The 11 chatbots surveyed affirm a user’s actions 49% more often than actual humans did, including in questions indicating deception, illegal or socially irresponsible conduct, the study found.