Probability assigned to a belief is estimated according to a goal that says “high Bayesian score is good”. It has a particular optimization target, and so it doesn’t run away from it. It is an “objective” fact about this optimization process that it tries to have a good Bayesian score. The probability that it produces is within its mind, and in this sense can be said to be “subjective”, but it is no more subjective than my decision to take an apple from the table, that also happens in my mind and is targeted on a goal of eating an apple. There is no point in assuming a state of mind just for the sake of reflecting the territory, if it doesn’t contribute to the goal. In this case, the goal happens to include the state of mind itself.
Probability assigned to a belief is estimated according to a goal that says “high Bayesian score is good”. It has a particular optimization target, and so it doesn’t run away from it. It is an “objective” fact about this optimization process that it tries to have a good Bayesian score. The probability that it produces is within its mind, and in this sense can be said to be “subjective”, but it is no more subjective than my decision to take an apple from the table, that also happens in my mind and is targeted on a goal of eating an apple. There is no point in assuming a state of mind just for the sake of reflecting the territory, if it doesn’t contribute to the goal. In this case, the goal happens to include the state of mind itself.