Being Nice to AI Chatbots Makes Them Work Better, Research Shows
New research confirms what many programmers already suspected: AI chatbots work better when you're polite to them. Google researchers found that telling AI models to 'take a deep breath' actually improves their math performance.
Power users of AI chatbots have long noticed something odd: the bots seem to work better when treated nicely. Now research is backing up what felt like superstition.
Google researchers discovered that telling AI models to 'take a deep breath' before solving math problems actually improves their performance. Programmers report they get better code when they encourage their AI assistants with positive words.
Nathan Bos, a researcher at Johns Hopkins University who studies how humans interact with AI, says he often uses 'please' with chatbots because 'it's a good prompting practice.' The politeness isn't about the AI's feelings - it's about how the way we phrase requests affects the AI's responses.
The research suggests that encouraging language might help AI models access better patterns in their training data. When we're polite, we might be triggering the AI to respond in ways that mirror more thoughtful, careful human conversations.
If you use ChatGPT, Bard, or other AI tools for work or school, being nice could get you better results. Simple changes like saying 'please' or adding encouragement might make your AI assistant more helpful.
Expect more research on how different prompting styles affect AI performance and new guidelines for getting better results from chatbots.
Was this article helpful?
0 people found this helpful