Should you be polite to LLMs?

By Marie Haynes
3 min read

Table of Contents

OK be honest. Do you say please and thank you to ChatGPT, Gemini and your other LLM tools? I do!

I think I mostly do this because it's in my nature to be polite.

It turns out though that research supports the idea that we get better responses if we are nice to our AI friends.

Research says it pays to be nice to LLMs

Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance

This study found that the politeness of prompts can significantly impact LLM performance. Impolite prompts can lower the performance of LLMs. Highly respectful prompts don't seem to improve performance. The study concluded that "moderate politeness is better."

Manners Matter? The impact of politeness on human-LLM interaction

In this article, Vik Bogdanov ran some experiments to see if his tone influenced the quality of AI generated results. He used different accounts and asked questions about his business in varying degrees of politeness. Both ChatGPT and Gemini responded to the polite prompts with more detail and accuracy. The impolite prompts produced less effective or biased answers.

Let's test it out!

ChatGPT

I wrote 2 prompts in different ChatGPT conversations:

This second one was hard to write!

Both of the responses were similar.

I apologized to ChatGPT (lol) and then asked what the differences were in the 2 responses.

It turns out that for the polite question, the LLM gave me more information, and a slightly better structure.

Gemini Advanced

I did the same prompts both polite and then mean in different conversations. Gemini shared similar information about my company. I once again, explained what I was doing.

Gemini told me that the rude prompt produced a more formal and concise answer. The polite prompt's response was warmer and more conversational. The rude prompt focused more on the company while the polite one had more information about me as a person.

LLMs often try to mimic your conversation style. I always have a good laugh when people tell me their responses from LLMs are lacking in intelligence. 😜

I do think it pays to be polite to LLMs. Just like in human conversation, you're more likely to have an agreeable response if you are nice!

🤖
How was AI used in writing this post? This article was completely written by me, Marie, a human. The image for this post was created in Gemini Advanced with the prompt, make an image where two people are talking to a robot. One is mean and one is nice. Pixar style.

Last Update: November 07, 2024

About the Author

Marie Haynes

I love learning and sharing about AI. Formerly a veterinarian, in 2008, understanding Google search algorithms captivated me. In 2022 my focus shifted to understanding AI. AI is the future!

View All Posts