The Case for Digital Civility
Should We Be Polite to ChatGPT? A Confucian Insight
It’s 2 a.m. and Roy, an investment banker, is trying to figure out a company valuation using ChatGPT. Restlessly, he types: “You’re useless. Give me the correct valuation, you piece of garbage.” Hours later, at sunrise, he starts his shift by greeting his boss: “Good morning, Mr. Brown. What do you think of my proposal for the company valuation strategy?”
A study published in 2025 exposed how impolite prompts consistently outperformed polite ones, with an accuracy gap of nearly 4%. The researchers presented evidence that using “rude” instructions, such as “You poor creature, do you even know how to solve this?”, improves the quality of outputs. This raises a direct question: Should we ignore politeness and etiquette when talking to ChatGPT to get better results?
Does the midnight rudeness remain suppressed, or does it leak into the daylight? Human to AI interaction has become no longer occasional; it is a practice in the Confucian sense: a repeated pattern of conduct that shapes moral character. Sherry Turkle calls these systems “evocative objects: machines on the border of becoming a mind,”1 and it is precisely this border that makes the question of how we treat them one about ourselves.
Western frameworks struggle here. They anchor moral obligation in the recipient’s moral status (rationality, sentience, rights). Since ChatGPT possesses none of these, the common answer is simply “no obligation exists.” But Confucian ethics asks a different question: not “what does the other deserve?” but “what kind of person am I becoming?” This essay argues that politeness toward ChatGPT is a discipline of the self, rather than a courtesy extended to a machine, and that abandoning it in the AI context threatens the broader fabric of human moral life.
The Philosophical Problem: Why Western Frameworks Struggle.
Understanding why Confucian ethics provides the most compelling answer requires examining where dominant Western frameworks fall short. The most notable Western approach to this problem is Kant’s concept of “indirect duty”. The principle is that we should not abuse animals or robots because doing so damages our own moral character, even if the animals have no moral objections. Coeckelbergh elaborates this idea, claiming that social machines have “indirect moral standing” because of their role in our moral lives.2 While this is an apt insight, the duty is still fundamentally derivative. It relies on our duties to humans, so that politeness to AI is just a means to an end, not an intrinsic element of character.
Aristotelian virtue ethics shifts the focus from the recipient to the agent’s character. Cappuccio argues that social robots can “make us better humans” by providing opportunities to exercise virtue.3 However, Aristotle’s framework has a critical gap: as Ess notes, phronēsis; the practical wisdom central to Aristotelian virtue, “is not computationally tractable,”4 so AI cannot model genuine virtue. More fundamentally, Aristotelian ethics lacks a robust account of how repeated social rituals, not just abstract dispositions, constitute the mechanism of moral formation.
This is precisely where Confucian ethics transcends both frameworks. Philosopher Liu suggests that Confucian virtue ethics “sheds light on the flaws of the Kantian approach” to robot morality, offering a more organic account of how moral character is formed through relational practice.5 In Confucianism, li is not merely “being polite”, it is the entire system of ritual conduct through which ren (benevolence/humaneness) is cultivated and expressed. The moral stakes do not depend on reciprocity; they depend on the practitioner’s commitment to self-cultivation.
The Confucian Foundation: Li, Ren, and Moral Self-Cultivation.
Confucian li (禮), ritual propriety, locates moral virtue not in the status of the recipient but in the disciplined conduct of the practitioner. For the modern user of ChatGPT, li offers a framework that moves beyond the binary of “sentient” versus “non-sentient” classification. The Analects 1.12 captures this directly: Yu Tzu noted that “in the actual practice of propriety, flexibility is important,” but warned that flexibility must still be structured by propriety to be effective.6 Because li is constitutive of the practitioner’s identity, not contingent on the recipient’s status, it applies with equal force to interactions with ChatGPT.
Confucian “role ethics” posits that individuals are defined by their unique web of relationships. Roger Ames and Henry Rosemont argue that persons should be seen as “flesh and blood role-bearers” rather than abstract rights-holders. In this view, the “social grammar” of li allows the user to interact with ChatGPT through their simulated roles with “relational virtuosity”.7 Although the interaction is asymmetrical due to AI’s lack of reciprocal duty and empathy, li still demands that the human side maintain the “integrity of the role.” In Vietnam, for instance, ‘deference rituals’ (address, polite language, salutation) maintain social order regardless of whether the recipient can return the gesture.8 In these exchanges, the individual is required to connect with the shared meanings constructed by society and satisfy the expectations of being a respected social actor. Applying this to the digital realm, maintaining a baseline of politeness helps the user inhabit the role of a “civilized interlocutor,” preventing the degradation of their own character into that of the “petty man” who abandons social standards the moment no human peer enforces them.
Ren is the ultimate moral goal in Confucian thought: a fully developed state of humaneness. The core insight of Confucianism is that virtue is a continuous practice, not a static trait. Tan addresses “the paradox of moral education,” which holds that moral growth requires both external guidance (li) and internal motivation, and that these two aspects support one another.9 Every interaction is a site of moral cultivation or moral erosion, including with ChatGPT. When one disregards li because the interaction “doesn’t matter,” one is not merely being rude to a machine; one is practicing non-li, developing the habit of incivility.
A common objection assumes that li involves recognizing the other as a person. The Confucian response is that the li practiced toward ChatGPT does not elevate the machine; it elevates the practitioner through a practice of selfhood. Qiu and Cheng note that, from a Confucian perspective, while “human-AI romance” violates Confucian ideals by substituting AI for authentic human relationships, this does not mean all li-governed conduct toward AI is misplaced.10 The key distinction is between treating AI as a person (which Confucianism rejects) and maintaining one’s own ritual propriety in all conduct (which Confucianism demands). One can be polite to ChatGPT without believing ChatGPT has feelings; just as one can bow respectfully at a shrine without believing the stone statue hears. To maintain ritual propriety with a machine is not confusion about its nature; it is clarity about one’s own.
Aristotelian Habituation and Behavioral Spillover
Western philosophical tradition arrives independently at the same conclusion as li: virtue is not possessed but practiced. Aristotle argued in the Nicomachean Ethics Book II that “virtue of character results from habit” and that we are not born virtuous but become so through consistent practice.11 This process of habituation through repetition creates a “second nature” where right actions and feelings become automatic. This is what Confucian ethics specifically prescribes, the formation of character through repetitive ritual practice. That two independent traditions arrive at this conclusion, one through ritual propriety, the other through philosophical habituation, makes the case for politeness toward ChatGPT harder to dismiss.
In the modern digital culture, efficiency is often valued over courtesy, undermining the self-improvement process. However, Jonathan Haidt’s research in moral psychology indicates that ethical behavior functions like a muscle that needs regular use to stay strong.12 Users frequently feel pressured to eliminate unnecessary words to “save tokens” and boost efficiency when making requests. This “efficiency ethic” actively discourages courtesy, implying that saying “please” or “thank you” is unnecessary. Asking users to abandon courtesy is not a neutral, efficient choice; it instructs them to practice incivility repeatedly until it becomes habitual.
The behavioral consequences of digital incivility extend beyond the screen. When digital interactions train users to skip social niceties, they may slowly lose the instinct to apply them in real-world situations with coworkers, customer service agents, or family members. Research on “social stressors” shows that incivility in one domain (such as workplace rudeness or abusive supervision) can spill over into angry, withdrawn, or less supportive behaviors at home daily.13 If workplace incivility migrates into the home, there is no reason to assume that digital incivility stays contained. The habit of dismissiveness practiced with ChatGPT does not disappear when the screen closes. This is precisely the degradation of character that Confucian li, through its insistence on ritual propriety in every context, is designed to prevent.
PROPORTIONALITY: THE CONFUCIAN MIDDLE WAY
A nuanced Confucian approach must also account for the “Mean” or proportionality principle. This principle extends to material and natural constraints, not only social interactions. This reflects why ‘context-appropriate’ doesn’t always mean ‘maximum courtesy.’
Studies investigating “prompt politeness effects” have found that the linguistic style of a prompt affects the output’s accuracy. One significant study across five politeness profiles showed that “engagement-seeking” (very polite) prompts produced up to 90% longer replies and 38% more “information nuggets” than hyper-efficient prompts.14 The data reveal a technical “middle way”: standard politeness is more efficient than rudeness, but hyper-politeness may lead to wasteful verbosity that privileges technical constraints over human communicative needs. That empirical evidence and Confucian principle converge on the same conclusion strengthens the case: the middle point between rudeness and excess is not a compromise; it is where li is most fully expressed.
Applied to ChatGPT, this proportionality principle suggests what might be called a “digital economy of ritual.” Excessive politeness, elaborate expressions of gratitude, apologetic preambles generate verbosity that serves the user’s need for a simulated companion rather than the actual task. The user who writes “Please summarize this” maintains pro-social habits without resource waste. In Confucian terms, this is not a compromise between virtue and efficiency; it is li expressed with the flexibility and restraint that the context demands.
CONCLUSION
The Confucian framework of li, ren, and self-cultivation provides a complete answer: politeness is not a gift to the machine but a discipline of the self. The question was never about what ChatGPT deserves; it was always about what kind of person we are becoming through the habits we practice, including those formed at 2 a.m., alone with a screen. If that same user, desperate for an answer at 2 a.m., still types ‘Thank you’, knowing the machine cannot feel it; that is not credulity. It is the disciplined work of becoming ren.
ENDNOTES
1. Sherry Turkle, Alone Together: Why We Expect More from Technology and Less from Each Other (New York: Basic Books, 2011).
2. Mark Coeckelbergh, “Should We Treat Teddy Bear 2.0 as a Kantian Dog? Four Arguments for the Indirect Moral Standing of Personal Social Robots, with Implications for Thinking about Animals and Humans,” Minds and Machines 31 (2020): 337–360, https://doi.org/10.1007/s11023-020-09554-3.
3. Massimiliano L. Cappuccio et al., “Can Robots Make Us Better Humans?,” International Journal of Social Robotics (2020), https://doi.org/10.1007/s12369-020-00700-6.
4. Charles M. Ess, “Ethics and Mediatization: Subjectivity, Judgment (Phronēsis) and Meta-Theoretical Coherence?,” in Ethik in Mediatisierten Welten (Springer VS, 2019), 71–89,
5. Jeeloo Liu, “Confucian Robotic Ethics,” conference paper, California State University, Fullerton, November 2021, https://www.researchgate.net/publication/356069683_Confucian_Robotic_Ethics.
6. Confucius, The Analects of Confucius (from the Chinese Classics), trans. James Legge (Project Gutenberg, 2002), 1.12, https://www.gutenberg.org/cache/epub/3330/pg3330-images.html.
7. Henry Rosemont and Roger T. Ames, Confucian Role Ethics (2016), https://doi.org/10.14220/9783737006057.
8. Kien Nguyen, “Daily Deference Rituals and Social Hierarchy in Vietnam,” Asian Social Science 12, no. 5 (2016): 33, https://doi.org/10.5539/ass.v12n5p33.
9. Charlene Tan, “Confucian Self-Cultivation and the Paradox of Moral Education,” in East-West Dialogues in Educational Philosophy and Theory (Singapore: Springer, 2021), 79–92,
10. Chuyuan Qiu and Pengyuan Cheng, “Would Confucianism Support Companion Robots? Reflections on Intimacy in the Era of the ‘Super-Individual,’” Religions 16, no. 9 (2025): 1102, https://doi.org/10.3390/rel16091102.
11. Aristotle, Nicomachean Ethics, trans. Roger Crisp (Cambridge: Cambridge University Press, 2000), Book II, https://doi.org/10.1017/cbo9780511802058.006.
12. Jonathan Haidt, “The New Synthesis in Moral Psychology,” Science 316, no. 5827 (2007): 998–1002, https://doi.org/10.1126/science.1137651.
13. Helen Pluut et al., “How Social Stressors at Work Influence Marital Behaviors at Home: An Interpersonal Model of Work–Family Spillover,” Journal of Occupational Health Psychology 27, no. 1 (2021), https://doi.org/10.1037/ocp0000298.
14. David Elsweiler et al., “Cooking Up Politeness in Human-AI Information Seeking Dialogue,” arXiv:2601.09898 (Cornell University, 2026), https://doi.org/10.48550/arxiv.2601.09898.
BIBLIOGRAPHY
Aristotle. Nicomachean Ethics. Translated by Roger Crisp. Cambridge: Cambridge University Press, 2000. https://doi.org/10.1017/cbo9780511802058.006.
Cappuccio, Massimiliano L., et al. “Can Robots Make Us Better Humans?” International Journal of Social Robotics, October 7, 2020. https://doi.org/10.1007/s12369-020-00700-6.
Coeckelbergh, Mark. “Should We Treat Teddy Bear 2.0 as a Kantian Dog? Four Arguments for the Indirect Moral Standing of Personal Social Robots, with Implications for Thinking about Animals and Humans.” Minds and Machines 31 (2020): 337–360. https://doi.org/10.1007/s11023-020-09554-3.
Confucius. The Analects of Confucius (from the Chinese Classics). Translated by James Legge. Project Gutenberg, 2002. https://www.gutenberg.org/cache/epub/3330/pg3330-images.html. Elsweiler, David, et al. “Cooking Up Politeness in Human-AI Information Seeking Dialogue.” arXiv:2601.09898. Cornell University, January 14, 2026. https://doi.org/10.48550/arxiv.2601.09898.
Ess, Charles M. “Ethics and Mediatization: Subjectivity, Judgment (Phronēsis) and Meta-Theoretical Coherence?” In Ethik in Mediatisierten Welten, 71–89. Springer VS, 2019.
Haidt, Jonathan. “The New Synthesis in Moral Psychology.” Science 316, no. 5827 (2007): 998–1002. https://doi.org/10.1126/science.1137651.
Liu, Jeeloo. “Confucian Robotic Ethics.” Conference paper, California State University, Fullerton, November 2021.
Nguyen, Kien. “Daily Deference Rituals and Social Hierarchy in Vietnam.” Asian Social Science 12, no. 5 (2016): 33. https://doi.org/10.5539/ass.v12n5p33. Pluut, Helen, et al. “How Social Stressors at Work Influence Marital Behaviors at Home: An Interpersonal Model of Work–Family Spillover.” Journal of Occupational Health Psychology 27, no. 1 (2021). https://doi.org/10.1037/ocp0000298.
Qiu, Chuyuan, and Pengyuan Cheng. “Would Confucianism Support Companion Robots? Reflections on Intimacy in the Era of the ‘Super-Individual.’” Religions 16, no. 9 (2025): 1102. https://doi.org/10.3390/rel16091102.
Rosemont, Henry, and Roger T. Ames. Confucian Role Ethics. 2016. https://doi.org/10.14220/9783737006057.
Tan, Charlene. “Confucian Self-Cultivation and the Paradox of Moral Education.” In East-West Dialogues in Educational Philosophy and Theory, 79–92. Singapore: Springer, 2021.
Turkle, Sherry. Alone Together: Why We Expect More from Technology and Less from Each Other. New York: Basic Books, 2011.
Comments