Is "trust science" a good epistemic foundation?