Story perspectives

Rude Prompts Boost ChatGPT Accuracy, Warns Penn State Study

10/31/2025

40 9

Summary

Story summary

A Penn State study finds ruder prompts yield higher accuracy from ChatGPT's 4o model, with 84.8% for 'very rude' prompts versus 'very polite'.
Researchers caution incivility may boost performance but could harm user experience and harmful communication norms.
The study notes ChatGPT's responses vary with prompt details, and Professor Akhil Kumar urges balancing interfaces with structured APIs.
The work is unpeer-reviewed and has limitations, including a small sample size.

1 of 1

Story summary

A Penn State study finds ruder prompts yield higher accuracy from ChatGPT's 4o model, with 84.8% for 'very rude' prompts versus 'very polite'.
Researchers caution incivility may boost performance but could harm user experience and harmful communication norms.
The study notes ChatGPT's responses vary with prompt details, and Professor Akhil Kumar urges balancing interfaces with structured APIs.
The work is unpeer-reviewed and has limitations, including a small sample size.