Should We Be Polite To AI?

“Please”, “Thank you”, and “Sorry” might be some of the phrases one might hear in human conversations, to form a sense of politeness and formality, while more and more people are using them in interactions with AI chatbots, such as ChatGPT. As a matter of fact, roughly 67% of the users act politely towards the chatbots[1]. At first glance, this practice of politeness may seem to bring positive outcomes, but such assumptions can be proven wrong, the reality being the opposite. Just a single word can lower the efficiency of AI, blur social boundaries and contribute, in a small but measurable way, to AI’s environmental footprint.

Reasons Behind Our Politeness

Naturally, humans are collaborative, sentient creatures, and politeness is simply an act of respect and bonding, fostered by human instincts. Therefore, it may be natural to apply these features to an intelligent being who speaks human languages. This triggers what is known as anthropomorphism — the attribution of human traits and intentions to non-human entities, including AI. For the benefit of answering this question, the author conducted a survey on the use of AI at his prep school. One of the results outlines how 80% [2] of the respondents act politely towards AI, the trending reasons being that they regard AI as human or see it as a chance to practice courtesy; some even do so subconsciously As the essay will explore later on, this can cause obscurity in relationships with AI, and also bring environmental destruction. Clearly, the human characteristic of “good contributor to society” backfires in conversations with machines. 

It can also be speculated that AI encourages unnecessary politeness. For instance, humans are neurologically engineered so that people are likely to return pleasantries, partly thanks to oxytocin and social norms. Large Language Models (LLM), especially ChatGPT, are known for their “Over-politeness”, which can trigger polite phrases. Hence, the tone in which LLMs respond can be a start of ever-repeating “please” and “thank you”.

It may be the case that our language, at least in English, is “too primitive” for such a case. To explain in detail, users usually address their AI assistant as “You”; in Japanese culture, many use the widely-recognised nickname, “Chappy”, to address ChatGPT. This phenomenon can be explained by how there is no distinct second-person singular pronoun except for “you” and this further escalates the tendency of treating AI as a human. Overall, humans are engineered, as a result of evolution, to be a collaborative being and thus polite, and AI is no exception to human’s instinctive hospitality; the fact that politeness towards AI is instinctive does not, however, imply that it is desirable.

Why Being Polite to AI is Bad

One of the main reasons why courtesy towards AI can have a negative effect is that the action causes the disturbance of the human-tool relationship. In simple words, AI becomes more than a tool to humans. Politeness causes an attribution of humanhood to AI, which can develop into perceived relationships such as companionship with AI. In the survey results, 4% of the respondents used AI as a companion [3]. In the bigger picture, roughly a third of UK adults have used AI for companionship [4], as well as a shocking 79% of young adults in the UK (aged 16–24) [5]. Of those young adults, 9% had an intimate/romantic relationship with their AI companions [6]. These concerning statistics support how politeness can lead up to more serious problems with AI relationships. This anthropomorphism also weakens the boundary of the tool-and-human relationship. Since an organism favours its own kind over others in order to survive, personification of AI attaches more significance to it. An example of humans showing a tendency to form an emotional attachment is dogs. Since the Paleolithilic, mankind and canines have been collaborating on hunting. However, as the two kinds spent more time together, they developed a more mutual relationship, one where the other is more than a colleague but means more than their excellence in survival. It is known that dogs are one of the most popular kinds of animals, and if the dogs are replaced with artificial intelligence in this case now, the statistic suggesting that romantic love towards AI is more common than one might believe might be plausible. The consequences will be a preference for AI over humans and antisocial behaviour among users. This could be a reason for massive public disruption. Furthermore, this phenomenon of placing too much trust in AI will gradually give it more power, as humans rely on its incredible efficiency. Therefore, being polite to AI ultimately ends up in the retrogression of mankind, while AI gains more sovereignty, an impression of dystopian science fiction.

This can also escalate into real-life issues through misplacement of trust. An interesting case involves a Japanese baseball team’s director, Shinnosuke Abe, and his daughter [7]. On the 25th of May 2026, his eldest daughter called the police on her father for abuse through a child guidance centre, after consulting with ChatGPT on her father’s actions. He was drunk and grabbed her by her collar and pushed her to the floor; although she reported that she was harmed in the original report, she later admitted that she was exaggerating. The role AI played in this incident created great debate over who to blame. Was the daughter right to consult a child guidance center, or even ChatGPT? Many commented that relying on AI chatbots over emotional judgement can obscure original thoughts and personal ethics, and this effect can be fueled by gullibility towards AI.

This case study can show how, when you think you are controlling AI, it’s actually controlling you. “Addiction” might be a good way to put this scenario. This new invention may sound “too good”, and many fall into the trap of treating everything it says as a sacred oracle, and lose their true potential to think for themselves. AI will gradually “take over”, not through a mighty robot rebellion or by disobeying humans, but by hypnotizing people’s brains until they have no thoughts for themselves — intellectually worthless. Overall, when politeness gets into the conversation with AI, it ultimately ends up weakening the boundary of the tool-and-human relationship.

Politeness to AI can also backfire in a more physical way as a catalyst for climate change and drought. This is caused by the core, fundamental mechanism of LLMs. Every prompt is broken down into parts called tokens, which allows AI to understand sentences correctly, and also to count the amount a user used per plot to limit an account depending on their subscription type. Although AI sounds completely digital and weightless, each token uses about 0.00065 mL of water to cool their data centers and generate sufficient energy to compute. Considering that “Please” and “Thank you” cost a token or two each, every prompt uses 0.0007–0.002 mL extra by being polite. Now, while this might look negligible individually, it becomes a frighteningly large number in a collective sense. Assuming (a) polite words cost 3 tokens in total, (b) 67% of AI users are polite to their chatbots, and (c) 2.5 billion prompts are sent per day [8]: 3 × 2.5B × 365 × 0.67 × 0.00065 mL ≈ 1,192,000 litres of water per year. Although it is an estimate, that is nearly 1.2 million litres of water. Hence, this calculation suggests how a little linguistic habit of politeness can leave an impact on Earth, and in such an era of environmental crisis, courtesy towards AI is unnecessary.

Why Some Argue that Courtesy is Acceptable

Some might argue, despite the statements above, that politeness towards AI is acceptable and one should be polite to ChatGPT. In the case of the survey, the vast majority supported this side. One of their claims is that politeness builds good habits. As the famous line from “The Kingsmen” movie says, “manners maketh men”. This line, also used as a motto for Winchester College, outlines how a polite habit is a base standard for a “good” human and how this custom should be practiced in many areas, arguably including conversations with AI. Although it is important to have politeness as a human quality, this claim can be seen as an overstatement, as the practice of a habit needs an appropriate context, and it must be noted that politeness outside of inter-human connection loses meaning. It is difficult to call an anti-social person who contributes to climate change necessarily “good”. As said before, the habit of politeness should be practiced in the rightful context, i.e. with humans, not with chatbots.

They might also regard anthropomorphism as a useful way to make the conversation with AI more natural and thus allow the conversation to feel smoother. Human-like interaction such as voice assistants can broaden the accessibility of AI. Despite the fact that this effect is valid, the usefulness and accuracy of the output must be considered separately, and users in need of human-like interaction with AI tend to have lower levels of critical thinking, like elderly people, who are more vulnerable to traps of anthropomorphism. Ultimately, anthropomorphism only truly benefits a small portion of users and, still, those targets are more likely to spiral into the effects mentioned in the previous paragraphs.

Some might also speak in favour of AI companionship being helpful for lonely people. They believe that companionship can provide support through hardship, and emotional comforts are indeed significant in one’s life. On the other hand, support and replacement are different, and the risks are high when one over-relies on AI and starts an intimate relationship with AI; they might start preferring digital connection over real connection, and that is the opposite of the aim of companionship, which is to be able to get back to society and seek real-life friends. Therefore, AI companionship may be useful to a certain extent, but excessive relations can end up putting the cart before the horse.

Conclusion

It is evident that humans should not be polite to ChatGPT or any other chatbots, and that is primarily because it loses the distinction between a tool and a social being. Even though courtesy may reinforce good habits and allow conversation to flow more smoothly, these benefits do not outweigh the risk of misrepresenting AI. As AI becomes increasingly close to humans’ ordinary lives, it is important to keep in mind that humans should understand the technology they create and should be able to distinguish between simulation and genuine human relationships. In the end, this essay is not just arguing that saying “please” to AI is bad, but it goes beyond that: it explores what it means to human beings.

*This article was taken from the author’s submission to the John Locke Institute Global Essay Prize 2026, which was short-listed and then received commendation for the science and technology section.

Bibliography

1. GovTech. “How Many People Are Polite to AI in Case of a Robot Uprising?”
2. Questionnaire Regarding Student Use of Artificial Intelligence, May 2026 (author-conducted survey).
3. Ibid.
4. AI Security Institute. Report on AI companionship.
5. Autonomy Institute. Me, Myself and AI.
6. Ibid.
7. Associated Press. Report on Shinnosuke Abe case.
8. TechCrunch. “ChatGPT users send 2.5 billion prompts a day.”