Bayesian confirmation by uncertain evidence: epistemological and psychological issues