Why being polite to AI might be hurting your results
- A study titled 'Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy' was presented at NeurIPS 2025.
- The research found that accuracy of LLMs increased as prompts became less polite.
- Very polite prompts achieved an average accuracy of 80.8%, while very rude prompts reached 84.8%.
The findings of this research challenge conventional wisdom about politeness in AI interactions, suggesting that users may need to rethink their approach to prompting LLMs. While the results are intriguing, they also raise ethical questions about how AI systems interpret social cues and the implications for factual accuracy. Further exploration is needed to understand the broader impact of tone on AI performance across different tasks and contexts.
