I understand this move but I don't like it. I think that in the fullness of time, we'll see that probability is not a kind of preference, and there is a "fact of the matter" about the effects that actions have, i.e. that reality is objective not subjective.

I think that probability is a tool for preference, but I also think that there is a fact of the matter about the effects of actions, and that reality of that effect is objective. This effect is at the level of the sample space (based on all mathematical structures maybe) though, of "brittle math", while the ways you measure the "probability" of a given (objective) event depend on what preference (subjective goals) you are trying to optimize for.