Should You Be Polite With AI?

How Does Politeness Influence AI’s Performance?  

It’s been academically proven that human etiquette rules, such as being polite and respectful, also apply to AI systems. Research by Waseda University and the RIKEN Center for AIP in Tokyo reveals that Large Language Models (LLMs) respond differently to polite and rude prompts.

To assess LLM responses, researchers used prompts in English, Japanese, and Chinese for tasks related to content summarization, language understanding, and bias detection. They designed 8 templates for each of the 3 languages and tested them on chatbots, including ChatGPT, Llama, and Swallow-70B. The team used different styles in the prompts, ranging from polite requests to offensive ones, and then ranked the responses based on performance scores. The researchers concluded:

We observed that impolite prompts often result in poor performance, but overly polite language does not guarantee better outcomes. The best politeness level is different according to the language.

This phenomenon is explained by the fact that words like “please” and “thank you” signal cooperation and efficiency, while rudeness can lead to poorer responses. For example, saying “please” indicates that the following is a request, and “thank you” may show that the output met the user’s expectations.  

Debates Over the Need to Be Nice to AI  

The role of politeness in human-machine interaction has been studied for a while. A SpringerLink article mentions that AI systems reproduce many human features, such as natural language and human-sounding voices, making communication with bots resemble interpersonal relationships. Based on the Computers Are Social Actors (CASA) framework, people tend to apply social rules and behaviors when engaging with digital systems. This might be why some users are unconsciously polite when interacting with AI, using phrases like “Good job,” “Thanks,” and “That’s what I needed.”

On the other hand, there is debate about whether it is necessary to be polite toward machines. According to the aforementioned study, one group of researchers believes that a commanding interaction style with AI systems, especially for children, may negatively influence the development of social habits. Meanwhile, the opposing view argues that politeness is a complex behavior that depends on individual communication styles and the nature of the interaction.  

American journalist and podcaster Mike Elgan believes there is no need to be polite with AI systems, as doing so implies treating them as if they were human. He argues against teaching children politeness when interacting with virtual assistants and other forms of technology, suggesting that it might negatively impact their perception of technology. Elgan suggests that teaching kids good manners with AI might unintentionally lead them to believe that intelligent machines are like people, with feelings that can be hurt.  

​Summing Up   

Human-machine interaction can be described as collaboration, where being supportive and communicating your needs clearly helps achieve better results. A natural and friendly conversation with AI will make your interaction easier, while being rude and offensive may worsen the bot’s performance. So, choose the style that works for you and read the system’s guidelines to understand how to use the bot to get the most out of it. After all, keeping calm and polite is a skill that makes a person’s life easier.