I am not convince confidences and beliefs can be meaningfully discriminated.
Perhaps you might try to discriminate the two by saying something like: one believes a proposition if one holds that the proposition is more likely to be true than false, while one is more or less confident that a proposition is true / false only after one believes it to be true or false.
But why can’t we drop belief-talk entirely and speak only in confidences? Seemingly we could in every instance replace ‘belief’ with ‘confidence.’ E.g. rather than ‘I believe the painting is a Vermeer,’ we could say ‘I am confident that the painting is a Vermeer’ (indeed, I would say the latter is preferable, given fallibilism).
We could even give a plausible definition of evidentialist epistemic justification like this:
S knows p iff S is confident that p is true, S’s evidence supports the degree of confidence S has in p, and S ranges his confidence in p on the evidence.
If we can plausibly replace ‘beliefs’ with ‘confidences,’ then are the two distinguishable?
A couple things. Even if we could always replace belief-talk with confidence-talk I don’t think that would show that we should drop belief-talk. We can replace table-talk with particles-arranged-table-wise-talk, but we shouldn’t drop table-talk since there are tables. Tables are part of the world even though we could paraphrase them out of our language. Just like tables are a part of the universe, so are beliefs. There are states of concluding or accepting that a proposition is true. So, we shouldn’t take out belief-talk even if we could paraphrase it away.
But, I don’t think we can successfully replace belief-talk with confidence-talk. I take it that I can believe something with no confidence — I can conclude that it is true but have no confidence that it is true. If so, then at least one of those attitudes (the belief or the degree of confidence) would be irrational, but such a case nevertheless seems *possible* and so beliefs and confidence are to be distinguished.
You’re right that there is no threat to evidentialism here (at least to the spirit of it).