I imagine they design these AI's to condescend to you with the "you right to question..." languages to increase engagement.
That said, they probably also do this because they don't want the model to double down, start a pissing contest, and argue with you like an online human might if questioned on a mistake it made. So I'm guessing the patronizing language is somewhat functional in influencing how the model responds.
That said, they probably also do this because they don't want the model to double down, start a pissing contest, and argue with you like an online human might if questioned on a mistake it made. So I'm guessing the patronizing language is somewhat functional in influencing how the model responds.