Could one deception check underpin almost all AI interpretability?

One post suggests that almost everything in interpretability may come down to asking whether a statement is deceptive—and getting as confident as possible in that check.