Over-imitation is better explained by norm learning than by distorted causal learning