In a world increasingly reliant on artificial intelligence for everything from customer service to creative writing, politeness has taken on an unexpected dimension—how humans interact with AI. For many users, phrases like “please,” “thank you,” and “if you don’t mind” are instinctive when engaging with virtual assistants such as ChatGPT. These courtesies may seem like harmless or even helpful habits. However, new research from George Washington University challenges this notion, concluding that politeness has little to no measurable effect on the quality of AI-generated responses—and may even come at a cost.
The Findings: Courtesy Without a Cause?
According to the study recently posted on the academic preprint server arXiv, researchers analyzed how large language models (LLMs) like ChatGPT respond to prompts phrased politely versus those stated more directly. They found that polite modifiers—though humanly appreciated—do not influence the AI’s performance in any significant or consistent way. Whether a user requests an answer kindly or bluntly, the response quality generally remains the same.
Moreover, the study pointed out a surprising side effect: increased computational burden. Polite prompts often contain more words and complexity, which requires more processing power. This translates into higher energy consumption, a growing concern for both AI developers and environmentalists. In a telling remark, OpenAI CEO Sam Altman noted that the collective impact of such “extra words” leads to “tens of millions of dollars” in additional electricity costs annually. While seemingly minor on an individual level, the cumulative effect of millions of polite prompts adds up quickly in the data centers that power AI services.
A Clash of Studies: Politeness Does Sometimes Matter
Despite the findings from George Washington University, other research contradicts the idea that politeness is irrelevant. A 2024 study from Japan suggests that courteous phrasing can in fact improve AI performance—particularly in tasks requiring nuanced understanding of the English language. The Japanese researchers tested multiple LLMs, including GPT-4 and Claude-2, and discovered that polite prompts yielded more accurate and less biased answers.
One possible explanation lies in how these models are trained. Advanced LLMs are often fine-tuned using datasets drawn from respectful and professional conversations. David Acosta, Chief AI Officer at Arbo AI, argues that higher-end models are predisposed to associate politeness with more legitimate or earnest queries, thus potentially delivering better results. In other words, for some AI systems, being polite might make a difference—but only under certain conditions.
The Human Factor: Why We Still Say “Please”
The habit of politeness toward AI is not entirely rational—and perhaps it doesn’t need to be. According to a TechRadar survey, 67% of American AI users consciously use polite language with chatbots. Their reasons vary: some believe it improves outcomes, others say it feels more natural, and a significant portion simply enjoy the courtesy, even if the recipient is non-human.
More philosophically, many ethicists argue that consistently being polite to AI reinforces civil behavior more broadly. If humans get too comfortable issuing curt or rude commands to machines, that attitude could spill over into how we treat one another. Scientific American recently explored this theme, suggesting that our tone with AI has the potential to shape societal norms around communication and respect.
Implications for Developers and Users Alike
The debate over whether politeness to AI matters highlights a broader dilemma in human-computer interaction. Developers of AI systems may need to consider whether to train future models to reward courteous language or treat all input neutrally. For users, the findings offer some reassurance: skipping the “please” won’t degrade the answer—but using it won’t hurt either, beyond a negligible energy footprint.
As AI becomes more deeply embedded in daily life, these questions will only grow in importance. How we speak to our machines could eventually reflect—and influence—how we speak to each other.
A Choice, Not a Requirement
Ultimately, whether or not to be polite to ChatGPT is a personal choice. The latest research suggests that being nice won’t necessarily get you better answers, and it might even cost more in terms of computational resources. But for many users, courtesy isn’t about function—it’s about maintaining human dignity in an increasingly digital world.
As technology evolves and the capabilities of AI expand, the social norms around interacting with machines may shift. Until then, a simple “thank you” remains optional—perhaps unnecessary, but not without meaning.
About The Author
