“A study at the Wharton Generative AI Labs found that for certain math problems, being polite to an AI by saying "please" resulted in more accurate answers.”