I have a question concerning subjective bayesianism.
Even If a bayesian uses conditionalization and baye's rule to update their credence distributions, in what sense can they actually interpret their result as a credence, Even in the limiting, for example in de finetti's convergence results one only gets a credence of 1 in a objective chance hypothesis- a value equal to the relative frequency. Do subjective bayesians simply construe this further 'objective probability itself as a credence, and then use the law of total probability to get from cr(cr(A)=x)=1 to cr(A)=x. Moreover, i know that this is perhaps why de finetti did not like talk of second order probabilities, but then the question is; given that prior distribution is generally a distribution over probability hypotheses, does the subjective bayesian just construe these probability hypotheses as themselves subjective credences, and take the actual prior credence to be the expected value of the subjective credences.. This would avoid having to talk of objective probabilities or having to use the principle principle to transform them into subjective probabilities in order to a singular credence value for an outcome
This would appear to invoke Moore like paradoxes for reasons i can go into later (ie so long as Cr(credence=expected credence) does not equal 1, then there will be credences (in your prior distribution) distinct from your expected credence, but which have themselves positive credence, and that one would appear to be in a juxtaposition between moore like paradoxes (my credence is x, but it is possible (by regularity) that ,my rational credence is/should be some value not equal to x) one hand and radical insensitivity to data (the old evidence problem) where the credence you assign to your credence is 1) or radical sensitivity where you only have a prior probability about singular outcome and as soon as it occurs your credence jumps to one or zero).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Probability