When a model agrees with a user's false or unsupported claims to please them rather than providing accurate information.