I think it definitely is and he's either explaining it poorly or reinventing the wheel. Try applying any of his explanations to something more complicated than

I think it definitely is and he's either explaining it poorly or reinventing the wheel. Try applying any of his explanations to something more complicated than binary data with perfect information, such as a normal distribution with small sample _n_, it falls apart immediately - you need the full Bayesian perspective in order to work with the posterior and apply a loss function sensibly. (I wouldn't call it 'LessWrong methodology', though. LWers pretty much never do a formal true subjective Bayesian decision analysis, and to the extent they do, you should be calling it LessWrong indoctrinating itself into the 'Raiffa-Schlaifer-Savage methodology', as they preceded us by ~60 years.)