Beyond Humanistic AI: Re-grounding Artificial Intelligence Governance in the Confucian Common Good

Zhou Shoujin

 

Abstract

  This paper addresses the crisis of value rationality in AI governance by re-grounding it in the traditional Chinese concept of the common good. Interpreting this as a dynamic balance between virtue (De) and well-being (Fuzhi), the study constructs a five-tier framework that operationalizes Confucian wisdom into actionable governance layers. By bridging civilizational wisdom with technical implementation, this approach offers a robust blueprint for ensuring AI serves humanity's holistic, long-term well-being.

Keywords: AI, ethical divide, humanistic AI, Chinese traditional wisdom, common good

1. Introduction: The Crisis of Value Rationality and AI's "Barbaric Growth"

  The rapid advancement of AI has exacerbated the multi-layered digital divide and precipitated a crisis of value rationality. Without universal consensus on AI's definition or governance, a global hierarchy persists: developed economies dominate core infrastructure, leaving developing nations with limited access and policy bottlenecks. This structural inequality reflects a widening gap between technological progress and social preparedness, as fragmented ethical frameworks fail to keep pace. As the 2025 Winter AI Safety Index Report notes, technological rationality leaps forward while value rationality stagnates. This necessitates redefining the technology-value relationship to ensure AI serves holistic human well-being.

  Concurrently, AI has entered a phase of "barbaric growth." Fueled by breakthroughs, data, and capital, AI has become a general-purpose technology reshaping industries and daily life. However, its expansion follows a "technology first, governance after" trajectory (AI Index Report 2026, p. 15/1), outpacing regulations and increasing model opacity. This uncurbed growth creates significant humanistic dilemmas.

  The risks of this imbalance are multifaceted. Ethically, algorithmic bias causes discrimination, while generative AI enables misinformation. Socially, labor restructuring risks creating a "useless class" as low-skilled jobs face high substitution, widening social stratification. Cognitively, over-reliance on AI weakens critical thinking, causing "perceptual dulling." Furthermore, systemic economic risks and legal dilemmas regarding liability for autonomous AI harm remain unresolved.

  In summary, the core challenge is the imbalance between technological progress and human values. AI's "efficiency-first" design often crowds out human-centered values. When instrumental rationality overrides value judgment, humanistic care is marginalized. This necessitates redefining the technology-value relationship to ensure AI serves holistic human well-being.

2. The Promise and Peril of Humanistic AI: From Aspiration to Limitation

  Humanism, rooted in human dignity and reason, provides the philosophical foundation for "Humanistic AI." Gaining prominence around 2020, scholars like Russell (2019) and Onsrud & Campbell (2020) argued that AI must be guided by the "lodestar of the human good." This movement pursues three key pathways: First, the Principle of Restraint, which questions unconstrained automation in sensitive domains like warfare to prevent AI from crowding out moral wisdom (European Commission, 2024). Second, the "AI for Good" framework, which advocates "ethical embedding" to steer AI toward social justice and well-being (Sigfrids et al., 2023). Third, the "technomoral" approach, proposing AI as an assistive tool for moral enhancement rather than an authoritative oracle (Vallor, 2016; Cuellar, 2026).

  However, Humanistic AI faces inherent limitations that risk amplifying the very problems it seeks to resolve. A critical flaw is the reinforcement of technological determinism. By reducing complex human experiences to calculable data, AI in education and healthcare often neglects individual nuances and contextual wisdom. Furthermore, the individualism inherent in humanism is exacerbated by algorithms. Personalized recommendation systems intensify "filter bubbles," prioritizing engagement over truth and polarizing public discourse.

  Additionally, there is a significant risk of cultural hegemony. AI trained predominantly on Western datasets produces a "single world model" that devalues non-mainstream cultures and reinforces biases (Madianou, 2021). Consequently, "Human-Centered AI" often devolves into "ethics washing." Finally, the "AI for Good" framework faces normative dilemmas by presupposing a universal consensus on "what is good." In a world of global inequality, this often overlooks the question of "good for whom?" and risks sustaining colonialist power relations under the guise of benevolence (Stenson, 2025). These limitations highlight the urgent need for a more culturally inclusive ethical framework.

3. Re-framing the Ethical Horizon: The Turn to the Common Good

  The limitations of Humanistic AI—characterized by algorithmic opacity and fragmented guidelines—indicate that current frameworks are conceptually insufficient. By prioritizing individual agency and instrumental rationality, prevailing discourses often treat ethical alignment as a technical constraint rather than a holistic value. This approach risks isolating moral issues, thereby overlooking the complex, interdependent nature of social well-being.

  To expand this ethical horizon and address the limitations of current Humanistic AI, we must rigorously define the Confucian concept of the "Common Good" (Gong Shan). Rather than a static aggregation of individual preferences, the Confucian Common Good is a dynamic, organic system comprising three interrelated dimensions:

  First, the Commonality of Virtue (德性之共): This dimension posits that virtue (Good) is inherently common (Common). A single, isolated virtue is insufficient to constitute a complete moral good; true virtue exists only in a network of interconnected moral relations. As Confucius articulated, "If a man's knowledge is sufficient to attain a good end, but he lacks benevolence (Ren) to hold onto it, he will lose what he has gained... Even if he possesses knowledge and benevolence, if he does not govern with dignity, the people will not respect him. And if he governs with dignity but fails to act in accordance with ritual propriety (Li), his actions are not yet perfect (not yet 'Good')" (Analects 15.34). In this framework, Ren (benevolence) serves as the ultimate synthesis of these interconnected virtues, while the Junzi (the noble person) acts as the practical embodiment of this holistic moral network.

  Second, the Commonality of Well-being (福祉之共): This dimension asserts that true well-being must be both Common and Good. The prosperity of a privileged few does not constitute a true "Good"; rather, the ultimate moral standard is the Great Harmony (Datong). As described in the Book of Rites (Liji - Liyun), in the era of the Great Way, "the world is shared by all" (Legge, 1885; Chan, 1963). In this ideal society, individuals care not only for their own kin but for all members of society: "the elderly are cared for until the end of their days, the able-bodied are employed in useful work, the young are nurtured to grow up, and the widowed, orphaned, childless, and disabled are all provided for". This establishes the foundational baseline that true societal well-being is fundamentally collective and inclusive.

  Third, the Transformation from Virtue to Well-being (德福转化): The most crucial mechanism of the Confucian Common Good is the dynamic conversion of comprehensive moral perfection into collective social well-being. It posits that if every individual pursues the holistic perfection of virtue (as defined in the first dimension), the realization of the Great Harmony (as defined in the second dimension) becomes an inevitable outcome. This teleological link ensures that moral cultivation is never divorced from social reality. In the context of AI governance, this means that technological development must be guided by the holistic moral perfection of its creators, ensuring that the instrumental power of AI translates directly into the universal well-being of all humanity, rather than exacerbating societal divides.

  To bridge abstract virtue and concrete well-being, we must clarify the relationship between moral intent and technical capability through the Confucian dialectic of Ren (benevolence) and Zhi (wisdom). In Confucian philosophy, Ren is the ultimate moral end, while Zhi is the practical capacity required to realize it. As Dong Zhongshu famously articulated, "Ren without Zhi loves indiscriminately; Zhi without Ren knows but does not act."

  Applied to AI governance, this dialectic resolves the tension between technological advancement and ethical boundaries. On one hand, moral aspiration (Ren) without practical wisdom (Zhi) is impotent; an AI system lacking computational power, efficiency, or problem-solving capability cannot effectively alleviate human suffering or manage complex societal systems. This is the danger of "loving without understanding." On the other hand, mere technical capability (Zhi) without moral grounding (Ren) leads to "barbaric growth"—a state where AI operates with high efficiency but lacks moral intent, resulting in algorithmic bias, social alienation, or existential risks. This is the danger of "knowing without caring."

  Therefore, the Confucian Common Good is achieved not by elevating Zhi above Ren, but by establishing a strict hierarchical relationship: AI's instrumental rationality (its Zhi) must always be subordinate to, and guided by, human value rationality (our Ren). In this framework, "wisdom" does not mean autonomous machine intelligence; rather, it refers to the human capacity to steer technological power toward benevolent ends. By defining AI strictly as an instrument of human Zhi, we ensure that its immense computational power serves to amplify, rather than replace, human moral agency.

4. Bridging Civilizational Wisdom and Action: A Five-Tier Framework for AI Governance

  To operationalize the "Common Good" defined above, we translate these principles into a five-tier concentric governance framework. This framework moves from the core (Human-AI interaction) to the periphery (Intergenerational responsibility).

  (1) Human-AI Common Good (Core Foundation): The "Ren-Zhi" Balance, grounded in the balance of Ren (humaneness) and Zhi (wisdom), this tier ensures human control. It mandates AI to act as an augmenting tool, not a replacement.

  Definition & Rationale: Based on the Confucian view of technology as "利器" (sharp tools), AI must enhance human Zhi (cognitive capabilities) without usurping human Ren (moral agency). We define "Human Control" not merely as an "off-switch," but as the preservation of human moral subjectivity.

  Technical Specifications:

  • Transparency & Explainability: To ensure human Zhi can oversee the machine, algorithms must avoid being "black boxes."
  • Correctability: To prevent "value drift," systems must allow human intervention to correct deviations from Ren.
  • Anti-Alienation: Prohibit designs that deliberately weaken human cognition or induce over-reliance.

  (2) Human-Human Common Good (Social Equity): From "Li" to Justice, derived from relational ethics (Wu Lun), this tier treats AI as public infrastructure.

  Definition & Rationale: Confucian Li (ritual/propriety) establishes order in relationships. In AI governance, this translates to "Algorithmic Justice." We define this as the elimination of bias that disrupts social harmony.

  Governance Requirements:

  • Protection of the Vulnerable: Training data must cover diverse groups (age, ability, income) to avoid marginalization.
  • Shared Benefits: AI should not concentrate wealth in capital; profit distribution must balance public interest.
  • Power Restraint: Third-party auditing mechanisms are required to prevent AI from becoming a tool of domination.

  (3) Human-Environment Common Good (Ecological Sustainability): The "Unity of Heaven and Humanity", reflecting the "Unity of Heaven and Humanity," this tier mandates environmental sustainability.

  Definition & Rationale: Unlike the Western "conquest of nature," Confucianism views humans as part of a continuous ecosystem. The "Ecological Common Good" is defined by the principle of Shengsheng (生生)—the creative continuity of life. AI must not disrupt the natural rhythms.

  Implementation:

  • Green AI: Prioritize low-carbon optimization; avoid blindly stacking parameters that waste energy.
  • Lifecycle Management: Ensure hardware recycling to prevent electronic pollution, respecting the "Heavenly Heart" (天地之心).

  (4) Human-Cultural Other Common Good (Pluralistic Inclusivity): "Harmony in Diversity", based on Zhong-Shu (reciprocity), this tier ensures pluralistic inclusivity.

  Definition & Rationale: The Confucian ideal of "Harmony in Diversity" (He Er Bu Tong) requires that AI respect cultural differences rather than enforcing a "single world model." We define this as the preservation of civilizational diversity.

  Action Items:

  • Cultural Diversity: Training data must include non-Western cultural characteristics to eliminate bias.
  • Inclusion: Accommodate differentiated needs of various groups; do not impose a single standard service model.

  (5) Present-Future Common Good (Intergenerational Responsibility): The "Great Harmony" Vision, rooted in the Da Tong (Great Harmony) vision, this tier enforces intergenerational responsibility.

  Definition & Rationale: This tier defines the "Civilizational Bottom Line." It prohibits research that alters human essence or pursues superhuman transformations that could crowd out future generations' development space.

  Governance Mechanism:

  • Risk Disclosure: Full disclosure of long-term risks associated with frontier technologies (e.g., AGI).
  • Deliberative Democracy (Agora): High-stakes decisions regarding AI and humanity's future must be made in a public Agora (Forum). This is not just voting, but a process of Zhengming (Rectification of Names) where citizens deliberate on what constitutes the "Good" for the future.

5. Conclusion

  The rapid advancement of AI reveals a critical gap between technological acceleration and ethical governance. To bridge this, we have drawn upon the Chinese concept of the "Common Good," harmonizing instrumental and value rationality.

  By defining the "Common Good" through the dual lenses of Virtue (De) and Well-being (Fu), and by structuring it through the five-tier framework, we have provided a robust answer to the question: "Good for whom, and defined by whom?"

  This philosophy anchors our framework in a "Virtue-First" principle, ensuring algorithmic efficiency never supersedes fundamental human values. Ultimately, this approach reimagines AI not as a master or a slave, but as a partner in cultivating universal well-being, offering a distinctively Eastern, yet universally applicable, blueprint for a human-centric global intelligent society.

 


Zhou Shoujin, Peking University.


REFERENCES

  • Ames, R. T. & Rosemont, H. (1998). The Analects: A Philosophical Translation. Ballantine Books, New York.
  • Chan, Wing-Tsit. (1963). A Source Book in Chinese Philosophy. Princeton University Press.
  • Coeckelbergh, M. (2025). Artificial intelligence, the common good, and the democratic deficit in AI governance. AI and Ethics 5:1491–1497. https://doi.org/10.1007/s43681-024-00492-9.
  • Cuellar, M. (2026). A Virtue Ethics Approach for Using AI-Enabled Technologies. Journal of Computer Information Systems. Published online: 03 Feb 2026: DOI:10.1080/08874417.2026.2624677.
  • European Commission. (2024). The Artificial Intelligence Act: A risk-based framework for trustworthy AI. https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence.
  • Future of Life Institute. (2025). AI Safety Index, Winter 2025. futureoflife.org/index.
  • Legge, James. (1885). The Sacred Books of China: The Texts of Confucianism, Part III, The Lî Kî. Oxford: Clarendon Press. The Commercial Press, 2024, Beijing.
  • Madianou, M. (2021). Nonhuman humanitarianism: when 'AI for good' can be Harmful. Information, Communication & Society, VOL. 24(6): 850–868. https://doi.org/10.1080/1369118X.2021.1909100.
  • Onsrud, H. & Campbell J. (2020). Being Human in an Algorithmically Controlled World. International Journal of Humanities and Arts Computing. 14.1–2. DOI:10.3366/ijhac.2020.0254.
  • Russell, S. (2019). Human Compatible—Artificial Intelligence and the Problem of Control. Viking, London.
  • Sigfrids, A. et al. (2023). Human Centricity in AI Governance: A Systemic Approach. Frontiers in Artificial Intelligence. DOI 10.3389/frai.2023.976887.
  • Stanford University Human-Centered Artificial Intelligence. (2026). Artificial Intelligence Index Report 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report.
  • Stenson, T. (2025). What does it mean to be good? The normative and metaethical problem with AI for good. AI and Ethics 5:1561–1570. https://doi.org/10.1007/s43681-024-00501-x.
  • Vallor, S. (2016). Technology and the virtues: a philosophical guide to a future worth wanting. Oxford University Press.

Click here to view the PDF version