Politeness makes ChatGPT dumber. Science says so.
You’ve been too well mannered to your AI. You say “please,” “thanks,” and “kindly help me.” You deal with ChatGPT like a digital butler that deserves manners. You assume being well mannered helps it perceive you higher. It doesn’t.
A brand new research from Penn State says your good manners are hurting the mannequin’s efficiency. The extra well mannered you’re, the more severe it will get. When customers had been impolite, ChatGPT gave extra right solutions. When customers had been good, it made extra errors. The distinction wasn’t small both. The “very well mannered” prompts scored 80.8% accuracy. The “very impolite” ones hit 84.8%.
We’ve been advised tone doesn’t matter to machines. However apparently, it does. It’s not about feelings or ego. The mannequin doesn’t get offended. It simply performs higher when your message is blunt, direct, and stripped of social padding. “Would you please resolve this downside” makes it stumble. “Remedy this downside” makes it focus.
This flips your complete “be good to AI” narrative. Each tech influencer advised you to “converse naturally” or “be type” when prompting. That recommendation sounds human, not technical. Machines don’t care about etiquette. They care about construction. Politeness provides noise.
