The Holistic Reconstruction of Ecological Civilization in the Age of Artificial Intelligence

In the current rapid development of artificial intelligence, the construction of ecological civilization has entered an entirely new technological context. The evolution of technology signifies not merely the advancement of the means of production but fundamentally drives the transformation of the mode of production, triggering a restructuring of social organizational structures, forms of interaction, and value concepts, thereby profoundly influencing the transformation of civilizational forms. Current academic research on artificial intelligence and ecological civilization mostly focuses on dimensions such as technological empowerment and digital transformation, emphasizing the role of AI as a tool in optimizing the ecological governance system, with little in-depth exploration of its impact on ecological civilization itself or its significance as a structural force of civilization. From the perspective of historical materialism, civilization is a phenomenon unique to human society; it is the sum total of the material and spiritual achievements formed by humans, as subjects, transforming the natural world with the aid of production tools. When AI, with its algorithmic logic and automatic generation mechanisms, deeply participates in the interactive relationship between humans and nature, the original "subject–object" and "ethics–behavior" schemas within the category of ecological civilization also face the possibility of being restructured. Therefore, this article will attempt to elucidate that, in the age of artificial intelligence, the construction of ecological civilization requires not only updating technological pathways but also re-examining the basic structure of ecological civilization. This is a practical requirement for the continuous development of ecological civilization theory in responding to the profound adjustments in the relationship between humans and nature under new technological conditions.

在人工智能快速发展的当下，生态文明建设进入一个全新的技术语境。技术的演进不仅意味着生产工具的进步，更是从根本上推动生产方式的变革，引发社会组织结构、交往形式以及价值理念的重构，进而深刻影响文明形态的变迁。目前学界关于人工智能与生态文明的研究，大多聚焦于技术赋能、数字转型等维度，强调人工智能作为工具在优化生态治理体系中的作用，鲜有深入探讨其对生态文明本身的影响及其作为文明结构性力量的意义。在历史唯物主义看来，文明是人类社会所特有的现象，是人作为主体借助生产工具改造自然世界所形成的物质成果和精神成果的总和。当人工智能以其算法逻辑和自动生成机制深度参与人与自然的互动关系时，原有的生态文明范畴中的“主体—客体”“伦理—行为”图式也面临被重构的可能。因此，本文将试图阐明在人工智能时代，生态文明建设不仅需要更新技术路径，更需要对生态文明的基本结构进行再审视。这是生态文明理论在新技术条件下回应人与自然关系的深刻调整而不断发展的现实要求。

I. Human–Machine Collaboration Shapes a New Pattern of Subjects in Ecological Civilization Construction

一、人机协同塑造生态文明建设主体的新格局

Generally speaking, the core of ecological civilization is correctly handling the relationship between humans and nature, with technology regarded as a tool serving ecological governance, merely a material means or intermediary for humanized nature. However, artificial intelligence, especially strong AI, is an "intelligent agent" [1]1 "Intelligent agent" (智能体) is a term in AI research referring to an entity that perceives its environment and takes actions to achieve goals, often used to describe autonomous AI systems. capable of learning, judgment, and even generating multiple solutions, interacting dynamically with nature in the form of a quasi-subject. This significant change drives the original binary "human–nature" relationship to evolve into a ternary interactive structure of "human–intelligent agent–nature," marking the entry of ecological civilization construction into a new stage of intelligent participation and human–machine collaboration.

通常而言，生态文明的核心是正确处理人与自然的关系，技术被视为服务于生态治理的工具，只是人化自然的物质手段或者中介。然而，人工智能特别是强人工智能，是具备学习、判断乃至多元方案生成能力的“智能体”,以拟主体的形态与自然产生动态的交互关系。这个重大变化推动原有的“人—自然”二元关系演化为“人—智能体—自然”的三元互动结构，标志着生态文明建设进入智能参与、人机协同的新阶段。

(1) The Interactive Mode of "Quasi-Subjective" Artificial Intelligence with Nature

(一)“拟主体性”人工智能与自然的交互方式

Whether artificial intelligence possesses subjectivity has been a hotly debated topic in academic circles in recent years. Some scholars argue that AI does not possess genuine subjectivity because it lacks the self-consciousness, free will, capacity for constructing social relations, and moral responsibility necessary for human subjects. However, the widespread application of generative AI and its performance surpassing human capabilities in many fields have prompted an increasing number of scholars to re-examine the connotation and boundaries of subjectivity. For example, some scholars propose that, according to the three basic requirements of autonomous self-consciousness, free will, and moral responsibility, AI gradually exhibits characteristics of weak subjectivity and, upon meeting certain criteria, can be regarded as a subject within limited domains. Other scholars advocate redefining subjectivity from a "subject–interaction" perspective: "Subjectivity can be seen as the potential action capacity of entities participating in interaction, determined by the driving factors of interaction." The subjectivity of generative AI is defined by the interactive capacity driven by language generation. This interactive subjectivity is no longer based on traditional philosophy's presupposition of the subject's intrinsic attributes (such as intentionality) but uses the goal-directedness and autonomy displayed by entities in interaction as criteria, reflecting a completely new form of artificial subjectivity. In short, regardless of one's stance on AI, the issue of its subjectivity has become a major topic that the humanities and social sciences must confront directly. Humanity needs to carefully assess the paradigm shift in subjectivity that intelligent technology may trigger while maintaining its own subjectivity.

人工智能是否具备主体性，是近年来学界广泛讨论的热点议题。有学者认为，人工智能不具备真正主体性，原因是欠缺人类主体所必需的自我意识、意志自由、社会关系建构能力和道德责任能力。然而，生成式人工智能的广泛应用及其在诸多领域展现出超越人类能力的表现，促使越来越多的学者重新审视主体性的内涵和边界。例如，有学者提出，按照自主性的自我意识、自由意志和道德责任这三个基本要件，人工智能逐步呈现弱主体性特征，在达到某些标准时，可以在限定领域被视为主体。还有学者主张从“主体-交互”的视角重新定义主体性，“主体性可视为参与交互实体的潜在行动能力，由交互的驱动力因素所决定”,生成式人工智能的主体性是由语言生成驱动的交互能力所定义，这种交互主体性不再基于传统哲学对主体内在属性(如意向性)的预设，而是以实体在交互中展现的目标导向性和自主性为判据，体现出一种全新的人工主体性形式。总之，无论人们对人工智能持何种立场，其主体性问题已成为人文社会科学必须直面的重大课题，人类需要在保持自身主体性的同时，审慎评估智能技术可能引发的主体性范式变革。

Although artificial intelligence does not yet possess organic life, through data-driven learning and generation mechanisms, it fundamentally transcends general technology and exhibits functional characteristics similar to a subject. Specifically, it can actively capture, extract, and deeply process information, rather than merely displaying phenomena; it can continuously self-optimize and self-improve in a goal-oriented manner during learning, rather than mechanically executing instructions linearly according to pre-set rules; it can automatically and flexibly respond to environmental changes based on information feedback, rather than only unidirectionally influencing objects under human instruction; and it can generate entirely new semantic combinations, such as text and images, through the probability distribution of existing information, rather than merely copying or reproducing existing materials. Although these processes and mechanisms are fundamentally different from the conscious mental activities of human subjects, in terms of behavioral patterns, they highly simulate the practical activities of humans actively adapting to, reflecting on, and transforming the external environment. In view of this, we should actively utilize this quasi-subjective characteristic of AI, grant it corresponding actor status, and incorporate it into extended reflections on the relationship between humans and nature. This provides an explanatory pathway for understanding the interactive relationship between AI and nature, as well as its position and role in the ecosystem, to collaboratively construct a higher-level ecological civilization.

人工智能虽然尚不具备有机生命，但通过数据驱动的学习和生成机制，从根本上超越了一般性技术，展现出近似主体的功能特征。具体而言，它能够主动捕捉、提取并深层处理信息，而不只是展示现象；能够在学习中以目标为导向持续进行自我优化和自我改进，而不是事先设定规则后只能机械地线性执行指令；能够依据信息反馈自动而灵活地响应环境变化，而不只是接受人的指令单向度地对客体施加影响；能够通过既有信息的概率分布，生成文字、图像等全新的语义组合，而不只是复制或再现已有材料。尽管这些过程和机制在本质上有别于人类主体的能动性意识活动，但在行为模式上高度模拟了人类主动适应、反映和改造外部环境的实践活动。鉴于此，我们应积极利用人工智能的这种拟主体性特征，赋予其相应的行动者地位，将其纳入对人与自然关系的延伸性思考中，从而为理解人工智能与自然之间的交互关系及其在生态系统中的地位与作用提供解释路径，协同构建更高层次的生态文明。

At the ontological level, natural information constitutes the object domain for the generation of AI's learning capabilities. The name "artificial intelligence" originates from the idea of "using machines to imitate human learning and other aspects of intelligence," including knowledge representation, automated reasoning, and even autonomously solving complex tasks. Human thinking ability and spiritual production require nature to provide a realistic premise. As Marx pointed out, nature is "a part of human consciousness" and "the inorganic body of the human spirit." [2]2 Karl Marx, *Economic and Philosophic Manuscripts of 1844* (Moscow: Progress Publishers, 1959), p. 75. Like humans, nature constitutes an important field for the development of AI. The dynamic variability and spatial heterogeneity of the natural environment provide rich training data for AI: "The use of massive data can avoid model overfitting, giving large models good generalization ability, i.e., out-of-sample prediction ability." [3]3 This is a quotation from a Chinese academic source on AI model generalization. Harsh extreme environments, in particular, prompt AI to adopt adaptive training strategies such as reinforcement learning. Through repeated trial and error and feedback correction, innovations are made in data acquisition, model training, and algorithm optimization, thereby forming more precise modeling and generation capabilities for the movement processes and future evolutionary trends of nature. For example, the extreme temperature differences and sandstorms in deserts impose requirements on the stability of equipment and algorithms, driving the development of control algorithms with fault tolerance to ensure that intelligent agents maintain stable performance during long-term autonomous operation. Moreover, these AI technological improvements fostered in deserts can also be generalized to other complex field environments, enhancing the proactive adaptive capacity of AI systems in diverse natural settings.

就本体层面而言，自然信息是人工智能学习能力生成的对象域。“人工智能”之得名，源起于“用机器来模仿人类学习以及其他方面的智能”的设想，包括知识表示、自动推理，甚至自主解决复杂任务等。人类的思维能力和精神生产需要自然提供现实前提。正如马克思所指出的，自然是“人的意识的一部分”,是“人的精神的无机界”。同人类一样，自然也构成人工智能发展的重要场域。自然环境的动态变化性与空间异质性为人工智能提供了丰富的训练数据，“海量数据的使用可避免模型过度拟合，使大模型具有较好的泛化能力即样本外预测能力”,而严苛的极端环境更是促使人工智能采用强化学习等自适应训练策略，通过反复试错和反馈纠正，在数据获取、模型训练、算法优化等方面创新，从而形成对自然的运动过程及其未来演化趋势更精准的建模与生成能力。比如，沙漠极端温差和沙尘对设备和算法稳定性提出要求，推动了具有容错能力的控制算法的发展，以保证智能体在长时间自主运行中保持性能稳定。不仅如此，这些在沙漠中催生的AI技术改进，亦有望泛化到其他复杂野外环境中，提升人工智能系统在多样自然环境下的主动适应能力。

At the epistemological level, AI can analyze the intrinsic connections between natural information. Through satellite remote sensing, big data analysis, ecological modeling, and other methods, AI automatically conducts efficient and in-depth analysis of natural systems. For instance, it can monitor trends in climate change in real time, precisely analyze data such as temperature, humidity, and air pressure globally or regionally; identify the spatial distribution patterns of species, track the movement paths of endangered animals, and assess the impact of habitat changes on species survival; and analyze fluctuations in water resources, predicting extreme climate events such as droughts and floods. In comparison, although general technology has to some extent expanded the human capacity to obtain natural information, its functional boundaries remain limited to the extended sensory presentation of natural information and cannot yet automatically extract and analyze natural information. Engels pointed out that humans "surpass all other living beings" precisely because they are "able to recognize and correctly apply the laws of nature." [4]4 Friedrich Engels, *Dialectics of Nature* (Moscow: Progress Publishers, 1976), p. 180. So-called laws are the essential, necessary, and stable connections in the movement and development of things. AI's ability to reveal deep, subtle structural correlations among the complex variables in the natural world makes it a "translator" of ecological information, thereby helping humans discover more natural laws in a more efficient and precise manner.

就认识层面而言，人工智能可以分析自然信息之间的内在联系。人工智能通过卫星遥感、大数据分析、生态建模等方式，自动对自然系统进行高效而深入的解析。比如，实时监测气候变化的趋势，精确分析全球或区域内的温度、湿度、气压等数据；识别物种的空间分布格局，追踪濒危动物的活动路径，判断栖息地变化对物种生存的影响；分析水资源的波动，预测干旱、洪涝等极端气候事件的发生。相较而言，一般性技术虽然在一定程度上拓展了人类获取自然信息的能力，但其作用边界仍限于自然信息的延伸性感官呈现，尚且不能对自然信息进行自动提取和分析。恩格斯指出，人“比其他一切生物强”,就在于“能够认识和正确运用自然规律”。所谓规律，就是事物运动发展中本质的、必然的、稳定的联系。人工智能的这种展现自然界中庞杂变量之间深层次的、微妙的结构性关联的能力，使其成为生态信息的“翻译者”,从而帮助人类以更高效、更精准的方式发现更多自然规律。

At the practical level, artificial intelligence can automatically regulate its impact on nature according to specific goals. Although AI does not possess subjective purpose in the human sense, under the guidance of preset goals, through dynamic modeling and parameter optimization of the variable relationships in natural systems, it adaptively exerts influence on the process of natural metabolism. For example, intelligent greenhouse systems use sensors to perceive environmental information such as temperature, humidity, and light, and control irrigation, ventilation, and other devices through algorithmic decision-making to maintain optimal conditions for plant growth. Engels said that other organisms can only “simply by their mere existence bring about changes in nature, whereas man, by the changes he makes, serves his own ends, dominates nature,” and “gradually learns to recognize the more remote, indirect social effects of his productive activity, and so is enabled to control and regulate these effects.” [5]5 Friedrich Engels, “The Part Played by Labour in the Transition from Ape to Man” (1876), in *Dialectics of Nature*. In this sense, AI differs from other organisms, which can only passively adapt to environmental changes and find it difficult to actively regulate natural systems, and also differs from general technology, which relies entirely on step-by-step instructions from human operators. Instead, by means of algorithmic judgment and real-time sensor feedback, it “constructs its own ‘associated environment,’ independently regulating the exchange of energy and matter,” [6]6 Reference to the concept of “associated environment” (milieu associé) from the work of French philosopher of technology Gilbert Simondon. exhibiting behavior similar to human “domination” of nature, and thus becoming a new active factor in the ecosystem.

就实践层面而言，人工智能可以根据特定目标自动调控对自然的影响。人工智能虽然不具有人类意义上的主观目的性，但在预设目标导向下，通过对自然系统变量关系进行动态建模与参数优化，自适应地对自然新陈代谢的过程施加影响。例如，智能温室系统利用传感器感知温湿度、光照等环境信息，通过算法决策控制灌溉、通风等装置，维持植物生长的最佳条件。恩格斯说，其他生物只能“简单地通过自身的存在在自然界中引起变化，而人则通过他所作出的改变来使自然界为自己的目的服务，来支配自然界”,并“渐渐学会了认清我们的生产活动在社会方面的间接的、较远的影响，从而有可能去控制和调节这些影响”。在这个意义上，人工智能既不同于其他生物只能被动适应环境变化而难以主动调控自然系统，也不同于一般性技术完全依赖操作人员的逐步指令，而是借助算法判断与传感器实时反馈，“构建自身的‘缔合环境’,独立调节能量、物质的交换”,呈现出类似人类“支配”自然界的行为，因而成为生态系统中新的能动因素。

In summary, AI, relying on machine deep learning, analyzes natural data, predicts trends in environmental change, and proposes solutions to ecological problems. At the same time, it is continuously “trained” by information from the natural environment, developing higher-order capabilities to address more complex ecological issues. This two-way interactive mechanism lays the foundation for constructing a new type of ecological structure characterized by the ternary symbiosis of “human—intelligence—nature.”

综上所述，人工智能依托机器深度学习，分析自然数据、预测环境变化趋势并提出生态问题的解决方案，同时，也在不断被自然环境的信息“训练”中，发展出应对更加复杂的生态问题的高阶能力。这种双向交互机制为构建“人—智能—自然”三元共生的新型生态结构奠定了基础。

(2) Shaping the Civilizational Leap of Ecological Intelligence and Human-Machine Collaboration

(二)塑造生态智能与人机协同的文明跃迁

Ecological civilization, as a new form of human civilization, marks a historic advance in the relationship between humanity and nature. However, from the perspective of the Marxist view of development, each higher form itself still contains the possibility and space for further evolution, and differences in the level and quality of development will also exist within society. This means that we must continuously promote the development of ecological civilization to a higher level in practice. The rise and widespread application of AI, this emerging force, poses new epochal propositions for the construction of ecological civilization. It requires us not only to continue the traditional ecological concept of respecting and protecting nature but also to promote the embodiment of an ecological orientation in the operational mechanisms of quasi-subjective intelligence. Only in this way can ecological civilization achieve timely reconstruction and development in the era of artificial intelligence.

生态文明作为人类文明的新形态，标志着人与自然关系的历史性进步。然而，从马克思主义发展观来看，每个高级形态本身依然包含着进一步演进的可能和空间，社会内部也会存在发展层次和质量的差别。这意味着，我们必须在实践中不断推动生态文明向更高水平发展。人工智能这一新兴力量的兴起与广泛应用，对生态文明建设提出新的时代命题，不仅需要我们赓续尊重和保护自然的传统生态理念，还应推动拟主体性智能在其运行机制中体现生态导向。如此，生态文明才能在人工智能时代实现与时俱进的重构与发展。

First, enrich the systematic input of natural knowledge to provide the foundation for AI ecological learning. The essence of AI is a “rule-based program,” in which “a large body of knowledge/beliefs is represented as a set of rules linking conditions to actions.” These programs “simulate associative memory, pattern recognition, meaning, language, problem-solving, learning, image and perceptual-motor control by identifying statistical probabilities in the environment.” [7]7 This is a composite description of rule-based AI systems; the specific phrasing is a paraphrase of common definitions in AI literature. This indicates that AI is a system trained on the basis of human linguistic data, and its “intelligence”—manifested in abilities such as learning, judgment, and generation—is a mapping and reorganization of human knowledge systems, value presuppositions, and modes of expression within the technical system. The way humans understand nature, describe nature, and produce knowledge about nature determines the path by which AI will present natural information, influence ecosystems, and intervene in environmental affairs. Therefore, we need to continuously enrich and develop the system of ecological knowledge, especially by integrating interdisciplinary natural knowledge. This is not only to enhance humanity’s own understanding of nature but also to provide an ecological blueprint for AI to learn from and train on.

其一，丰富自然知识的系统输入，提供人工智能生态学习的基础。人工智能本质是一种“基于规则的程序”,“大量知识/信念被表示为一套将条件与动作联系起来”的“规则集”,这些程序“通过识别环境中的统计概率，模拟联想记忆、模式识别、意义、语言、问题求解、学习、图像和感知运动控制”④。这表明，人工智能是一种以人类语料为基础训练而成的系统，其“智能”所表现出的学习、判断、生成等能力，均是人类知识体系、价值预设与表达方式在技术系统中的映射与再组织。人类认识自然、描述自然、生产自然知识的方式，决定了人工智能将以何种路径展现自然信息、影响生态系统和介入环境事务。因此，我们需要不断丰富发展生态知识体系，特别是整合跨学科的自然知识，这不仅是为了提升人类自身对自然的认知，更是为人工智能提供一个可供学习与训练的生态图谱。

Second, establish the goal orientation of ecological intelligence, promoting the universal embodiment of ecological values in all AI systems. Virginia Dignum points out that “regardless of the degree of autonomy, social awareness, and learning ability of the system, AI systems are tools built by people for specific purposes,” and “even if the system develops by learning from interaction with its environment, what it learns depends on the construction purpose and the functions it possesses.” [8]8 Virginia Dignum, *Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way* (Springer, 2019). This shows that although AI possesses the ability to continuously learn, the scope of its learning is always limited by the functional boundaries set by humans. Just as a floor-cleaning robot cannot “learn” to do laundry, and a self-driving car will not “evolve” the ability to fly, AI’s performance in ecological and environmental protection also depends on whether it was endowed with an ecological protection-oriented goal at the time of its design. Therefore, shaping ecological intelligence should not be limited to developing specialized AI systems for environmental protection. Instead, ecological goals should be “built into” the design principles, corpus training, and model construction of all AI systems, so that AI simultaneously carries the generation, expression, and realization of ecological values in its diverse functional evolution. This truly constitutes an ecological civilization embodied in the entire process and all aspects of the era of artificial intelligence.

其二，确立生态智能的目标导向，推动生态价值取向在所有人工智能系统中普遍体现。弗吉尼亚·迪格纳姆指出，“无论系统的自主程度、社会意识和学习能力如何，AI系统都是人们为特定目的而构建的工具”,“即使系统已经发展，通过从与环境的交互中学习，它所学到的东西也取决于建构目的和所具备的功能”。这表明，人工智能虽具备持续学习的能力，但其学习范围始终受限于人类设定的功能边界。正如扫地机器人无法“学会”洗衣，自动驾驶汽车不会“进化”出飞行能力一样，人工智能在生态环境保护中的表现，也取决于其设计之初是否被赋予了生态保护的导向性目标。因此，塑造生态智能，不应仅限于研发面向环境保护的专门型人工智能系统，而应在所有人工智能的设计原则、语料训练与模型构建中“内置”生态目标，使人工智能在多元功能演化中同步承载生态价值的生成、表达与实现，这才真正构成人工智能时代体现在全过程和各方面的生态文明。

Third, establish an ecological behavior assessment indicator system to ensure that AI always maintains value alignment. “Value alignment” is an important concept in AI ethical governance proposed by scholars such as Dylan Hadfield-Menell in 2016. It refers to “ensuring that AI systems, in the process of pursuing their specific goals, can conform to the greatest extent possible to the true intentions of human users,” [9]9 Dylan Hadfield-Menell et al., “The Off-Switch Game,” in *Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence* (2017), pp. 220-227. thereby avoiding technical systems outputting content that contradicts social values. The key to achieving this goal lies in the design and implementation of values. “The value design process should ensure that the connection between values and their specific interpretations in system design and operation can be traced and evaluated.” [10]10 This is a paraphrase of a principle in value-sensitive design (VSD) literature, associated with scholars such as Batya Friedman. That is to say, only by transforming value concepts (such as fairness, safety, privacy, etc.) into specific parameter configurations can the algorithmic system of AI establish a common logical context with human value concepts, thereby reflecting and aligning with human value goals during operation. Regarding ecological intelligence, developers must clarify the actual form of ecological values in the technical system, such as quantitative indicators like the amount of carbon emission reduction and the biodiversity impact factor. Based on this, they should continuously examine and evaluate whether the technology benefits the overall health of the ecosystem, establish a dynamic feedback mechanism, form an ecological discipline system oriented towards AI, and promote it to gradually establish behavioral norms for harmonious coexistence with nature in its “learning” process, growing into an ecological intelligent agent capable of automatically avoiding ecological risks, executing environmental protection, and enhancing natural benefits.

其三，建立生态行为评估指标体系，确保人工智能始终保持价值对齐。“价值对齐”(value alignment)是迪伦·哈德菲尔德-梅内尔(Dylan Hadfield-Menell)等学者在2016年提出的人工智能伦理治理的重要概念，它是指“确保人工智能系统在追求其特定目标的过程中，能够最大限度地符合人类用户的真实意图”,从而避免技术系统输出与社会价值观相悖的内容。实现这一目标的关键在于价值的设计与落实，“价值设计过程应确保可以追踪和评估系统设计和运行中价值及其具体解释之间的联系”。也就是说，只有将价值理念(如公平、安全、隐私等)转化为具体的参数配置，人工智能的算法体系才能与人类价值理念建立共同的逻辑语境，从而在运行过程中反映并对齐人类的价值目标。就生态智能而言，开发者必须明确生态价值在技术系统中的实际呈现形式，如碳排放减少量、生物多样性影响因子等量化指标，以此为依据持续考察评估该技术是否有益于生态系统的整体健康，并建立动态反馈机制，形成面向人工智能的生态规训体系，促使其在“学习”过程中逐步建立与自然和谐共处的行为准则，成长为能够自动规避生态风险、执行环境保护和提升自然效益的生态智能体。

II. Algorithmic Characteristics Raise New Requirements for Technological Ecological Ethics

二、算法特性提出技术生态伦理的新要求

The consideration of ecological issues has always been closely related to technological ethics. Modern ecological civilization has gradually developed precisely through a profound reflection on the technological ethics of industry. The behavioral consequences of traditional technology can be fully attributed to the direct operation of humans, so ethical norms mainly focus on “how humans use technology.” In contrast, AI exhibits a certain “decision-making ability,” but due to its inherent drawbacks such as algorithmic bias, AI hallucination, and systemic high energy consumption, it cannot bear responsibility at the moral and legal levels. This raises higher requirements for the technological ethics of AI in the construction of ecological civilization. It must shift from mere usage norms to proactive intervention, emphasizing “how humans design and supervise intelligent agents,” and implementing the ecological responsibilities that humans bear in developing, manufacturing, and using AI systems from the source.

生态问题的考量向来与技术伦理密切相关，现代生态文明正是在对工业技术伦理的深刻反思中逐步发展起来的。传统技术的行为后果可以完全归因于人的直接操作，因而伦理规范主要集中于“人如何使用技术”。相比之下，人工智能展现出一定的“决策能力”,但由于其存在算法偏向、AI幻觉、系统性高能耗等内在弊端，无法在道德和法律层面上承担责任，从而使得生态文明建设对人工智能的技术伦理提出了更高要求，即必须从单纯的使用规范转向前置性介入，强调“人如何设计和监督智能代理”,从源头上落实人们在开发、制造和使用人工智能系统时所承担的生态责任。

(1) Algorithmic Bias and the Establishment of an Attention Allocation Mechanism for Ecological Justice

(一)算法偏向与建立生态正义的注意力分配机制

Ecological civilization, at its deep structure, is a question of “how ecological justice is to be constructed.” Only by achieving a fair distribution of ecological interests among different species, regions, and human groups, and by realizing comprehensive justice in the relations between humans and between humans and nature, can we speak of a true ecological civilization. While artificial intelligence enhances the efficiency of ecological governance, its algorithmic biases are quietly eroding ecological justice, causing environmental governance decisions to be formally reasonable but substantively capable of exacerbating ecological inequality. So-called algorithmic bias, also known as algorithmic prejudice or algorithmic discrimination, refers to the phenomenon in which, during the processes of algorithm design, data collection, model training, and practical application, the subjective assumptions of developers, limitations in data sources, and other factors cause the system to discriminate against or neglect specific groups and regions in its output. Currently, many ecological models and monitoring systems rely during the algorithm training phase on data that is often concentrated in regions with abundant existing records or on common species, while ecologically fragile marginal areas such as mountains and deserts, as well as endangered species with scarce data, are marginalized in modeling due to a lack of sufficiently representative samples. As a result, they increasingly struggle to receive adequate monitoring, early warning, or restoration, posing greater challenges to the overall effectiveness of ecological governance and biodiversity protection.

生态文明在深层结构上是关于“生态正义如何建构”的问题。只有实现生态利益在不同物种、区域和人群之间的公平分配，达成人与人、人与自然关系的全面正义，才是真正的生态文明。人工智能在提升生态治理效率的同时，其算法偏向也在悄然侵蚀生态正义，导致环境治理决策在形式上合理，在实质上却可能加剧生态不平等。所谓算法偏向(algorithmic bias),也被称为算法偏见或算法歧视，是指人工智能系统在算法设计、数据采集、模型训练和实际应用过程中，由于开发者的主观假设、数据来源的局限性等因素，导致系统在输出结果中存在歧视或忽视特定群体和区域的现象。当前许多生态模型和监测系统在算法训练阶段所依赖的数据，往往集中于已有资料丰富的区域或常见物种，而山地、荒漠等生态脆弱的边缘地区以及数据稀缺的濒危物种，则因缺乏足够代表性样本而在建模中被系统边缘化，从而越发难以获得充分的监测、预警或修复，使得生态治理的整体效能与生物多样性保护面临更大挑战。

Going further, algorithmic bias may also strip ecological justice of its institutional safeguards through technological means. In traditional governance systems, ecologically fragile regions and endangered species could often, through public participation and the expression of opinions, receive human intervention in institutional design, thereby enjoying a degree of priority protection rights, such as policy preferences, resource compensation, or targeted restoration. However, the essence of an algorithm is a mechanism for allocating attention. “Attention plays a key role in information filtering and the allocation of cognitive resources across various information processing tasks.” That is, through the screening, ranking, and weighting of information, algorithms prioritize the preferences and expressions of mainstream groups, while marking ecologically fragile regions, marginalized groups, or non-mainstream issues as “low relevance” or “low priority,” thereby being overlooked or filtered out in data presentation, decision-making recommendations, and risk prediction. In short, as algorithms increasingly become the technical protagonists in government and corporate ecological governance, marginalized groups and the faint voices of ecological demands find it even harder than before to gain opportunities for expression and intervention; they are directly screened out of technical outputs, losing the possibility of being responded to in the policy feedback chain. Ultimately, ecological justice is not only not strengthened by human technological progress but also loses the last line of defense provided by institutional safeguards.

更进一步，算法偏向还可能使生态正义的制度保障被技术剥夺。在传统治理体系中，生态脆弱区域、濒危物种等往往能够通过民众参与和意见表达，在制度设计中获得人为干预，从而在一定程度上享有优先保护权利，如政策倾斜、资源补偿或定向修复等。然而，算法的本质是一种注意力分配机制，“注意力在各类信息加工过程中起到信息筛选和认知资源分配的关键作用”,即算法通过对信息的筛选、排序和权重设定，优先放大主流群体的偏好和表达，而将生态脆弱区域、边缘群体或非主流议题标记为“低相关”或“低优先级”,进而在数据呈现、决策建议和风险预测中被忽视或被过滤。简言之，随着算法日益成为政府和企业生态治理的技术主角，那些处于边缘的群体及生态诉求微弱的声音，较之以往更难获得表达与介入的机会，直接被筛除在技术输出之外，在政策反馈链条中失去了被回应的可能性，最终，生态正义不仅未能在人类技术进步中得到强化，反而失去了制度保障的最后防线。

Therefore, ecological justice in the age of artificial intelligence has a more complex and multidimensional connotation. It involves not only value orientations but also the implicit biases that are invisibly amplified by algorithms. Even some subtle potential value tendencies “will be discovered through machine learning algorithms, and will be identified by the system and used as indicators, thereby reinforcing bias.” To this end, on the one hand, it is necessary to expand the representativeness of training data, incorporating local ecological knowledge, the experiences of marginalized groups, and diverse ecological cultural practices into the data system, so as to avoid algorithmic exclusion caused by “data poverty.” On the other hand, it is necessary to introduce bias identification and correction mechanisms in the process of constructing algorithmic models. That is, by adjusting model weights, introducing representative samples, or employing manual correction pathways, guide algorithms to allocate sufficient “attention” to ecological issues and social groups originally marked as “low priority,” increase the visibility of non-mainstream issues in data and their responsiveness in models, and build a green algorithm system with social sensitivity. Only in this way can we truly achieve the synergistic advancement of technological progress and ecological justice.

因此，人工智能时代的生态正义具有更为复杂和多维的内涵，不仅涉及价值导向，还包括那些被算法无形中放大的隐性偏见，即便是一些细微的潜在价值倾向，“也会通过机器学习算法被发现，而且被系统识别并被用作指标，从而增强偏见”。为此，一方面，要扩大训练数据的代表性，将地方性生态知识、边缘群体经验以及多元生态文化实践纳入数据体系，避免“数据贫瘠”造成的算法排斥；另一方面，要在算法模型构建过程中引入偏见识别与纠正机制，即通过调整模型权重、引入代表性样本或人工校正路径，引导算法将足够的“注意力”分配给原本被标记为“低优先级”的生态问题和社会群体，增加非主流议题在数据中的可见性及在模型中的响应度，建设具有社会敏感性的绿色算法体系，从而真正实现技术进步与生态正义的协同共进。

(II) AI Hallucinations and the Accuracy Verification of the Information Base for Ecological Decision-Making

(二)AI幻觉与生态决策信息基础的准确性校验

For ecological civilization construction to achieve tangible results, the key lies in scientific decision-making. Generative artificial intelligence may produce “hallucinations” when providing decision support. So-called AI hallucinations refer to the generation of content that does not conform to facts or is unrelated to user requests, manifesting as “misattribution,” “fabrication,” or “irrelevant responses,” such as citing forged literature, fabricating data, or providing misleading narratives. The content generated by AI hallucinations often carries a high degree of “confidence”; even completely erroneous information is presented as if it were well-reasoned and evidence-based. At the same time, humans generally exhibit a “cognitive miser” psychology, meaning that when faced with complex problems, they tend to adopt the most effort-saving cognitive methods, including ignoring part of the information to reduce cognitive burden, relying on existing information to avoid seeking more, or accepting a “seemingly good enough” suboptimal choice. This cognitive inertia, combined with trust in technology, leads people, when confronted with complex ecological problems, to directly adopt the ready-made knowledge output by AI without delving into its logic or data sources. In the field of environmental governance, which has high technical thresholds, the misleading nature of AI hallucinations is even more insidious, because most people lack sufficient professional discernment to identify the pseudo-knowledge that others have accepted. This allows unverified erroneous information to enter formal texts or decision-making processes, leading to severe deviations in environmental risk assessments, the formulation of ecological governance plans, and resource allocation strategies, and even producing directional errors or catastrophic ecological consequences.

生态文明建设要取得实效，重在科学决策。生成式人工智能在提供决策支持时可能出现“幻觉”。所谓AI幻觉，指的是生成与事实不符或与用户请求无关的内容,表现为“张冠李戴”“无中生有”“答非所问”,如引用伪造文献、杜撰数据、误导性叙述等。AI幻觉的生成内容往往带有很高的“自信度”,即使是完全错误的信息，也会被呈现得仿佛有理有据。同时，人类普遍存在“认知吝啬鬼”(cognitive miser)心理，即在面对复杂问题时倾向于采用最省力的认知方式，包括忽略部分信息以减少认知负担，依赖已有信息以避免继续寻找更多信息，或者接受一个“看起来还不错”的次优选择。这种认知惰性与科技信赖叠加，使得人们在面对复杂生态问题时，可能直接采信人工智能输出的现成知识，而不对其逻辑与数据来源进行深究。在技术门槛较高的环境治理领域，AI幻觉所带来的误导更具隐蔽性，因为多数人缺乏足够的专业甄别能力，难以识别他人采信的伪知识，从而使未经核实的错误信息进入正式文本或决策流程，导致环境风险评估、生态治理方案制定、资源配置策略等出现严重偏差，甚至产生方向性错误乃至灾难性生态后果。

Furthermore, AI hallucinations can also become a source of information pollution, affecting the effectiveness of ecological decision implementation. The false information generated by AI poses a severe challenge to the public opinion ecosystem. It “fractures the cognitive landscape of public opinion facts, manipulates the public orientation of opinion discourse, and hijacks the social trust in opinion values.” The large-scale proliferation of false environmental information disrupts the public’s basic understanding of environmental issues, weakening the public’s correct comprehension of and support for ecological protection policies. AI hallucinations may also be improperly exploited to create an “information fog,” deliberately steering environmental public opinion toward extremes or opposition, provoking or intensifying conflicts and disputes in the ecological and environmental field, and undermining the social consensus and implementation willingness of all parties in environmental protection decision-making.

此外，AI幻觉还可能成为信息污染源，影响生态决策的执行成效。AI生成的虚假信息对舆论生态构成严峻挑战，它“割裂着舆论事实的认知图景、操纵着舆论话语的公共导向、绑架着舆论价值的社会信任”。大规模泛滥的虚假环境信息扰乱公众对环境问题的基本认知，削弱公众对生态保护政策的正确理解与支持配合。AI幻觉还可能被不当利用以制造“信息迷雾”,蓄意引导环境舆论走向极端化或对立面，挑起或激化生态环境领域的矛盾纠纷，破坏各方在环境保护决策上的社会共识和执行意愿。

The root cause of AI hallucinations is not, as some technologists believe, a problem of training data, model architecture, or generation strategies. Rather, it lies in the fact that generative artificial intelligence essentially lacks genuine “understanding” ability. Its “learning” is nothing more than the indiscriminate acceptance of all externally instilled knowledge, followed by the algorithmic matching, correspondence, and formal transformation of existing information. “If it is given something, then it searches according to what it is given, finds the consequent of the conditional, and outputs it as the result.” Since what AI generates is merely “meaning-representing language,” possessing “only syntax but no semantics,” then “the involvement with meaning remains solely a matter for the designers and operators.” [11]11 The original Chinese text contains a footnote marker ③ here, which is retained in the translation as the endnote reference. The source of this quotation is not provided in the input. Therefore, it is necessary to reaffirm the primacy and irreplaceability of human judgment in ecological governance, and to promote the establishment of a traceable algorithmic governance mechanism, ensuring that every environmental decision involving AI undergoes rigorous output verification, including expert review of information, retention of human oversight, and refinement of the chain of responsibility. This is not only a necessary measure to enhance the reliability of AI but also a fundamental requirement for maintaining humanity’s subject position in the construction of ecological civilization. Currently, China has begun to pay attention to this issue. For example, the “Interim Measures for the Management of Generative Artificial Intelligence Services” issued in 2023 clearly stipulates: “Adopt effective measures to enhance the transparency of generative artificial intelligence services and improve the accuracy and reliability of generated content.” [12]12 The original Chinese text contains a reference to the 2023 “Interim Measures for the Management of Generative Artificial Intelligence Services” (《生成式人工智能服务管理暂行办法》), a Chinese national regulation. Preventing AI hallucinations from interfering with and posing risks to public policy areas such as the ecological environment has entered the purview of national institutional supervision, laying the foundation for building a healthy and orderly technological application ecosystem.

AI幻觉产生的根本原因，并非部分技术派所认为的是训练数据、模型架构或者生成策略的问题，而在于生成式人工智能本质上并不具备真正的“理解”能力，它的“学习”不过是对外部灌输的知识不加分辨地全部接纳，然后根据算法对已有信息进行匹配、对应、形式转换，“如果被给予了什么，那么它就根据被给予的去搜索，把条件句的后件搜索出来，然后作为输出给出”。由于人工智能生成的只是“意义表征语言”,“只有句法，但是没有语义”,那么“对意义的涉及仍只是设计、操作人员的事情”③。因此，必须重申人类判断在生态治理中的首位性与不可替代性，推动建立可追责的算法治理机制，确保人工智能参与的每一项环境决策都经过严格的输出验证，包括专家审核信息、保留人工监督以及细化责任链条等。这不仅是提升人工智能可靠性的必要举措，更是维护人类在生态文明建设中主体地位的根本要求。目前，我国已经开始重视这个问题，如2023年出台的《生成式人工智能服务管理暂行办法》明确规定，“采取有效措施，提升生成式人工智能服务的透明度，提高生成内容的准确性和可靠性”。防范AI幻觉对生态环境等公共政策领域带来的干扰与风险，已进入国家制度监管视野，为构建健康有序的技术应用生态奠定了基础。

(III) Systemic High Energy Consumption and Full Lifecycle Carbon Footprint Information Disclosure

(三)系统性高能耗与全生命周期碳足迹信息披露

When people discuss energy-saving technologies, they usually focus on their potential to reduce energy consumption at the consumer end, yet rarely consider the energy consumption implicit in the design, manufacturing, and even background operation of such technologies themselves. Artificial intelligence operates in digital form and ostensibly produces no visible pollution, but in reality it is highly dependent on natural resources and its energy consumption is concealed. First, the facilities of AI are materially supported by key rare metals such as rare earths and silicon-based materials, and the extraction and processing of these resources involve complex industrial processes, with overall energy consumption far exceeding that of traditional industries. As Kate Crawford has pointed out, “We must attend to the physical reality of AI: the mineral extraction, energy consumption, and water use that underpin the glossy surface of ‘intelligent’ systems. AI is not an immaterial dream of pure computation; it is a rapacious industry rooted in the earth’s crust.” [13]13 Kate Crawford, *Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence* (New Haven: Yale University Press, 2021), p. 8. Second, the technical characteristic of AI systems being “online around the clock” means that even in an “idle” state, they still consume about 60% of their peak power to maintain data synchronization and model updates. Srnicek states: “A quick look at the energy consumption of data centers quickly proves (the entire internet uses about 9.2% of global electricity).” [14]14 Nick Srnicek, *Platform Capitalism* (Cambridge: Polity Press, 2017), p. 43. Finally, in the age of AI, the definition of efficiency has shifted from how much physical output corresponds to a unit of energy consumption to real-time processing of information, dynamic response to user behavior, and personalized push notifications. Under this service model, systems must constantly invoke algorithmic models and increase computational frequency to achieve precise adaptation to the specific preferences of individual users, thereby greatly increasing computational load and energy intensity. It can be said that AI technology is restructuring the energy use mechanism of modern society, transforming the energy consumption pattern from the traditional linear growth corresponding to output scale into a nonlinear, high-density dissipation process driven by high-frequency interaction and targeted algorithms.

人们在谈论节能技术时，通常关注其在消费终端降低能耗的潜力，却很少考虑这类技术本身在设计、制造乃至后台运行过程中所隐含的能源消耗。人工智能以数字化形态运作，表面上不产生可视化污染，实则高度依赖自然资源且能耗隐蔽。首先，人工智能的设施以稀土、硅基材料等关键稀有金属为物质支撑，而这类资源的提取与加工涉及复杂的工业流程，整体能耗远高于传统产业。正如凯特·克劳福德(Kate Crawford)所指出的，“我们必须关注人工智能的物理现实：支撑‘智能’系统光鲜外表的是矿物开采、能源消耗和用水量。人工智能并非纯计算的无形梦想；它是一个植根于地壳的贪婪产业”。其次，人工智能系统“全天候在线”的技术特性意味着，即便在“空闲”状态下，为保持数据同步与模型更新，仍会消耗其峰值功率的60%。斯尔尼塞克说：“随便看一下数据中心的能源消耗，都能快速被证明(整个互联网的用电量约占全球的9.2%)。”最后，在人工智能时代，效率的定义已不再是单位能耗对应多少物理产出，而是转向对信息的实时处理、用户行为的动态响应与个性化推送。在这种服务模式下，系统需不断调用算法模型、提高计算频率，以实现对具体用户特定偏好的精准适配，从而大幅增加运算负荷与能耗强度。可以说，人工智能技术正在重构现代社会的能源使用机制，使能耗模式从传统的与产出规模相对应的线性增长，转变为由高频交互和靶向算法驱动的非线性、高密度耗散过程。

In summary, the energy consumption of AI is not limited to the operational load of data centers but is also manifested in the coupled linkage between different technical levels, exhibiting typical systemic high-energy-consumption characteristics. We must comprehensively examine the energy consumption structure of AI technology and avoid creating a greater burden on the ecological environment in the name of digitalization. First, the scope of existing AI ethics should be expanded; it should not be limited only to concerns such as data privacy and ideological risks but should also incorporate consideration of the natural environment, promoting the extension of AI ethics from a moral circle exclusive to humanity to a broader vision of a community of life. Second, a mechanism for tracing environmental data covering the entire life cycle of AI should be established, incorporating not only the energy consumption generated during algorithm operation into the assessment system but also the resource input of its preliminary infrastructure construction and even the environmental degradation costs after equipment decommissioning into the scope of energy consumption calculation. A mandatory environmental information disclosure mechanism for digital energy consumption should also be established, making its ecological externalities quantifiable, traceable, and intervenable, thereby providing institutional guarantees for ensuring that AI is conducive to sustainable ecological development. Finally, an incentive mechanism for AI energy consumption governance should be established. By setting energy efficiency thresholds for data centers and introducing market-based means such as computing power quota trading and ecological compensation, the design philosophy of AI should be guided to transform from maximizing algorithmic performance to seeking optimal solutions under ecological constraints.

综上所述，人工智能的能耗不仅局限于数据中心的运行负荷，更体现在不同技术层级之间的联动耦合，呈现出典型的系统性高能耗特征。我们必须全面审视人工智能技术的能耗结构，避免以数字化之名制造更大的生态环境负担。首先，要拓展现有人工智能伦理的议题范围，不应仅局限于数据隐私、意识形态风险等方面的关切，还应纳入对自然环境的观照，推动人工智能伦理从人类独享的道德圈拓展到更大的生命共同体视野。其次，推动建立覆盖人工智能全生命周期的环境数据溯源机制，既将算法运行过程中产生的能源消耗纳入评估体系，也将其前期基础设施建设的资源投入乃至设备退役后的环境降解成本一并纳入能耗计算范围，并建立数字能耗的强制性环境信息披露机制，使其生态外部性影响变得可量化、可追踪、可干预，为确保人工智能有利于生态可持续发展提供制度保障。最后，建立人工智能能耗治理的激励机制，通过设定数据中心能效的阈值，引入算力配额交易、生态补偿等市场化手段，引导人工智能设计理念从追求算法性能的最大化向寻求生态约束层面的最优解转型。

III. New Forms of Constructing Representations of Natural Objects through Data and Information

三、数据信息建构自然客体表征的新形式

Public ecological awareness, including the level of cognition of nature and the degree of participation in ecological and environmental affairs, reflects the social penetration of the concept of ecological civilization and is an important indicator for measuring the level of development of ecological civilization in a society. Relying on the access of sensor networks, Internet of Things terminals, and algorithmic models, AI encodes ecological environmental conditions such as air quality, water pollution, and carbon emission concentrations into a dynamically generated, real-time feedback data system, presenting them in real time in daily life through smartphones, wearable devices, or public platforms. This not only reshapes people’s cognitive model of nature but also promotes the shift of natural information from the closed context of specialized technology to an open state available to the public, raising new demands on people’s ecological awareness in the age of AI.

公众的生态意识，包括对自然的认知水平以及在生态环境事务中的参与程度，体现了生态文明理念的社会渗透力，是衡量一个社会生态文明发展水平的重要标志。人工智能依托传感器网络、物联网终端与算法模型的接入，将空气质量、水体污染、碳排放浓度等生态环境状况，编码为一个动态生成、实时反馈的数据系统，通过智能手机、可穿戴设备或公共平台在日常生活中实时呈现。这不仅重塑了人们对自然的认知模式，也推动了自然信息从专业技术的封闭语境走向公共可用的开放状态，对人工智能时代人们的生态意识提出了新要求。

(I) Enhancing the Level of Knowledge about the “Data Nature”

(一)提升关于“数据自然”的知识水平

Knowledge of the ecological environment is an important cognitive foundation supporting the construction of ecological civilization. In the age of AI, although the popularization of digital technology has made it more convenient for people to acquire ecological knowledge, the requirements for public environmental literacy have also increased accordingly. We must not only understand the representational meaning presented by ecological environmental data but also grasp the logic of its generation and the technical mechanisms behind it.

生态环境知识是支撑生态文明建设的重要认知基础。进入人工智能时代，虽然数字技术的普及使人们获取生态知识变得更加便捷，但对公众环境素养的要求也随之提高了。我们不仅要理解生态环境数据所呈现的表征意义，还要把握其生成的逻辑及背后所依赖的技术机制。

In terms of knowledge content, there is a leap from empirical common sense of environmental protection to technically processed environmental data. Against the backdrop of AI transforming natural information into various visualized and dynamic data forms, people need to possess a basic understanding of technical semantics, be able to read the environmental conditions reflected by ecological data, and judge the trends of environmental change they portend. This ability to identify and understand the ecological information conveyed by the “data nature” is a basic requirement for people’s environmental literacy in the age of AI.

在知识内容上，从经验性的环境保护常识向技术化处理的环境数据跃升。在人工智能将自然信息转化成各种可视化、动态化的数据形式的背景下，人们需具备对技术语义的基本理解，看懂生态数据所反映的环境状况，判断它们所预示的环境变化趋势。这种识别和理解“数据自然”所传递的生态信息的能力，是人工智能时代对人的环境素养的基本要求。

In terms of depth of knowledge, there is a leap from “knowing what is” to “knowing why it is so.” With the support of AI technology, ecological cognitive ability is no longer a simple factual awareness of ecological phenomena themselves but is manifested in an understanding of the entire process of how ecological information is captured, transformed, and presented by technology. Specifically, ecological cognitive ability includes not only the static interpretation of the “data nature” but also a basic insight into the generation mechanisms and interfering factors behind its dynamic changes. Ecological information exhibits fluctuating characteristics, which are the result of the combined action of multiple factors. In addition to the traditionally familiar factors such as natural climate change and human activity intervention, these also include new variables introduced by AI algorithms in the process of collecting and processing natural data, such as the time frequency of monitoring, the density of data updates, and the sensitivity and response mechanisms of technical equipment. Only by comprehensively grasping the various variables and mechanisms affecting fluctuations in ecological data can one accurately interpret the “data nature,” avoiding the misjudgment of occasional large fluctuations as major changes in the ecological environment or the misinterpretation of short-term data fluctuations as trend shifts. This analytical ability, which is both data-based and maintains independent judgment, is an important quality of thought for scientifically carrying out environmental protection in the age of AI.

在知识深度上，从“知其然”向“知其所以然”跃升。在人工智能技术的支持下，生态认知能力不再是简单的对生态现象本身的事实知晓，而是体现在对生态信息如何被技术捕捉、转换与呈现的全过程理解。具体而言，生态认知能力不仅包括对“数据自然”的静态解读，也包括对其动态变化背后的生成机制与干扰因素的基本洞察。生态信息呈现出波动性特征，是多种因素共同作用的结果，除了传统上我们所熟知的自然气候变化、人类活动干预等因素，还包括人工智能算法在采集和处理自然数据过程中引入的新变量，如监测的时间频率、数据更新的密度、技术设备的灵敏度与响应机制等。只有全面掌握影响生态数据波动的各类变量与机制，才能准确解读“数据自然”,避免将偶发性的大幅波动误判为生态环境的重大变化，或者将短期的数据起伏误读为趋势性变迁。这种既以数据为依据又保持独立判断的辨析能力，是人工智能时代科学开展环境保护的重要思维品质。

In terms of knowledge judgment, there is a leap from technological trust to rational reflection. In the context of algorithm-generated ecological knowledge, humans cannot become mere “data consumers.” “What is important here is that we not only understand how machine learning works but also ensure the transparency of processes and decision-making methods and the proper management of data.” [15]15 Virginia Eubanks, *Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor* (New York: St. Martin’s Press, 2018), p. 178. Faced with seemingly objective and neutral data expressions such as ecological scoring systems, environmental quality indices, and carbon emission ratings, people should pay attention to technical biases such as the sources of training data, indicator settings, and sampling frequency, and actively examine the value orientations and value presuppositions behind them, such as which variables are given higher weight and which values—such as economic priority, risk tolerance, and efficiency orientation—are given priority. Only by deeply understanding the process of algorithm generation and mastering the knowledge production mechanisms and pathways of power intervention behind the technology can one avoid mistaking the technically constructed “ecological simulacrum” for the true state of nature itself, and then actively participate in algorithm design, data annotation, and model tuning processes, safeguard the indispensable humanistic values in ecological governance, and ensure that the development of ecological civilization is not alienated into a data-driven “techno-utopia.”

在知识判断上，从技术信赖向理性省思跃升。在算法生成生态知识的背景下，人类不能沦为单纯的“数据消费者”,在此，“重要的是我们不仅要了解机器学习的工作原理，还要确保流程和决策方法的透明性以及对数据的正确管理”。面对生态评分系统、环境质量指数、碳排放评级等看似客观中立的数据表达，人们应关注训练数据的来源、指标设定、采样频率等技术偏差，主动审视其背后的赋值取向和价值预设，如哪些变量被赋予更高权重、哪些价值如经济优先、风险容忍、效率导向等被优先体现。只有深入理解算法生成过程，掌握技术背后的知识生产机制与权力介入路径，才能避免将技术建构的“生态拟象”误认为自然本体的真实状态，进而主动参与算法设计、数据标注与模型调优过程，守护生态治理中不可或缺的人文价值，确保生态文明的发展不被异化为由数据驱动的“技术乌托邦”。

(II) Creating a New Type of AI-Driven Ecological Culture

(二)创造人工智能驱动的新型生态文化

Ecological culture is the spiritual soul of the ecological civilization system. Culture, as the ideological superstructure, is “directly interwoven with people’s material activities, with their material intercourse, and with the language of real life”; “it is not consciousness that determines life, but life that determines consciousness.” [16]16 Karl Marx and Friedrich Engels, *The German Ideology* (1845–1846). That is to say, the generation and efficacy of culture must be rooted in real life and participate in people’s daily interactions and meaning construction in order to achieve the effect of “transforming people through culture” [17]17 “以文化人” (yǐ wén huà rén): a classical Chinese expression meaning to use culture to educate and transform people. by nurturing value identification from social experience. Artificial intelligence promotes the universal sharing of knowledge, turning environmental information that originally belonged to specialized fields into information resources directly accessible in people’s daily lives. This popularization of the knowledge-power structure transforms the grand-narrative ecological discourse into a tangible field of reality perceptible to every individual, creating an unprecedented opportunity for people to reshape green lifestyles and practice ecological culture.

生态文化是生态文明体系的精神灵魂。文化作为观念的上层建筑，是“直接与人们的物质活动，与人们的物质交往，与现实生活的语言交织在一起的”,“不是意识决定生活，而是生活决定意识”。也就是说，文化的生成与效能必须植根于现实生活，参与人们的日常交往与意义建构，才能达到从社会经验中孕育价值认同的“以文化人”效果。人工智能推动了知识的普惠共享，使得原本属于专业领域的环境信息成为人们日常生活中可直接获取的信息资源。这种知识权力结构的大众化将宏大叙事的生态话语转化为每个人切身可感的现实场域，为人们重塑绿色生活方式和践行生态文化创造了前所未有的契机。

First, promoting the embedding of lightweight ecological applications into daily consumption terminals. “Consumption, as necessity, as need,” [18]18 Karl Marx, *Grundrisse* (1857–1858). is an important practical dimension of human lifestyle and cultural expression. Integrating functions such as energy-saving tips and low-carbon suggestions into smart home systems and intelligent travel platforms helps individuals conveniently access ecological information, adjust consumption behavior in a timely manner, and make life decisions aligned with environmental protection guidance. At the same time, constructing mechanisms such as carbon accounts and green points, and using artificial intelligence to record, quantify, and provide real-time feedback on individuals’ green behaviors, enables people to accurately perceive their actual impact on changes in environmental indicators, thereby enhancing the sustainability and sense of achievement of ecological actions. Furthermore, developing apps such as waste-sorting assistants and energy-saving check-in tools will also promote the cultivation of people’s ecological “micro-habits,” gradually building a diversified green lifestyle driven by artificial intelligence technology.

一是推动轻量化生态应用程序嵌入日常消费终端。“消费，作为必需，作为需要”,是人类生活方式和文化表达的重要实践维度。将节能提示、低碳建议等功能集成于智能家居系统、智慧出行平台，有助于个体便捷获取生态信息，及时调整消费行为，作出符合环保导向的生活决策。同时，构建碳账户、绿色积分等机制，借助人工智能对个体绿色行为进行实时记录、量化与反馈，使人们能够精准感知其对环境指标变动的实际影响，从而增强生态行动的持续性与成就感。此外，开发垃圾分类助手、节能打卡工具等APP,也将促进人们生态“微习惯”的养成，逐步构建人工智能技术驱动的多元化绿色生活样态。

Second, using artificial intelligence to promote the modern transformation of traditional ecological culture. “The Chinese nation has always respected and loved nature; the Chinese civilization, spanning more than five thousand years, has nurtured a rich ecological culture.” [19]19 Xi Jinping, speech at the National Conference on Ecological and Environmental Protection (18–19 May 2018). This long-standing concept of nature has been deeply integrated into the spiritual bloodline of the Chinese people, constituting a unique cultural gene of the Chinese nation. Today, the rise of artificial intelligence technology has injected new vitality into the traditional ecological culture rooted in thousands of years of Chinese agrarian civilization, enabling it to gain new forms of expression and channels of dissemination in the digital context. On the one hand, by leveraging artificial intelligence technology, traditional Chinese ecological cultural resources scattered in oral memory, ancient classics, and folk techniques can be systematically organized, analyzed, and reproduced, so that originally fragmented and ephemeral ecological life experiences and natural cultural concepts are preserved in digital form and enter people’s daily audio-visual experience through multimedia forms such as images, audio, and animation, allowing traditional ecological culture to permeate the life-world of contemporary people. On the other hand, relying on the high-dimensional semantic recognition and deep reconstruction functions of large models, traditional ecological concepts such as “harmony between man and nature,” “the Dao follows nature,” and “guiding according to circumstances,” [20]20 “天人合一” (tiān rén hé yī), “道法自然” (dào fǎ zì rán), “因势利导” (yīn shì lì dǎo): classical Chinese philosophical concepts. “Harmony between man and nature” originates in Confucian and Daoist thought; “the Dao follows nature” is from the *Daodejing*; “guiding according to circumstances” is a strategic principle from ancient Chinese military and political thought. as well as textual expressions praising and protecting nature in ancient poetry and decrees, can be incorporated into modern ecological governance systems and public education platforms, promoting the transformation of natural imagery and ecological wisdom in traditional culture into ideological resources for ecological protection policies, ecological value judgments, and knowledge production, thereby providing a richly layered discursive foundation for constructing new concepts and new expressions of the independent knowledge system of ecological civilization in the New Era.

二是运用人工智能推动传统生态文化的现代转化。“中华民族向来尊重自然、热爱自然，绵延五千多年的中华文明孕育着丰富的生态文化。”这种源远流长的自然观念已深深融入中国人的精神血脉，构成中华民族独特的文化基因。如今，人工智能技术的兴起为植根于中华民族数千年农耕文明的传统生态文化注入了新的活力，使其在数字语境下获得新的表达形式与传播路径。一方面，借助人工智能技术，对散见于口述记忆、古代典籍与民间技艺中的中华传统生态文化资源进行系统化整理、分析与再现，使原本零散、易逝的生态生活经验和自然文化观念以数字形式保存，并通过图像、音频、动画等多媒体形式进入人们的日常视听体验，让传统生态文化浸润当代人的生活世界。另一方面，依托大模型的高维语义识别与深度重构功能，将“天人合一”“道法自然”“因势利导”等传统生态理念以及古代诗词、律令中保护自然、赞美自然的文本表述，纳入现代生态治理体系与公共教育平台，推动传统文化中的自然意象与生态智慧转化为生态保护政策、生态价值判断和知识生产的思想资源，为建构新时代生态文明自主知识体系的新概念、新表述提供底蕴厚重的话语支撑。

Third, exploring the development of a digital ecological culture system. Generative artificial intelligence enables cultural creation to engage in cross-border imagination within an open semantic space, not only providing a technological fulcrum for breaking through existing forms of expression in the innovative development of ecological culture, but also capable of generating on a large scale diverse media content such as ecological-themed narrative texts and visual images, continuously enhancing the communicative penetration, emotional appeal, and visual impact of ecological culture. At the same time, relying on intelligent algorithms, immersive interaction, and virtual reality technology, it promotes the spatial reconstruction of the ecological culture field, constructing new cultural scenes such as green metaverses and intelligent ecological parks, so that the relationship between humans and nature is no longer limited to people’s direct perception of natural entities but shifts toward a deep interaction based on technological media, revealing a unique new appearance of ecological culture in the age of artificial intelligence.

三是探索发展数字生态文化体系。生成式人工智能使文化创作能够在开放性的语义空间中展开跨界想象，不仅为生态文化的创新发展提供了突破既有表达形式的技术支点，也能大规模生成生态主题的叙事文本、视觉影像等多元媒介内容，不断增强生态文化的传播渗透力、情感感染力与视觉冲击力。同时，依托智能算法、沉浸交互与虚拟现实技术，推动生态文化场域的空间重构，构建绿色元宇宙、智能生态园区等新型文化场景，使人与自然的关系不再局限于人对自然实体的直接感知，而是转向一种基于科技媒介的深度互动，展现人工智能时代独有的生态文化新样貌。

(3) Establishing and improving a public environmental participation system based on data feedback

(三)建立完善数据反馈的公众环境参与制度

Public environmental participation is an important component of modern ecological civilization construction. In traditional understanding, public participation is often limited to conscious expressive behaviors, such as opinion feedback and policy suggestions. However, from the perspective of ecosystem theory, humans, as natural beings, are constantly “participating” in material exchange with nature at all times; this existential participation has a real and profound impact on the natural environment. Through technical means such as sensor networks, wearable devices, and Internet of Things terminals, artificial intelligence can capture in real time the material-energy exchange between humans and nature and convert it into data, making originally hidden forms of environmental participation identifiable and regulable. In this context, the design of public environmental participation systems in the age of artificial intelligence must confront and respond to this new reality of “non-explicit participation,” redefine the boundaries of “participation,” endow daily behavioral data with institutional significance, and ensure that everyone’s role in the ecological environment is fully recognized and effectively reflected.

公众环境参与是现代生态文明建设的重要内容。在传统理解中，公众参与往往被限定为有意识的表达行为，如意见反馈、政策建言等。然而，从生态系统论的角度来看，人作为自然存在物，无时无刻不在“参与”与自然界的物质变换，这种生存性参与对自然环境产生真实而深刻的影响。人工智能通过传感器网络、可穿戴设备与物联网终端等技术手段，能够实时捕捉人与自然之间的物质能量交换并将其转化为数据，使原本隐性的环境参与形式获得可识别和可调控的可能。在此背景下，人工智能时代的公众环境参与制度设计必须正视并回应这个“非显性参与”的新现实，重新界定“参与”的边界，赋予日常行为数据以制度意义，使每个人在生态环境中的作用都被充分承认和有效体现。

First, improving the explicit expression mechanism for environmental participation. Artificial intelligence injects new technological momentum into environmental participation, enabling originally scattered and fragmented public expressions to be quickly classified and integrated, helping decision-makers more accurately grasp the environmental issues and governance demands of public concern, thereby enhancing the responsiveness of ecological civilization system construction to public opinion. More importantly, through intelligent systems that quantitatively model and publicly display indicators such as the degree of people’s participation, the adoption rate of suggestions, the areas where opinions are incorporated, and the effectiveness of environmental improvements, people can not only see that public opinion is incorporated into institutional decision-making but also directly witness the actual driving effect of public opinion on real problems, thereby forming a positive feedback mechanism that encourages more people to actively express environmental concerns. Research shows that artificial intelligence technology “has been proven to have higher trustworthiness” in decisions involving quantitative calculations. [21]21 Reference to studies on AI trustworthiness in quantitative decision-making; specific source not cited in the original. Therefore, fully leveraging the quantitative support role of artificial intelligence in enhancing the transparency of environmental participation and strengthening the responsiveness to public opinion is an innovative path for promoting whole-process people’s democracy in the field of ecological civilization in the New Era.

首先，完善环境参与的显性表达机制。人工智能为环境参与注入了新的技术动力，使原本零散、分散的公众表达能够被迅速分类整合，帮助决策者更准确地把握公众关切的环境问题与治理诉求，从而增强生态文明制度建设对民意的响应能力。更重要的是，通过智能系统对人们的参与程度、建议采纳率、意见吸纳领域以及环境改善成效等指标的量化建模与公开展示，人们不仅可以切实看到民意被纳入制度决策，还能直观见证民意对现实问题的实际推动作用，由此形成正向反馈机制，激励更多人积极表达环境关切。研究表明，人工智能技术在涉及定量运算的决策“被证明具有更高的可信任度”。因此，充分发挥人工智能在提升环境参与透明度、增强民意回应性方面的量化支撑作用，是新时代推动生态文明领域实现全过程人民民主的创新路径。

Second, exploring the establishment of an implicit expression mechanism for public environmental participation. Public environmental participation is not only reflected in the realization of expressive rights but also concerns the issue of representativeness in governance structures. Incorporating data on public daily environmental behaviors, such as individual carbon footprints, energy consumption records, and waste-sorting habits, as legitimate variables in indicator setting, resource allocation, and policy adjustment is an expansion and supplement to the traditional “representatives speak on behalf” mechanism, introducing a new mechanism for public participation in ecological civilization construction. Reconstructing the way the public is “present” through data forms, so that even in situations where they are absent from direct expression, they can still play a practical role through institutional arrangements, helps to break through the structural bias of “a few active individuals representing the majority of the silent” in traditional public participation, promoting a shift in governance logic from “discursive participation” to “behavioral representation,” thereby more prominently highlighting the subject status of “people” in the environmental governance system. At the same time, this mechanism also helps to promote the comprehensive inclusion of individual data in artificial intelligence systems, reducing to a certain extent the risk of algorithmic bias caused by missing or unbalanced data. Of course, this kind of participation must clearly define the scope of collection, use, preservation, and disclosure principles for environmental behavior data, not only preventing technological abuse and privacy infringement but also safeguarding the public’s basic right to know and right to participate. “It is necessary to ensure that people are widely aware of the potential impacts of AI and also that they realize they can participate in shaping the social development of AI”; [22]22 Quotation from an unspecified source on responsible AI development. this is an important measure for building responsible artificial intelligence.

其次，探索建立公众环境参与的隐性表达机制。公众环境参与不仅体现为表达性权利的实现，更关涉治理结构中的代表性问题。将个体碳足迹、能源消费记录、垃圾分类习惯等公众日常环境行为数据纳入指标设定、资源配置与政策调整的合法变量，是对传统“代表说事”机制的拓展与补充，为生态文明建设引入了公众参与的新机制。通过数据形式重构公众“在场”的方式，使其即便在缺席直接表达的情境中，也能通过制度安排发挥实际作用，有助于突破传统公共参与中“少数积极者代表多数沉默者”的结构性偏差，推动治理逻辑由“话语参与”向“行为代表”转变，从而更加凸显“人”在环境治理体系中的主体地位。同时，这一机制也有助于推动个体数据全面纳入人工智能系统，在一定程度上降低因数据缺失或不均衡而导致的算法偏向风险。当然，这种参与必须明确环境行为数据的采集范围、用途、保存与开放原则，既要防止技术滥用与隐私侵权，也要保障公众的基本知情权和参与权，“既要确保人们广泛了解AI的潜在影响，又要使人们意识到他们可以参与塑造AI的社会发展”,这是建设负责任的人工智能的重要举措。

Finally, establish a mechanism for the synergistic integration of multiple information channels. The opinion tendencies of explicit expression and implicit expression are not always highly consistent; sometimes they may conflict with each other. This is not only because the data sources differ and the dimensions of concern vary, but also because the expressers themselves are different groups, whose positions, preferences, and even interest demands often diverge. If public opinion is blindly followed without discrimination, it is easy to fall into the misunderstanding of “tailism,” which is instead detrimental to scientific environmental decision-making. Therefore, it is necessary to establish a multi-data synergistic integration mechanism in which explicit participation and implicit participation corroborate each other. Specifically, traditional political inquiry information, smart terminal sensing data, and algorithm prediction results should be uniformly incorporated into a system architecture. Machine learning algorithms should be used to conduct dynamic comparison and trend modeling of explicit and implicit data, identifying correlations and points of deviation. Through comprehensive analysis, a more complete environmental governance response system should be formed, and disclosed to the public in an appropriate manner. This not only opens up the transmission path of public participation from the micro-daily level to policy decision-making, but also helps the public understand why policies are formulated in this way, on which data they are based, and which groups’ wishes they reflect, thereby alleviating social contradictions in environmental governance to a certain extent, and providing an important pathway for artificial intelligence to play a coordinating role in ecological affairs.

最后，建立多元信息渠道的协同整合机制。显性表达与隐性表达的意见倾向并不总是高度一致，有时可能是相互冲突的。这不仅是因为数据来源不同、关注维度有别，更是因为表达者本身就是不同群体，其立场、偏好乃至利益诉求往往存在差异。如果不加辨析地盲从公众意见，容易陷入“尾巴主义”的误区，反而不利于科学的环境决策，因此，要建立显性参与和隐性参与互为印证的多元数据协同整合机制。具体而言，将传统问政信息、智能终端感知数据与算法预测结果统一纳入一个系统架构，利用机器学习算法对显性与隐性数据进行动态比对和趋势建模，识别其中的关联性与偏离点，通过综合研判形成更为全面的环境治理响应体系，并以适当方式向公众公开。这不仅打通了公众参与从微观日常到政策决策的传导路径，也有助于让公众理解政策为何这样制定、基于哪些数据、体现了哪些群体的意愿，从而在一定程度上缓和环境治理中的社会矛盾，为人工智能在生态事务中发挥协调作用提供了一种重要路径。

Conclusion

结 语

As the most cutting-edge technology, artificial intelligence expands the subject category of ecological civilization, pushes the ethics of ecological technology from outcome application to front-end design, and translates the movement processes of natural objects into encoded data forms, thereby profoundly changing the structural foundation of ecological civilization. However, the fact that artificial intelligence exercises active regulatory functions in specific scenarios is, in the final analysis, still a manifestation of the essential power of humanity, and still serves humanity’s purposive activities. Lewis Mumford pointed out, “Even the most automated machine requires conscious human participation at the beginning and end of the process: at the beginning, initial design; at the end, resolving flaws and making repairs.” [23]23 Lewis Mumford, *The Myth of the Machine: Technics and Human Development* (New York: Harcourt, Brace & World, 1967), p. 9. Therefore, the holistic reconstruction of ecological civilization by artificial intelligence does not mean that the role of humans is weakened; on the contrary, it places higher demands on human ecological awareness, value judgment, and active participation. As the ultimate responsible subject, humanity must deeply understand and be adept at utilizing the practical characteristics of artificial intelligence, guiding this emerging “quasi-subject” to always remain on the track of technical ethical norms, and constructing an artificial intelligence governance framework oriented toward green development. This is not only a new topic in current theoretical research on ecological civilization, but also an important opportunity for the deepening of human ecological rationality and the elevation of subjectivity in the age of artificial intelligence.

人工智能作为最前沿的技术，拓展了生态文明的主体范畴，推动生态技术伦理从结果应用走向前端设计，将自然客体的运动过程转译为编码化的数据形态，从而深刻改变了生态文明的结构基础。然而，人工智能在特定场景下行使主动调节功能，归根结底仍是人类本质力量的体现，依然服务于人的合目的性活动。刘易斯·芒福德指出，“即便是自动化程度最高的机器，在工艺的开头和结尾也需要人有意识的参与：开头需要最初设计，结尾需要解决瑕疵，做出修补”。因此，人工智能对生态文明的整体性重构，并不意味着人的角色被削弱，反而对人的生态意识、价值判断和主动参与提出了更高要求。人类作为终极责任主体，必须深刻认识和善于运用人工智能的实践特性，引导这种新兴的“拟主体”始终保持在技术伦理规范的轨道上，建构面向绿色发展的人工智能治理框架。这不仅是当前生态文明理论研究的全新课题，也是人工智能时代人的生态理性深化与主体性高扬的重要契机。

Author’s brief introduction: Li Juan, Professor at the School of Marxism, Beijing Normal University.

作者简介：李娟，北京师范大学马克思主义学院教授。