in the validity of Bayesian reasoning itself. If it were possible for people starting with different assumptions, looking at the same evidence, to come to different conclusions about the validity of Bayesian reasoning, then that would make the validity of Bayesian reasoning itself completely subjective, according to Matthews’s own argument.

Which means his argument itself is built on an entirely subjective basis, and is not to be taken seriously.

Hamilton, New Zealand