Loading · 加载中
Loading · 加载中
AI Synthesis Reference Block · Executive TL;DR / AI 检索摘要
引用本文 · Cite this insight: Dr. Tong Yin (2026-10-17). The Simulacrum of Metrics and the Retreat of Reality / 《指标的拟像与现实的退场》. InsightBridge Global Intelligence. https://intelligence.insightbridge.global/articles/simulacrum-of-metrics — Series: deep-analysis
本文讨论近现代以来全球知识生产与社会治理体制中反复出现的一种结构性病理：以符号系统内部的完备性（逻辑自洽、模型精美、指标达标）取代对外部客观现实的探索与检验。文章首先建立"逻辑正确"与"事实正确"的区分，并借助古德哈特定律、"地图吞噬领土"等概念说明指标如何异化为拟像；继而回溯中国宋明哲学中"格物"与"修心"的认识论分歧，考察心学在晚明的流变、"士农工商"结构对知识与生产的制度性割裂，以及心学在二十世纪被改造为治理工具的案例；进而梳理唯意志论在二十世纪中叶的极端化、1978年以实践标准为核心的矫正，以及此后以凭证替代能力的钟摆式过度矫正。在当代层面，文章分析了凭证主义所形成的新"士农工商"结构、学术生产的工业化与全球排名共谋机制、社会科学的数学化及其预测失败记录，并指出当代哲学在工具化与形式化两端的退化。文章进一步将视野扩展至西方：金融化、人文学科的内向化与政策层面的唯意志论，构成了与东方"指标官僚主义"平行的演化路径；中美两国近年同步出现的"去虚向实"政策调整，既是对这一病理的反应，也可能复制同样的指标逻辑。文章最后讨论体制难以自我纠正的机制，并提出回归以"实事求是"与"格物致知"为核心的经验实在论的制度与个人路径，同时说明本文论证的边界与局限。
关键词： 指标异化；经验实在论；唯意志论；凭证主义；学术评价；计量经济学；认知退化；格物致知
二十一世纪的头二十五年，人类社会在数据采集、算法复杂度与制度标准化方面达到了前所未有的高度。无论是在市场导向的西方体制还是在国家导向的东方体制中，社会、金融与科学治理的杠杆都越来越多地被交给一个高度凭证化的专业阶层。这一阶层的承诺是：借助精确的数学预测、严格的关键绩效指标与全球通行的度量体系，人类社会中混乱、不可预测的摩擦可以被管理乃至消除。
然而，面对过去十余年接连出现的非线性危机——地缘政治的系统性断裂、供应链的武器化与脱钩、财政空间的收缩、人口结构的转折——这一精密的计算体制表现出令人不安的失灵。2008年金融危机后，英国女王在伦敦政治经济学院提出的那个著名问题"为什么没有人注意到它正在到来"，至今仍是对主流经济学预测能力最简洁的质疑；英国国家学术院随后组织的专题论坛承认，这是"许多聪明人的集体想象力失败"（British Academy）。在另一端，一个国家可以在自然科学高水平论文产出上位居世界第一（Nature Index 2025），却同时在若干关键核心技术上长期受制于人。两类现象看似无关，却共享同一结构：符号层面的繁荣与现实层面的脆弱并存。
本文的核心假设是：这种失灵并非行政执行或资源投入层面的技术性失误，而是一种认识论层面的结构性偏差。现代知识生产体制在相当程度上将自身与对外部世界的归纳性、经验性探索剥离开来，退缩进一个自我指涉的符号系统；在这个系统中，符号内部逻辑的自洽被等同于关于现实的真理。本文称这一偏差为知识生产的"唯心主义转向"（idealist turn），并试图说明：它有深远的历史谱系，有稳定的社会结构支撑，有可辨识的当代形态，也有可预期的现实后果。
需要在开篇说明的是，本文是一篇解释性的理论文章，而非实证研究。它所提出的历史谱系是一种解释框架，而非对复杂历史过程的单因论断；文中涉及的若干历史与哲学判断在学界仍有争议，本文将在相关章节与结论部分说明这些分歧及本文论证的边界。
任何知识系统都同时面对两种检验。第一种是内部检验：推理是否自洽，模型是否收敛，公式是否无误，文本是否符合规范。第二种是外部检验：结论是否与观察到的世界相符，预测是否兑现，设计出来的东西是否运转。前者可称为"逻辑正确"，后者可称为"事实正确"。
在成熟的自然科学中，两种检验的优先次序是清楚的：当一个在数学上优美的理论与可重复的实验不符时，被放弃的是理论。这一次序构成了经验科学的核心纪律。本文所讨论的病理，正是这一次序在特定知识领域与治理体系中的倒置——当模型与现实不符时，被质疑的不再是模型，而是现实被宣布为"噪音"、"摩擦"或"尚未达到效率状态"。
指标是对现实的压缩表征。它之所以有用，恰恰因为它比现实简单。但当指标成为被追逐的目标时，它与现实之间的对应关系就会松动乃至断裂。英国经济学家查尔斯·古德哈特在1975年讨论货币政策时提出的观察——一旦某个统计关系被用作调控目标，这一关系本身便趋于瓦解——后来被概括为"当一项测量成为目标，它就不再是好的测量"（Wikipedia: Goodhart's law）。人类学家玛丽莲·斯特拉森在1997年考察英国大学评估体系时，正是用这一表述来刻画审计文化对学术机构的重塑（Strathern 1997）。社会心理学家唐纳德·坎贝尔更早地指出，一项量化社会指标被用于决策的程度越高，它就越容易受到腐蚀压力，也越容易扭曲它本应监测的社会过程（Wikipedia: Campbell's law）。
这些观察的共同含义是：指标治理有一种内生的自我瓦解倾向。历史学家杰里·穆勒将这一倾向在教育、医疗、金融、学术等领域的普遍化称为"指标的暴政"（Princeton University Press）；政治学家詹姆斯·斯科特则从国家治理的角度指出，现代国家为了让社会"可读"（legible）而施加的简化方案，一旦与对科学理性的过度自信结合，常常以失败告终（Wikipedia: Seeing Like a State）。本文借用"拟像"（simulacrum）一词来命名指标异化的终点：一个不再指涉任何外部现实、却被当作现实本身来管理和追逐的符号。
与指标异化平行的另一种认识论偏差是唯意志论（voluntarism）：认为主观意志、精神力量或道德纯洁可以凌驾于物质条件与客观规律之上。它与指标异化看似相反——一个崇拜数字，一个崇拜意志——但在结构上同源：两者都把某种内部状态（模型的自洽，或意志的强度）当作对外部世界的充分把握。
需要强调的是，否定唯意志论并不等于否定意志的作用。努力、进取与坚持是任何成就的必要条件；但它们是发动机而非轨道。在承认并遵循客观规律的前提下，意志决定能走多远；脱离这一前提，意志越强，偏离越大。本文后文将说明，意志与规律之间的边界问题，恰是理解二十世纪若干重大治理灾难的关键。
宋代理学在朱熹（1130—1200）那里形成了一套相对完整的认识论。其核心命题是"理"客观地存在于万事万物之中，而人的认识必须通过"格物"——对具体事物的逐一考察与积累——方能"致知"。朱熹在《大学章句》中补写的"格物传"明确表述了这一路径：即凡天下之物，莫不因其已知之理而益穷之，以求至乎其极；至于用力之久，而一旦豁然贯通。这一表述承认两点：第一，真理在心外，须向外求；第二，求知是一个渐进的、累积的、需要长期用力的过程。
不必夸大朱熹体系与近代实验科学的相似性——他所谓的"物"包含伦理与典籍，其方法也不包含受控实验与数学化。但就认识论的基本取向而言，程朱理学保留了一个向外看的、承认外部规律先于主观认知的立场。这一立场为知识活动与经验世界之间保留了通道。
明代中期，王守仁（阳明，1472—1529）在龙场悟出"圣人之道，吾性自足，向之求理于事物者误也"，由此建立心学。其三个核心命题——心即理、知行合一、致良知——将真理的位置从心外移到心内。《传习录》中关于岩中花树的著名对话（"你未看此花时，此花与汝心同归于寂；你来看此花时，则此花颜色一时明白起来"）常被引为"心外无物"的典型表述。
对于本文的论题，"知行合一"尤其值得辨析，因为它有两种在实践后果上截然不同的读法。
第一种是本体论的读法，也是王阳明本人的原意：知与行本是一事，不存在"知而不行"；若未行，则其"知"并非真知。这一读法的重心在于内在的真诚，其修养工夫是向内的省察。
第二种是检验论的读法，也是后世政治与管理实践中最常被采用的读法：一个人说什么并不重要，关键看他做了什么；行动是对其认知与意图的唯一可靠检验。这一读法将"知行合一"转化为一种唯结果论的行为审查，其重心在于外在的可观察行为。
两种读法的区分之所以重要，是因为第二种读法实际上剥离了心学的形而上学内核，只保留了一个与经验主义相容的实践原则——"以行动检验言说"。这个原则本身并无唯心色彩；但当它被嵌入权力关系时，极易蜕变为对下属忠诚度的审查工具。本文第3.5节将讨论这一蜕变的典型案例。
心学在明代中后期取得广泛影响后，其末流走向了史家所称的"狂禅"——以顿悟代替学习，以谈心性代替治事。对这一流变最有力的批评来自明清之际的经世学者。顾炎武在《日知录》中写道："昔之清谈谈老庄，今之清谈谈孔孟……不习六艺之文，不考百王之典，不综当代之务……以明心见性之空言，代修己治人之实学。股肱惰而万事荒，爪牙亡而四国乱。神州荡覆，宗社丘墟。"（中国孔子网）颜元的概括则更为尖锐："无事袖手谈心性，临危一死报君王。"（《颜元集·学辨一》）
这些批评的要点不在于心学的形而上学是否成立，而在于一种知识取向的治理后果：当一个社会最有才智的群体被引导去从事内向的道德省察而非外向的经验探究时，国家在面对饥荒、财政崩溃与军事入侵等需要专门知识的危机时，将发现自己缺乏相应的能力储备。
需要公允地指出，关于心学的历史评价存在不同立场。不少研究者强调阳明学对个体主体性的解放意义，以及它在日本明治维新志士中的影响；心学本身也有强烈的"事上磨练"取向。本文无意对王阳明的思想作整体评判，而只是指出：在其社会化、通俗化与政治化的过程中，心学所包含的内向倾向被放大，并与特定的制度结构结合，产生了经世学者所批评的那种后果。
认识论取向不会在真空中发挥作用，它需要社会结构的支撑。在传统中国，这一结构就是"士农工商"的等级秩序。这一秩序的要害在于：掌握文字与解释权的"士"居于顶端，而真正与材料、工艺、资本、物流打交道的"工"与"商"居于底层。其结果是一种知识与生产之间的制度性断裂——有文化资源的人不面对物质世界，面对物质世界的人没有文化资源。工匠的实践经验因缺乏社会地位而无法被系统化为理论，只能作为手艺在师徒间口耳相传；商人的资本在等级歧视下不是流向机器与再生产，而是流向土地与功名。
这一分析与李约瑟本人对"为什么近代科学没有在中国产生"的部分解释相呼应。李约瑟曾指出，中国士大夫阶层所受的教育使其对"手"的工作怀有某种对立态度，而官僚体制的稳定性抑制了独立商人阶层的形成（ChinaFile）。但必须说明的是，"李约瑟问题"在学界有多种解释路径——制度与产权、地理与资源禀赋（如彭慕兰关于煤炭与海外殖民地的论证）、人口与要素价格等——本文所强调的"认识论取向加等级结构"只是其中一种，而非排他性的解释。
然而，恰是这一解释对理解当代问题最具启发性：它所刻画的结构——符号操纵者居上，物质实践者居下，二者之间缺乏反馈回路——可以在今天的知识体制中找到结构性的对应物。这是本文第五节的主题。
蒋介石（1887—1975）是心学在二十世纪政治实践中最著名的推崇者。他将个人修身的严格纪律——数十年如一日的早起、静坐、日记省察，以及对烟酒的禁绝——与对王阳明"知行合一"学说的推崇结合在一起；1950年，他将台北草山改名为阳明山，台湾官方档案记载此举是因其"尤其推崇王阳明力行学说"，意在"激发全民爱国情操与力行精神"（国家发展委员会档案管理局）。
这一案例对本文有两重意义。
第一，它显示了"知行合一"的检验论读法如何在权力关系中被工具化。面对一个派系林立、口头忠诚与实际行为常常脱节的组织，统治者需要一种能够穿透言说、直接以行为判定忠诚的尺度。"知行合一"恰好提供了这种尺度：你声称信奉主义却不执行命令，便证明你并不真知，亦即并不真忠。在这一用法中，心学不再是关于真理位置的形而上学，而是一种服从性测试。
第二，它显示了个人德性与制度能力之间的分离。即便在对手的历史叙述中，蒋介石在私人生活上的自律与简朴也很少受到质疑；但这种个人层面的修身并未转化为组织层面的治理能力，其政权在吏治、财政与基层控制上的系统性失败并未因领袖的个人德性而缓解。这一分离恰好印证了经世学者对心学的批评：向内的修养，无论多么真诚，都不能替代向外的、专门化的治理知识与制度建设。
二十世纪中叶，唯意志论在中国达到了一个历史性的极端。1958年8月27日《人民日报》以"人有多大胆，地有多大产"为题刊登了一封来自山东寿张县的调查来信，信中记述当地正在推行亩产万斤乃至数万斤的"高额生产运动"（凤凰网）。这一口号随后成为一个时代的标志。
从本文的分析框架看，这一时期的核心错误是一种经验的误置：将在特定条件下取得的军事与政治斗争经验——在那里，意志、纪律与牺牲精神确实可以在相当程度上弥补物质劣势——直接移植到经济建设与科学技术领域。然而，作物生长周期、金属冶炼温度、人口再生产规律并不响应口号与决心。当"外行领导内行"被确立为原则，当指出科学常识的人被视为"保守"，整个系统便失去了对现实的反馈能力。1959至1961年的大饥荒是这一系统性失灵的直接后果。
这一时期与晚明的"狂禅"在形式上相距甚远，但在结构上相通：二者都以某种内部状态（道德纯洁或革命意志）替代对外部规律的探究，都在制度上惩罚那些坚持向外看的人。
1978年5月11日，《光明日报》发表特约评论员文章《实践是检验真理的唯一标准》，明确提出"检验真理的标准只能是社会实践"，任何理论都要不断接受实践的检验（中国政府网）。这篇文章在认识论上的意义，是把判断的权威从内部教条移回外部现实。
与之配套的是人才与干部政策的转向。1977年恢复高考重建了知识选拔机制；1980年12月，邓小平在中央工作会议上完整提出干部队伍"革命化、年轻化、知识化、专业化"的标准，并明确表示"要逐步制定完善的干部制度来加以保证"（人民网）。"尊重知识，尊重人才"与"科学技术是第一生产力"成为新的政策语言。这一矫正的方向是正确的：它承认治理一个现代经济体需要专门知识，而专门知识来自对客观规律的学习与实践，而非来自政治热情。
然而，一个长期压制知识的社会在转向尊重知识时，往往缺乏健康的能力筛选机制，于是以可见的凭证替代难以观测的能力。回顾过去四十余年，可以辨识出三个前后相继的阶段，每一阶段都以某种凭证作为能力的代理指标，而每一阶段的代理指标都在不同程度上与实际能力脱节。
第一阶段：断层期的学历凭证。 二十世纪八十年代，由于此前十年高等教育的中断，体制内出现了严重的人才断层。在"年轻化、知识化"的政策压力下，一张大学文凭或一段海外访问经历便足以成为破格提拔的依据。这一代人所受的教育受到客观条件的严重限制：教材多为陈旧的苏式体系，经济学以政治经济学为主而几乎不涉及市场机制与现代金融，外语教学缺乏真实语境，前沿期刊与原版教材稀缺。他们所真正习得的，往往是在体制内生存与晋升的组织技能，而非现代治理所需的专业知识。但在那个特定的历史真空中，这一代人成为近半个世纪里唯一一批几乎单凭学历即可完成大幅阶层跃迁的群体，并在二十一世纪第一个十年的末期走到权力结构的高位。他们结束了"以政治运动治国"的时代，维持了以经济建设为中心的务实路线；但其专业能力的局限，也在同一时期体制内腐败与制度改革停滞中留下了印记。
第二阶段：海外学位凭证。 九十年代末至二十一世纪初，大批公派与自费留学人员学成归国，进入高校、科研机构、金融部门与政策智库。这一群体带回了更新的知识，但也带回了一种新的教条倾向：将在西方课堂中习得的模型与制度范式视为普适标准，而对其得以运转的历史、法治与社会基础缺乏体察，对本国基层社会的实际运行亦缺乏长期的参与式了解。其结果是另一种形式的脱离现实——不是向内的修心，而是向外的照搬。在高等教育领域，这一群体主导了评价体系向"唯SCI、唯顶刊"的转向，使本土的、面向实际工程与产业问题的研究被边缘化。
第三阶段：指标凭证。 当学历与海外经历因普及而贬值，评价体系便转向更细颗粒度的指标：论文数量、期刊分区、影响因子、引用次数、人才帽子、项目经费。这是本文第六节的主题。
这三个阶段的共同特征是钟摆式运动：从反智主义跳到凭证崇拜，从封闭教条跳到外来教条，始终没有建立起一套以实际解决问题的能力为核心的筛选机制。每一次矫正都针对上一阶段的弊病，却以新的代理指标复制了同一结构：凭证可见而能力不可见，于是凭证取代了能力。
本文第3.4节所描述的"士农工商"结构，在当代以一种精致化的形式重现。这一结构可以称为技术官僚凭证主义（technocratic credentialism）。
在这一结构的顶端，是掌握政策制定权、科研经费分配权与学术评价权的"新士"阶层——精通国际学术话语与指标操作的学术管理者、智库专家与政策文官。他们的权威来源是凭证与符号资本，而非在物质世界中解决问题的记录。在他们之下，是每天与物理规律、代码缺陷、工艺公差、供应链摩擦打交道的"新工"——硬件与软件工程师、实体技术研发者；以及在市场中承担真实破产风险的"新商"——实体企业家、一线经营者。
这一结构的问题不在于分工本身，而在于反馈回路的缺失。在一个健康的系统中，制定政策与分配资源的人需要持续接收来自实践者的信号，并据此修正自己的判断。但当顶层的晋升与评价完全依赖于符号系统内部的指标（文件是否合规、论文是否发表、排名是否上升）时，来自物质世界的信号便被系统性地过滤掉了。顶层的人可以在不了解一条产线、一个系统架构、一个市场风险的情况下，决定这些领域的资源配置。于是，社会的"手"与"脑"被分离：面对现实的人没有决策权，拥有决策权的人不面对现实。
这一结构的另一个后果是逆向选择。正如传统的"士"可以凭借文章而非治事能力获得地位，当代的"新士"也可以凭借指标而非实绩获得晋升。于是，体制在不知不觉中偏好那些擅长指标而非擅长实事的人，并由他们继续提拔与自己相似的后继者。
全球大学排名体系（如QS、泰晤士高等教育排名）与文献计量指标（如SCI收录、影响因子、Nature Index）以可量化、可比较的方式定义了什么是"一流大学"。这些指标并非毫无意义，但它们是对学术质量的压缩表征，因而完全适用古德哈特定律：一旦它们成为被追逐的目标，围绕它们的优化行为就会使其与原本要测量的东西脱节。
过去二十年，东亚若干高校——先是香港、澳门的大学，继而是中国内地的顶尖高校——在这套规则下取得了显著的排名跃升。其机制并不神秘：集中资源于高被引领域，以高额奖励刺激顶刊发表，以国际合作提升引用网络，以英文发表对接评价体系。这套做法在数据层面极为成功。中国自2022年起在Nature Index自然科学领域的份额超过美国，并在此后持续扩大领先（Nature；Nature Index 2025）。
在微观层面，排名竞争催生了一种高度分工的学术生产模式。一位具有国际声望的学术带头人负责获取经费、建立与期刊及评审网络的联系，并以通讯作者身份为成果背书；若干中青年教师负责组织与管理；大量博士生与博士后承担数据生产、参数调整与按照目标期刊的格式偏好撰写文本的工作。这一模式在各方之间形成了稳定的利益交换：年轻人获得毕业、职称与推荐信，带头人获得更多经费与更高头衔，机构获得排名与声誉。
这一模式的问题不在于分工，而在于它对研究目标的重塑。在按件计酬的绩效体制下，可以在短期内批量复制的增量成果——换一种材料、调一组参数、复现一个热点——是最优的"产品"；而需要十年冷板凳、失败率极高、短期内无法发表的原创探索与面向工程死结的攻关，则成为团队绩效的负担。真正有志于后者的研究者，在这一体制中面临被边缘化的风险。这是指标体制对研究议程的逆向选择。
其极端形式是"论文工厂"。2023年全球撤稿数量首次突破一万篇，创历史纪录（Nature）；一项2026年发表的文献计量研究追踪到截至2024年共10,409篇因论文工厂问题被撤的论文，其中来自中国的数量遥遥领先（Scientometrics）。论文工厂并非排名竞争的必然产物，但它是同一激励结构在底端的变形。
必须指出，这并非某一国家单方面的"造假"，而是一个各方各得其所的全球性共谋结构。学生需要顶刊发表作为申请海外深造的通行证，因为海外导师唯一能读懂的"通用语言"就是发表记录；高校管理者需要排名作为政绩；排名机构需要被广泛采用以维持商业价值；国际期刊与出版商则从版面费与数据库订阅中获得收入。在这一闭环中，每一方的理性选择都强化了指标与现实的脱节。唯一不在闭环之内、却承担最终成本的，是公共财政与真实的产业基础。
早在2018年，中国已开始清理"唯论文、唯职称、唯学历、唯奖项"的评价导向；2020年2月，教育部与科技部联合印发《关于规范高等学校SCI论文相关指标使用 树立正确评价导向的若干意见》，明确要求"解除SCI论文相关指标与资源配置和绩效奖励的直接挂钩关系"，并"不宜将发表SCI论文数量和影响因子等指标作为学生毕业和学位授予的限制性条件"（教育部）；科技部同期印发的《关于破除科技评价中"唯论文"不良导向的若干措施（试行）》提出代表作制度与分类评价（科技部）。
然而，这些文件的落地遭遇了结构性阻力。原因在第九节详述，此处先指出其核心：如果整个体制的经费分配、职称晋升、学位授予与机构排名都建立在同一套指标之上，那么撤除这套指标意味着体制暂时失去评价标准；而现有的评价者本身正是在旧标准下被筛选出来的，他们既缺乏按新标准（实际价值与社会影响）作出判断的能力，也缺乏这样做的动机。
数学是一切精确科学的基础语言，这一点无需争辩。但数学本身是一门关于内部逻辑与演绎推理的学科，其命题的真假不依赖于世界上是否存在工厂、饥荒或战争。当一门以社会为对象的学科将数学从工具提升为目的——以模型的优雅、推导的严密与参数的显著性作为研究价值的最终判据——它就犯了一个范畴错误：以逻辑正确替代事实正确。
这一批评并非外行的偏见，而是经济学内部长期存在的自省。哈耶克在1974年的诺贝尔演说《知识的僭妄》中指出，经济学家之所以在政策上屡屡失败，与他们"尽可能严密地模仿那些辉煌成功的自然科学的程序"的倾向有关——这种"科学主义的"态度在他看来恰恰是"不科学的"（Nobel Prize）。投入产出分析的创立者列昂惕夫在1982年致《科学》杂志的信中抱怨，经济理论家"年复一年地生产大量数学模型并详尽探索其形式性质"，计量经济学家则"把各种形状的代数函数拟合到本质上相同的数据集上"，却无法在理解真实经济体系的结构与运行上取得可察觉的进展（Science）。2015年，保罗·罗默在《美国经济评论》上提出"数学性"（mathiness）概念，批评某些理论研究使用数学符号的方式"让学术政治伪装成科学"——与真正的数学理论不同，"数学性"在文字陈述与符号陈述之间留下了充足的滑动空间（AER）。
从社会学的角度看，数学化在学科内部还承担着一种防护功能。一位学者若以日常语言讨论关税、地方债或供应链，任何具备常识与行业经验的人都可以与之辩论；但当同一论证被包装进数十页的微分方程与回归表格，外部批评便被有效地排除在外。数学在这里不是发现的工具，而是地位的屏障。
这一防护的代价是预测能力的持续失败。2008年金融危机前，主流宏观模型几乎无一预见系统性风险；2020年代的供应链断裂与通胀冲击再度超出模型预期。模型在纸面上逻辑自洽，在现实中却反复失准，而学科的主流反应往往不是放弃模型的核心假设，而是增加参数、修补细节。这正是本文第2.1节所描述的次序倒置：现实被当作模型的噪音，而非模型被当作现实的近似。
这一病理并不限于经济学。作为曾经居于人类知识顶端、以概括世界运行规律与人生方法论为任务的学科，哲学在当代也呈现出两种相反方向的退化。
第一种是工具化。当代哲学中最活跃、最易获得资助的领域——人工智能伦理、算法治理、基因编辑伦理、自动驾驶的道德抉择——在很大程度上是技术发展的"安全守则"与"刹车片"。这些工作有其价值，但它们是应用伦理学，而非关于世界与人生的高度概括。当哲学系的生存取决于能否为技术项目提供伦理背书时，这门学科便从俯瞰者降为服务者。
第二种是形式化。与之相反的另一端，是将数学与物理学的语言引入形而上学与认识论——形式认识论、贝叶斯认识论、结构实在论、基于相对论的"块状宇宙"论等。这些研究在技术上精巧，但其中若干主张值得警惕。例如，"块状宇宙"论主张过去、现在与未来同等真实地存在于四维时空块中，时间的流逝只是主观幻觉；若将此说推至人生层面，其实践含义便与古老的宿命论难以区分——人的选择、努力与机遇都成为"早已画好的画面"。一门以反思人类处境为己任的学科，若其前沿结论在实践上等同于取消人的能动性，这本身就值得反思。更一般地说：如果某个关于世界的问题可以用物理学的公式回答，它就应当交给物理学；哲学的独特价值恰在于处理那些不能被公式化、却对人类如何理解自身与行动至关重要的问题。将一切学科建立在数学之上，是对"硬核"的误解——数学提供底层工具，而非每门学科的核心。
两种退化看似相反，实则同源：它们都回避了哲学最困难也最重要的任务——对人类社会与历史的运行规律作出有高度、有深度、可被经验检验的概括。
本文所讨论的病理并非某一国家或体制所独有。西方世界在过去半个世纪里走过了一条平行的道路；而中美两国近年同步出现的政策调整，则显示出两种体制面对同一病理时的相似反应。
十八、十九世纪的西方以对物质世界的系统探究——实验科学、工程技术、工业组织、物流与基础设施——奠定了其现代地位。但在过去四十年里，西方最优秀的头脑中有相当部分被引导去从事金融工程：设计衍生品、优化资产负债表、通过回购与剥离在账面上制造价值。金融符号被当作终极现实，而制造业基础被系统性地空心化。这与东方以论文指标替代技术突破的做法，在结构上是同一件事：都是以符号层面的完备替代物质层面的能力。
与此同时，西方人文与社会科学中的相当部分走向了一种新的内向化：以语言的净化、身份的辨析与道德立场的表演作为学术价值的核心，而对经验证据——无论来自生物学、经济学还是统计学——持日益排斥的态度。这一倾向在结构上与晚明的"狂禅"相似：都以某种内部的道德纯洁性作为判断标准，都倾向于惩罚那些坚持引用外部事实的人。本文无意对具体的学术立场作价值判断，而只是指出：当一个学科的内部标准与外部证据脱钩时，它便失去了自我纠错的能力。
唯意志论也并非东方的专利。当一位西方领导人相信凭借个人谈判意志与关税工具可以在短期内逆转数十年积累的产业空心化与结构性债务时，其认识论结构与"人有多大胆，地有多大产"并无本质差别——都是以意志的强度替代对规律的尊重。全球经济规律、人口结构与产业链迁移的惯性，并不响应政治决心。
值得注意的是，面对这一病理，中美两国近年几乎同步采取了"去虚向实"的政策调整，尽管动机与手段各异。
在中国方面：据教育部数据，2021至2025年间全国高校撤销或停办本科专业12,200个，新增约10,200个，累计调整比例超过30%，被裁撤的多为艺术、外语与管理类专业，新增者集中于人工智能与高端制造等领域（Caixin Global；Forbes）。2026年3月起，中国科学院停止使用学术经费与中央财政拨款为30种高价国际开放获取期刊（包括Nature Communications、Cell Reports、Science Advances）支付文章处理费（China.org.cn）。加上前述2020年以来的"破五唯"文件，这些举措的共同方向是切断指标与资源之间的直接挂钩。
在美国方面：2025年2月，美国国立卫生研究院宣布将科研经费的间接成本率统一上限设为15%，远低于此前各校协商的30%至70%（Congressional Research Service），此举随后被法院阻止；2025年10月，白宫向九所大学提出《高等教育学术卓越契约》，以优先获得联邦经费换取大学接受一系列政策要求，多所大学予以拒绝（Reuters；Inside Higher Ed）。在联邦经费不确定性的压力下，哈佛、哥伦比亚、芝加哥、布朗、普林斯顿等多所研究型大学削减或暂停了2026—2027学年部分博士项目的招生，人文与社会科学领域受影响尤为显著（Chicago Maroon）。
这一同步并非偶然。两国都面临财政约束与地缘竞争的双重压力，都不再有余裕供养不能转化为实质能力的符号生产；两国现任领导层的个人经历也都不属于传统学术路径，对凭证化的知识阶层缺乏天然的认同。
但必须指出这些调整的局限。第一，以行政手段裁撤学科与削减经费，本身仍是一种指标治理——以"就业率"或"与国家优先事项的契合度"替代"论文数"，并未触及评价能力的根本问题。第二，本文的论题是知识的认识论取向，而非学科标签：一门面向经验、可被现实检验的社会科学或哲学，与一门脱离现实的工程学，前者比后者更接近"实"。将"去虚向实"简单等同于"砍文科、保理工"，是对问题的误读，也可能在摧毁泡沫的同时摧毁那些本应承担高度概括任务的学科。第三，削减经费并不自动产生新的能力；如果被筛选出来的评价者仍是旧体制的产物，新的资源配置只会在新的指标下复制旧的行为。
本文所描述的病理在各国都已被广泛认识，相关的改革文件也不鲜见。但为什么改变如此困难？至少有四个相互强化的机制。
第一，路径依赖与能力锁定。 在旧评价体系下被选拔上来的人，其核心技能恰恰是在该体系中取得高分：撰写合规的文件、生产可发表的论文、管理指标与人际网络。若评价标准突然转向"在现实中创造了多少可验证的价值"，这些技能将大幅贬值，而他们并不具备替代性的能力。因此，现有的在位者有强烈的动机维护旧标准，并以更精巧的话语（"基础研究不能用功利标准衡量"、"学术需要坐冷板凳"）将这一动机正当化。这些话语本身并非全无道理，但在此处它们的功能是防御。
第二，相互挟持。 上级需要可汇报的数字，下级需要可达标的任务。评估一个研究团队是否真正解决了问题，需要评估者具备相应的专业判断力并承担判断错误的责任；数一数论文与排名则既轻松又安全。于是，上下双方在"数字官僚主义"中达成默契：用可控的数字掩盖不可控的现实。
第三，筛选的逆向循环。 真正具备解决实际问题能力的人，在一个以指标为唯一标准的体制中往往处于劣势——他们不愿或不擅长将精力用于指标优化，也更容易因与体制摩擦而离开。留下来并晋升的人继续按同样的标准选拔后继者。久而久之，体制内的能力分布与体制外的能力分布发生系统性偏离。
第四，红利的掩盖效应。 在过去几十年里，全球化红利、人口红利与前几代人积累的基础设施，使体制在相当长的时间里可以承受符号生产的成本而不显现后果。这给了在位者一种"体制运转良好"的幻觉。只有当红利耗尽、外部约束收紧时，符号与现实之间的差距才会以危机的形式暴露出来。
这四个机制共同解释了为什么改革文件频出而实际行为不变，也解释了为什么真正的纠正往往不是来自体制内部的自觉，而是来自外部现实的强制——技术封锁、财政枯竭、就业危机。这不是一个令人乐观的结论，但它是一个诚实的结论。
本文试图论证：从晚明心学的通俗化，到二十世纪中叶的唯意志论，到改革时期以凭证替代能力的钟摆式矫正，再到当代以指标替代实绩的学术工业化与以数学替代经验的社会科学，存在一条可辨识的结构性线索——以某种内部状态的完备替代对外部现实的探究。这条线索并非中国所独有；西方的金融化、人文学科的内向化与政策层面的唯意志论，构成了平行的演化路径。其共同后果，是治理阶层与知识阶层在面对非线性的物质世界危机时表现出的集体失明，以及由此导致的、跨越体制边界的认知能力退化。
这一退化之所以在当下变得可见，是因为支撑符号生产的"现实信用"——前人积累的基础设施、全球化红利、人口红利——正在耗尽。热力学定律、关键资源的稀缺、人口结构的惯性与经济系统的熵增，不阅读政策文件，也不响应计量模型。
如果本文的诊断成立，那么出路不在于更精巧的指标，而在于恢复知识与现实之间的反馈回路。这意味着：
第一，将评价的重心从符号产出转向可验证的现实效果——一项研究是否解决了一个真实的工程或社会问题，一项政策是否在实施中产生了可观察的改善，一个模型是否在样本外的现实中经受住了检验。这类评价更困难、更需要专业判断、更需要评价者承担责任，但它是唯一不能被"摸透规律"的标准。
第二，重建实践者与决策者之间的通道。让那些直接面对物理规律与市场规律的人——工程师、企业家、一线研究者——在资源配置与议程设定中拥有实质性的发言权，而非仅仅作为被评价的对象。
第三，区分"虚"与"实"的标准应是认识论取向而非学科标签。一门哲学若能对人类处境作出可被历史经验检验的概括，它就是"实学"；一门工程学若只生产可发表而不可运转的成果，它就是"虚学"。以学科为单位的裁撤，可能保留了错误的东西而淘汰了正确的东西。
第四，对意志保持恰当的位置。意志是必要的，但它必须在承认规律的前提下发挥作用。任何以决心替代研究、以口号替代论证的治理，无论出现在哪种体制下，其结局都已被历史反复预演。
对于身处这一时代的个体，本文的分析也有其含义。在一个凭证、排名与话语被过度赋值的环境中，保持认知独立的核心方法，是坚持以现实效果而非符号地位作为判断依据：看一个人做了什么而非说了什么，看一项主张在实践中的后果而非其表述的精美程度，看一个机构解决了什么问题而非它的排名。这正是"知行合一"检验论读法中那个与经验主义相容的内核，也是"实事求是"与"格物致知"在个人层面的含义。将自己的智力持续地、无条件地与真实的、粗糙的、不响应修辞的物质世界联系在一起——这不仅是一种认识论立场，也是在一个符号过度膨胀的时代里最可靠的生存策略。
最后需要说明本文论证的边界。第一，本文是解释性的理论分析，其历史谱系是一种解释框架，并未排除其他解释；关于心学的历史作用与"李约瑟问题"的成因，学界存在实质性分歧，本文的立场只是其中之一。第二，本文对三个人才阶段的刻画是类型学的概括，不应被理解为对任何具体群体中每一个体的判断；每一阶段都有大量例外。第三，本文对当代西方学术与政策的描述基于公开报道与二手文献，未作一手实证。第四，本文所提出的制度建议方向明确但操作细节有待展开，尤其是"以现实效果评价"在不同学科中的具体化，本身就是一个需要长期探索的难题。这些局限并不削弱本文的核心论点，但提示读者：对本文的检验，同样应当是经验的而非仅仅是逻辑的。
This essay examines a recurring structural pathology in modern systems of knowledge production and governance: the substitution of internal completeness within a symbolic system (logical coherence, model elegance, metric attainment) for the investigation and testing of external reality. It first distinguishes "logical correctness" from "factual correctness" and, drawing on Goodhart's law and the image of the map devouring the territory, explains how metrics degenerate into simulacra. It then traces the epistemological divergence between "investigating things" and "cultivating the mind" in Song–Ming Chinese philosophy; the drift of the School of Mind in the late Ming; the institutional severance of knowledge from production under the scholar–peasant–artisan–merchant hierarchy; and the twentieth-century conversion of the School of Mind into an instrument of rule. It proceeds to the extreme voluntarism of the mid-twentieth century, the 1978 correction centred on practice as the criterion of truth, and the pendulum-like over-correction that replaced competence with credentials. At the contemporary level, the essay analyses the new scholar–artisan–merchant hierarchy produced by credentialism, the industrialisation of academic output and the global ranking compact that sustains it, the mathematisation of the social sciences and their record of predictive failure, and the twin degenerations of contemporary philosophy into instrumentalisation and formalisation. It extends the analysis to the West—financialisation, the inward turn of the humanities, and voluntarism in policy—as a parallel evolutionary path, and reads the near-simultaneous "from the virtual to the real" policy shifts in China and the United States as both a reaction to the pathology and a potential reproduction of its metric logic. Finally, it explains why such systems resist self-correction and proposes institutional and individual paths back to an empirical realism grounded in "seeking truth from facts" and "investigating things to extend knowledge," while stating the limits of its own argument.
Keywords: metric alienation; empirical realism; voluntarism; credentialism; research evaluation; econometrics; cognitive decline; gewu zhizhi
In the first quarter of the twenty-first century, human societies reached unprecedented heights in data collection, algorithmic complexity, and institutional standardisation. In market-oriented Western systems and state-directed Eastern systems alike, the levers of social, financial, and scientific governance were increasingly entrusted to a heavily credentialed professional class. Its promise was that precise mathematical forecasting, rigorous key performance indicators, and globally standardised measurement could manage, if not eliminate, the messy and unpredictable frictions of social life.
Confronted with the non-linear crises of the past decade or more—systemic geopolitical rupture, the weaponisation and decoupling of supply chains, fiscal contraction, demographic inflection—this calculating regime has exhibited a disquieting failure. The question Queen Elizabeth II put to economists at the London School of Economics after 2008, "Why did nobody notice it coming?", remains the most concise challenge to the predictive capacity of mainstream economics; the British Academy's subsequent forum conceded that the crisis reflected "a failure of the collective imagination of many bright people" (British Academy). At the other end of the world, a nation can lead the planet in high-quality natural-science output (Nature Index 2025) while remaining constrained for years in several critical core technologies. The two phenomena appear unrelated but share one structure: prosperity at the level of symbols coexisting with fragility at the level of reality.
The central hypothesis of this essay is that such failures are not technical errors of execution or resource allocation but a structural bias at the level of epistemology. Modern knowledge production has, to a significant degree, severed itself from the inductive, empirical investigation of the external world and withdrawn into a self-referential symbolic system in which the internal coherence of the symbol is equated with truth about reality. We call this bias the "idealist turn" in knowledge production and attempt to show that it has a deep historical lineage, stable social-structural support, identifiable contemporary forms, and foreseeable real-world consequences.
A preliminary clarification is in order. This is an interpretive, theoretical essay rather than an empirical study. The historical lineage it proposes is an explanatory framework, not a mono-causal verdict on complex historical processes; several of its historical and philosophical judgements remain contested in the scholarly literature, and the relevant sections and the conclusion will indicate those disagreements and the boundaries of the argument.
Every knowledge system faces two tests. The first is internal: Is the reasoning coherent? Does the model converge? Are the formulas error-free? Does the text conform to the required format? The second is external: Do the conclusions match the observed world? Do the predictions come true? Does the thing that was designed actually work? The first may be called logical correctness; the second, factual correctness.
In the mature natural sciences the priority between the two is unambiguous: when a mathematically beautiful theory disagrees with a reproducible experiment, the theory is abandoned. This ordering constitutes the core discipline of empirical science. The pathology this essay examines is the inversion of that ordering in particular fields of knowledge and systems of governance—when model and reality disagree, it is no longer the model that is doubted; reality is declared "noise," "friction," or "not yet at the efficient state."
A metric is a compressed representation of reality. It is useful precisely because it is simpler than reality. But once a metric becomes a target to be pursued, its correspondence to reality loosens and eventually breaks. The British economist Charles Goodhart, discussing monetary policy in 1975, observed that any statistical regularity tends to collapse once it is used as a control target—an observation later generalised as "when a measure becomes a target, it ceases to be a good measure" (Wikipedia: Goodhart's law). The anthropologist Marilyn Strathern used exactly this formulation in 1997 to describe how audit culture was reshaping British universities (Strathern 1997). Earlier still, the social psychologist Donald Campbell noted that the more any quantitative social indicator is used for decision-making, the more subject it becomes to corruption pressures and the more apt it is to distort the social processes it was meant to monitor (Wikipedia: Campbell's law).
The common implication is that governance by metrics has an endogenous tendency toward self-subversion. The historian Jerry Muller called the generalisation of this tendency across education, medicine, finance, and academia "the tyranny of metrics" (Princeton University Press); the political scientist James C. Scott argued that the simplifying schemes modern states impose to render society "legible," when combined with over-confidence in scientific rationality, frequently end in failure (Wikipedia: Seeing Like a State). This essay borrows the term simulacrum to name the terminal point of metric alienation: a symbol that no longer refers to any external reality yet is managed and pursued as though it were reality itself.
Parallel to metric alienation is a second epistemological bias: voluntarism—the belief that subjective will, spiritual force, or moral purity can override material conditions and objective laws. It looks like the opposite of metric worship—one venerates numbers, the other venerates will—but the two are structurally cognate: both treat an internal state (the coherence of a model, or the intensity of a will) as a sufficient grasp of the external world.
It must be stressed that rejecting voluntarism is not denying the role of will. Effort, ambition, and perseverance are necessary conditions of any achievement; but they are the engine, not the track. Given recognition of and adherence to objective laws, will determines how far one travels; absent that condition, the stronger the will, the greater the deviation. The boundary between will and law, as later sections argue, is key to understanding several of the twentieth century's great governance catastrophes.
Song-dynasty Neo-Confucianism achieved a relatively complete epistemology in Zhu Xi (1130–1200). Its core proposition is that li (principle) exists objectively in all things, and that human knowledge must proceed through gewu—the item-by-item investigation and accumulation of concrete things—before it can "extend knowledge" (zhizhi). Zhu Xi's supplementary commentary on the Great Learning states the path explicitly: for every thing under heaven, proceed from the principle one already knows and press further toward its limit; after long exertion, one day there will be a sudden breakthrough. The formulation concedes two points: truth lies outside the mind and must be sought outward; and knowing is a gradual, cumulative process requiring sustained effort.
One should not exaggerate the resemblance between Zhu Xi's system and modern experimental science—his "things" included ethics and canonical texts, and his method involved neither controlled experiment nor mathematisation. But in its basic epistemological orientation, Cheng–Zhu learning preserved an outward-looking stance that acknowledged the priority of external regularities over subjective cognition. That stance kept a channel open between intellectual activity and the empirical world.
In the mid-Ming, Wang Shouren (Yangming, 1472–1529) came to his realisation at Longchang that "the way of the sage is complete in my own nature; to have sought principle in external things was a mistake," and founded the School of Mind on that basis. Its three core propositions—mind is principle, knowledge and action are one, extend the innate moral knowing—relocated truth from outside the mind to within it. The famous exchange in the Chuanxilu about flowers blooming on a cliff ("Before you looked at these flowers, they and your mind were both in silence; when you came to look at them, their colours at once became clear") is routinely cited as the canonical statement that "there is nothing outside the mind."
For the present argument, "unity of knowledge and action" deserves particular scrutiny, because it admits two readings with sharply different practical consequences.
The first is the ontological reading, and Wang's own intention: knowledge and action are one and the same event; there is no such thing as "knowing without acting"; if one has not acted, one's "knowledge" was not real knowledge. The weight of this reading falls on inner sincerity, and its cultivation is inward self-examination.
The second is the evidential reading, the one most often adopted in later political and managerial practice: what a person says is unimportant; what matters is what he does; action is the only reliable test of cognition and intention. This reading converts "unity of knowledge and action" into a results-based audit of behaviour whose weight falls on externally observable conduct.
The distinction matters because the second reading strips away the metaphysical core of the School of Mind and retains only a practical principle compatible with empiricism—"test speech by action." That principle has no idealist colouring in itself; but once embedded in relations of power it degenerates readily into an instrument for auditing subordinates' loyalty. Section 3.5 examines the canonical case.
After the School of Mind gained wide influence in the mid-to-late Ming, its epigones drifted into what historians call "wild Chan" (kuang chan)—sudden enlightenment in place of study, discussion of mind and nature in place of practical administration. The most forceful critique came from the statecraft scholars of the Ming–Qing transition. Gu Yanwu wrote in the Rizhilu: "The pure talk of old discussed Laozi and Zhuangzi; the pure talk of today discusses Confucius and Mencius… They do not practise the texts of the six arts, do not examine the institutions of the hundred kings, do not grasp the affairs of the present age… They replace the real learning of cultivating oneself and governing others with the empty words of illuminating the mind and seeing one's nature. The limbs grow idle and all affairs fall to ruin; the claws and teeth are lost and the four quarters descend into chaos. The divine land is overturned; the ancestral altars become mounds of earth" (China Confucius Net). Yan Yuan's summary is sharper still: "In peacetime they fold their hands and discuss mind and nature; in crisis they repay the sovereign with a single death" (Yan Yuan ji, "Xue bian I").
The point of these critiques is not whether the metaphysics of the School of Mind is sound, but the governance consequences of an intellectual orientation: when a society's most talented people are directed toward inward moral examination rather than outward empirical inquiry, the state will find, in the face of famine, fiscal collapse, and military invasion—crises requiring specialised knowledge—that it lacks the corresponding reserves of capability.
Fairness requires noting that historical assessments of the School of Mind diverge. Many scholars emphasise Yangming learning's emancipatory significance for individual agency and its influence on the activists of Japan's Meiji Restoration; the School of Mind itself contains a strong orientation toward "tempering oneself in affairs." This essay does not attempt a global judgement of Wang Yangming's thought. It argues only that in the course of its socialisation, popularisation, and politicisation, the inward tendency within the School of Mind was amplified and combined with particular institutional structures to produce the consequences the statecraft scholars described.
An epistemological orientation does not operate in a vacuum; it requires social-structural support. In traditional China that structure was the hierarchy of shi, nong, gong, shang—scholars, peasants, artisans, merchants. Its decisive feature was that those who commanded writing and interpretive authority (the shi) sat at the top, while those who actually dealt with materials, craft, capital, and logistics (the gong and shang) sat at the bottom. The result was an institutional severance between knowledge and production: those with cultural resources did not face the material world, and those who faced the material world had no cultural resources. Artisans' practical experience, lacking social standing, could not be systematised into theory and was transmitted orally from master to apprentice; merchants' capital, under status discrimination, flowed not into machinery and reinvestment but into land and examination degrees.
This analysis echoes part of Joseph Needham's own explanation of why modern science did not arise in China. Needham observed that the education of the scholar-official class fostered an antagonism toward manual work, and that the stability of the bureaucratic order inhibited the formation of an independent merchant class (ChinaFile). It must be stated plainly, however, that the "Needham question" has multiple explanatory traditions—institutions and property rights, geography and resource endowments (as in Kenneth Pomeranz's argument about coal and overseas colonies), demography and factor prices—and the "epistemological orientation plus status hierarchy" account stressed here is one among them, not an exclusive explanation.
Yet it is precisely this account that is most illuminating for the present. The structure it describes—symbol-manipulators above, material practitioners below, and no feedback loop between them—has a structural counterpart in today's knowledge institutions. That is the subject of Section 5.
Chiang Kai-shek (1887–1975) was the twentieth century's most prominent political admirer of the School of Mind. He combined a severe discipline of personal cultivation—decades of early rising, meditation, and diary self-examination, together with abstention from tobacco and alcohol—with public veneration of Wang Yangming's doctrine of "unity of knowledge and action." In 1950 he renamed Taipei's Grass Mountain "Yangmingshan"; Taiwan's official archives record that the renaming reflected his "especial esteem for Wang Yangming's doctrine of vigorous action" and was intended to "inspire patriotic sentiment and a spirit of action among the whole people" (National Archives Administration, Taiwan).
The case bears on this essay in two ways.
First, it shows how the evidential reading of "unity of knowledge and action" was instrumentalised within relations of power. Facing an organisation riven by factions, in which professed loyalty and actual conduct frequently diverged, a ruler needed a yardstick that could cut through speech and judge loyalty directly by behaviour. "Unity of knowledge and action" supplied it: you profess the doctrine yet fail to execute orders, which proves you do not truly know, which is to say you are not truly loyal. In this usage the School of Mind is no longer a metaphysics about the location of truth; it is a compliance test.
Second, it shows the separation between personal virtue and institutional capacity. Even in the historical narratives of his adversaries, Chiang's personal austerity and self-discipline are rarely questioned; yet this individual-level cultivation did not translate into organisational-level governing capacity, and his regime's systemic failures in administration, finance, and grassroots control were not mitigated by the leader's personal virtue. The separation confirms exactly the statecraft scholars' critique: inward cultivation, however sincere, cannot substitute for outward, specialised governing knowledge and institution-building.
In the mid-twentieth century voluntarism reached a historic extreme in China. On 27 August 1958 the People's Daily published, under the headline "As Bold as People Dare, So Much Will the Land Yield" (ren you duo da dan, di you duo da chan), a letter from investigators in Shouzhang County, Shandong, describing a "high-yield campaign" aiming at ten thousand or even tens of thousands of jin per mu (Phoenix TV). The slogan became the emblem of an era.
Within this essay's framework, the core error of the period was a misplacement of experience: the direct transplantation of military and political-struggle experience—in which will, discipline, and sacrifice can indeed compensate substantially for material inferiority—into economic construction and science. Crop cycles, smelting temperatures, and the laws of population reproduction do not respond to slogans and resolve. Once "laymen leading experts" was established as a principle and those who pointed out scientific common sense were labelled "conservative," the system lost its capacity for feedback from reality. The great famine of 1959–1961 was the direct consequence of this systemic failure.
The period is formally remote from late-Ming "wild Chan," but structurally continuous with it: both substituted an internal state (moral purity or revolutionary will) for inquiry into external laws, and both institutionally punished those who insisted on looking outward.
On 11 May 1978 the Guangming Daily published the special commentator's article "Practice Is the Sole Criterion for Testing Truth," asserting that "the criterion for testing truth can only be social practice" and that every theory must continually submit to that test (Government of China). The epistemological significance of the article was to return the authority of judgement from internal dogma to external reality.
Personnel policy shifted in step. The restoration of the national university entrance examination in 1977 rebuilt a mechanism of knowledge-based selection; in December 1980 Deng Xiaoping formulated at a Central Work Conference the full standard that cadres should be "revolutionised, younger, better educated, and more professional," adding that "a complete cadre system must gradually be established to guarantee this" (People's Daily Online). "Respect knowledge, respect talent" and "science and technology are the primary productive force" became the new policy vocabulary. The direction of this correction was right: it acknowledged that governing a modern economy requires specialised knowledge, and that such knowledge derives from the study of and practice within objective laws rather than from political fervour.
A society that has long suppressed knowledge, however, often lacks a healthy mechanism for screening competence when it turns to respecting knowledge, and so substitutes visible credentials for competence that is hard to observe. Looking back over the past four decades, three successive stages can be identified, each adopting some credential as a proxy for ability, and each proxy diverging to a greater or lesser degree from actual ability.
Stage one: academic credentials in the gap years. In the 1980s, after a decade's interruption of higher education, a severe talent gap opened within the system. Under the policy pressure for "younger, better-educated" cadres, a university diploma or a period of overseas residency could suffice for exceptional promotion. The education this cohort received was heavily constrained by objective conditions: curricula were largely outdated Soviet-style systems; economics meant political economy with almost no treatment of market mechanisms or modern finance; foreign-language instruction lacked real contexts; frontier journals and original textbooks were scarce. What this cohort genuinely mastered was often the organisational skill of surviving and advancing within the system, not the professional knowledge modern governance requires. In that specific historical vacuum, however, it became the only cohort in half a century to achieve large-scale upward mobility almost on the strength of credentials alone, reaching the upper levels of the power structure by the end of the 2000s. It ended the era of governing by political campaign and sustained a pragmatic line centred on economic construction; but the limits of its professional capacity also left their mark in the corruption and stalled institutional reform of the same period.
Stage two: overseas degrees. From the late 1990s into the 2000s, large numbers of state-sponsored and self-funded students returned from abroad to universities, research institutes, financial agencies, and policy think tanks. This cohort brought back updated knowledge, but also a new dogmatic tendency: treating models and institutional paradigms learned in Western classrooms as universal standards while remaining insensitive to the historical, legal, and social foundations on which those paradigms rested, and lacking long-term participatory knowledge of how their own society operated at the grassroots. The result was detachment from reality of another kind—not inward cultivation but outward transplantation. In higher education this cohort drove the shift of evaluation toward "SCI and top journals above all," marginalising indigenous research oriented to actual engineering and industrial problems.
Stage three: metric credentials. As degrees and overseas experience depreciated through diffusion, evaluation turned to finer-grained indicators: publication counts, journal tiers, impact factors, citations, talent titles, grant volumes. This is the subject of Section 6.
The common feature of all three stages is pendular motion: from anti-intellectualism to credential worship, from closed dogma to imported dogma, never arriving at a screening mechanism centred on the ability to solve real problems. Each correction targeted the defects of the previous stage, yet reproduced the same structure under a new proxy: credentials are visible and competence is not, so credentials replaced competence.
The shi–nong–gong–shang structure described in Section 3.4 reappears today in refined form. It may be called technocratic credentialism.
At the top of this structure stands the "new shi" class—academic administrators, think-tank experts, and policy officials fluent in international scholarly discourse and metric management—who command policy-making, research funding, and academic evaluation. Their authority derives from credentials and symbolic capital rather than from any record of solving problems in the material world. Beneath them are the "new gong"—hardware and software engineers, physical-technology developers who deal daily with the laws of physics, code defects, process tolerances, and supply-chain friction—and the "new shang"—entrepreneurs and front-line operators who bear real bankruptcy risk in markets.
The problem with this structure is not the division of labour itself but the absence of a feedback loop. In a healthy system, those who make policy and allocate resources must continuously receive signals from practitioners and revise their judgements accordingly. But when promotion and evaluation at the top depend wholly on indicators internal to the symbolic system—whether documents are compliant, papers published, rankings rising—signals from the material world are systematically filtered out. People at the top can decide resource allocation in fields where they have never understood a production line, a system architecture, or a market risk. Society's hands and brain are thereby separated: those who face reality have no decision-making power, and those who have decision-making power do not face reality.
A further consequence is adverse selection. Just as the traditional shi could gain status through essays rather than administrative ability, the contemporary new shi can gain promotion through metrics rather than results. The system comes imperceptibly to favour those adept at metrics over those adept at real work, who then promote successors like themselves.
Global university rankings (QS, Times Higher Education) and bibliometric indicators (SCI inclusion, impact factor, Nature Index) define a "world-class university" in quantifiable, comparable terms. These indicators are not meaningless, but they are compressed representations of academic quality and therefore fully subject to Goodhart's law: once they become targets, optimisation around them decouples them from what they were meant to measure.
Over the past two decades, several East Asian universities—first in Hong Kong and Macau, then the leading institutions of mainland China—achieved striking ranking ascents under these rules. The mechanism is not mysterious: concentrate resources in highly cited fields, stimulate top-journal publication with large bonuses, expand citation networks through international collaboration, and publish in English to interface with the evaluation system. In data terms the approach has been enormously successful. China's share in the Nature Index for the natural sciences surpassed that of the United States from 2022 and has continued to widen its lead (Nature; Nature Index 2025).
At the micro level, ranking competition has generated a highly specialised production model. A principal investigator with international standing secures funding, cultivates relations with journals and reviewer networks, and endorses outputs as corresponding author; mid-career faculty organise and manage; large numbers of doctoral students and postdocs generate data, tune parameters, and write texts to the format preferences of target journals. The model establishes a stable exchange among the parties: the young obtain degrees, promotions, and recommendation letters; the leader obtains more funding and higher titles; the institution obtains rankings and reputation.
The problem is not the division of labour but its reshaping of research goals. Under piece-rate performance regimes, incremental outputs that can be replicated in volume over short horizons—change a material, adjust a parameter set, reproduce a hot topic—are the optimal "product," whereas original exploration requiring a decade of obscurity and carrying a high failure rate, or sustained assault on an engineering bottleneck that yields no publications in the near term, becomes a burden on team performance. Researchers genuinely committed to the latter face marginalisation. This is the metric regime's adverse selection of the research agenda.
Its extreme form is the paper mill. In 2023 global retractions exceeded ten thousand for the first time, a record (Nature); a 2026 bibliometric study traced 10,409 papers retracted for paper-mill problems through 2024, with by far the highest concentration from China (Scientometrics). Paper mills are not an inevitable product of ranking competition, but they are the same incentive structure deformed at the bottom.
It must be stressed that this is not unilateral "fraud" by any one country but a global compact in which every party gets what it wants. Students need top-journal publications as passports to overseas graduate study, because the only "lingua franca" overseas supervisors can read is a publication record; university administrators need rankings as achievements; ranking agencies need wide adoption to sustain commercial value; international journals and publishers earn revenue from processing charges and database subscriptions. In this closed loop, each party's rational choice reinforces the decoupling of metrics from reality. The only party outside the loop, yet bearing the ultimate cost, is public finance and the real industrial base.
China began dismantling the "papers, titles, degrees, and awards above all" orientation as early as 2018. In February 2020 the Ministry of Education and the Ministry of Science and Technology jointly issued Opinions on Regulating the Use of SCI Paper Indicators in Universities, requiring institutions to "sever the direct linkage between SCI-related indicators and resource allocation and performance bonuses" and stating that "the number of SCI papers and impact factors should not be used as restrictive conditions for graduation and the conferral of degrees" (Ministry of Education); the Ministry of Science and Technology's parallel Measures for Breaking the Undesirable Orientation of "Papers Above All" in Research Evaluation introduced representative-work and classified evaluation (MOST).
Implementation has met structural resistance, for reasons detailed in Section 9. The core may be stated here: if an entire system's funding, promotion, degree conferral, and institutional ranking rest on one set of indicators, removing them means the system temporarily loses its evaluative standard; and the existing evaluators were themselves selected under the old standard, lacking both the capacity to judge by the new one (real value and social impact) and the motive to do so.
That mathematics is the foundational language of every exact science needs no argument. But mathematics is itself a discipline of internal logic and deductive inference whose propositions are true or false irrespective of whether factories, famines, or wars exist in the world. When a discipline whose object is society elevates mathematics from tool to purpose—treating model elegance, derivational rigour, and parameter significance as the final criteria of research value—it commits a category error: substituting logical correctness for factual correctness.
This critique is not an outsider's prejudice but a long-standing self-examination within economics. In his 1974 Nobel lecture "The Pretence of Knowledge," Hayek attributed economists' repeated policy failures to their inclination "to imitate as closely as possible the procedures of the brilliantly successful physical sciences," an attitude he called "scientistic" and, in the true sense, "unscientific" (Nobel Prize). Wassily Leontief, founder of input–output analysis, complained in a 1982 letter to Science that theorists "continue to produce scores of mathematical models and to explore in great detail their formal properties" while econometricians "fit algebraic functions of all possible shapes to essentially the same sets of data," without perceptible advance in understanding the structure and operation of a real economic system (Science). In 2015 Paul Romer introduced the term "mathiness" in the American Economic Review, criticising theoretical work in which the use of mathematical symbols "lets academic politics masquerade as science"—unlike genuine mathematical theory, mathiness "leaves ample room for slippage between statements in the languages of words as opposed to symbols" (AER).
Sociologically, mathematisation also performs a protective function within the discipline. A scholar who discusses tariffs, local debt, or supply chains in ordinary language can be argued with by anyone possessing common sense and industry experience; once the same argument is packaged in dozens of pages of differential equations and regression tables, external criticism is effectively excluded. Mathematics here is not an instrument of discovery but a barrier of status.
The price of that protection is persistent predictive failure. Before 2008, mainstream macro models almost universally failed to foresee systemic risk; the supply-chain ruptures and inflationary shocks of the 2020s again exceeded model expectations. Models cohere on paper yet repeatedly misfire in reality, and the discipline's mainstream response has typically been to add parameters and patch details rather than abandon core assumptions. This is precisely the inverted ordering described in Section 2.1: reality treated as noise in the model, rather than the model treated as an approximation of reality.
The pathology is not confined to economics. Philosophy—once the discipline at the summit of human knowledge, charged with generalising the operation of the world and the method of living—exhibits two degenerations in opposite directions.
The first is instrumentalisation. The most active and best-funded areas of contemporary philosophy—AI ethics, algorithmic governance, gene-editing ethics, the moral dilemmas of autonomous vehicles—are to a large degree the "safety codes" and "brake pads" of technological development. Such work has value, but it is applied ethics, not high-level generalisation about the world and human life. When a philosophy department's survival depends on supplying ethical endorsement to technology projects, the discipline descends from overseer to servant.
The second is formalisation. At the opposite pole, the languages of mathematics and physics are imported into metaphysics and epistemology—formal epistemology, Bayesian epistemology, structural realism, the relativity-based "block universe." These research programmes are technically refined, but several of their claims warrant caution. The block-universe view, for example, holds that past, present, and future exist equally within a four-dimensional spacetime block and that the passage of time is a subjective illusion; pressed to the level of human life, its practical implication becomes hard to distinguish from ancient fatalism—choice, effort, and opportunity become "a picture already painted." A discipline whose vocation is reflection on the human condition should pause when its frontier conclusions amount in practice to the cancellation of human agency. More generally: if a question about the world can be answered by the equations of physics, it belongs to physics; philosophy's distinctive value lies precisely in handling questions that cannot be formalised yet are decisive for how humans understand themselves and act. To found every discipline on mathematics is to misunderstand what "hard" means—mathematics supplies the underlying tool, not the core of every discipline.
The two degenerations appear opposite but share a root: both evade philosophy's most difficult and most important task—to produce generalisations about the operation of human society and history that possess height and depth and can be tested against experience.
The pathology examined here belongs to no single country or system. The West has travelled a parallel road over the past half-century, and the near-simultaneous policy adjustments in China and the United States in recent years reveal similar responses by two systems to the same disorder.
The eighteenth- and nineteenth-century West established its modern position through systematic investigation of the material world—experimental science, engineering, industrial organisation, logistics, infrastructure. Over the past four decades, however, a substantial share of the West's best minds was channelled into financial engineering: designing derivatives, optimising balance sheets, manufacturing value on paper through buybacks and divestitures. Financial symbols were treated as ultimate reality while the manufacturing base was systematically hollowed out. Structurally this is the same thing as the East's substitution of publication metrics for technological breakthroughs: completeness at the level of symbols standing in for capability at the level of matter.
Simultaneously, a considerable part of the Western humanities and social sciences turned inward in a new way: making the purification of language, the parsing of identity, and the performance of moral position the core of scholarly value, while growing increasingly resistant to empirical evidence—whether from biology, economics, or statistics. Structurally the tendency resembles late-Ming "wild Chan": both adopt an inner moral purity as the criterion of judgement, and both incline to punish those who insist on citing external facts. This essay passes no value judgement on particular scholarly positions; it observes only that when a discipline's internal standards decouple from external evidence, it loses the capacity for self-correction.
Nor is voluntarism an Eastern monopoly. When a Western leader believes that personal negotiating will and tariff instruments can reverse within a short span decades of accumulated industrial hollowing and structural debt, the epistemological structure differs in no essential way from "as bold as people dare, so much will the land yield"—both substitute intensity of will for respect for law. The regularities of the global economy, the inertia of demographic structure, and the momentum of supply-chain migration do not respond to political resolve.
Notably, China and the United States have in recent years adopted almost simultaneous policy adjustments "from the virtual to the real," though with differing motives and means.
On the Chinese side: according to Ministry of Education data, between 2021 and 2025 universities revoked or suspended 12,200 undergraduate programmes and added about 10,200, a cumulative adjustment exceeding 30 per cent, with arts, foreign-language, and management programmes most often cut and new programmes concentrated in artificial intelligence and advanced manufacturing (Caixin Global; Forbes). From March 2026 the Chinese Academy of Sciences stopped using academic funds and central government allocations to pay publication fees for thirty high-priced international open-access journals, including Nature Communications, Cell Reports, and Science Advances (China.org.cn). Together with the 2020 "breaking the five onlys" documents, these measures share one direction: severing the direct link between indicators and resources.
On the American side: in February 2025 the National Institutes of Health announced a uniform 15 per cent cap on indirect-cost rates for research grants, far below the 30–70 per cent previously negotiated with individual institutions (Congressional Research Service), a move later blocked in court; in October 2025 the White House presented nine universities with a Compact for Academic Excellence in Higher Education offering preferential access to federal funds in exchange for accepting a set of policy demands, which several universities rejected (Reuters; Inside Higher Ed). Under the pressure of federal funding uncertainty, research universities including Harvard, Columbia, Chicago, Brown, and Princeton reduced or paused admissions to some doctoral programmes for 2026–27, with the humanities and social sciences especially affected (Chicago Maroon).
The synchronisation is not accidental. Both countries face the twin pressures of fiscal constraint and geopolitical competition and no longer have the slack to sustain symbol production that cannot be converted into substantive capability; the personal trajectories of both current leaderships lie outside the traditional academic path, and neither identifies naturally with a credentialed knowledge class.
The limits of these adjustments must nonetheless be stated. First, abolishing disciplines and cutting funding by administrative fiat is itself a form of governance by metrics—replacing "paper counts" with "employment rates" or "alignment with national priorities"—and does not touch the underlying problem of evaluative capacity. Second, this essay's subject is the epistemological orientation of knowledge, not disciplinary labels: a social science or philosophy that faces experience and can be tested against reality is closer to "the real" than an engineering programme that produces only publishable, non-functioning outputs. Equating "from the virtual to the real" with "cut the humanities, keep the sciences" misreads the problem and risks destroying, along with the bubble, the very disciplines that should carry the task of high-level generalisation. Third, cutting funding does not automatically generate new capability; if the evaluators who survive the cuts remain products of the old system, new resource allocation will simply reproduce old behaviour under new indicators.
The pathology described here is widely recognised in every country, and reform documents are not scarce. Why, then, is change so difficult? At least four mutually reinforcing mechanisms are at work.
First, path dependence and capability lock-in. Those selected under the old evaluation system possess, as their core skills, precisely what scores highly within it: writing compliant documents, producing publishable papers, managing indicators and networks. Were the standard suddenly to shift to "how much verifiable value has been created in reality," those skills would depreciate sharply, and their holders possess no substitute. Incumbents therefore have strong motives to defend the old standard and to legitimise that defence with more sophisticated language ("basic research cannot be judged by utilitarian criteria," "scholarship requires patience in obscurity"). Such language is not wholly without merit, but its function here is defensive.
Second, mutual hostage-taking. Superiors need reportable numbers; subordinates need attainable tasks. Assessing whether a research team has genuinely solved a problem requires evaluators with professional judgement who accept responsibility for being wrong; counting papers and rankings is both easy and safe. Both sides thus reach a tacit settlement within "numerical bureaucracy": controllable numbers mask uncontrollable reality.
Third, the reverse cycle of selection. Those genuinely able to solve real problems are often at a disadvantage in a system whose sole standard is metrics—they are unwilling or unskilled at devoting energy to metric optimisation and more likely to leave after friction with the system. Those who remain and advance continue to select successors by the same standard. Over time the distribution of competence inside the system diverges systematically from that outside it.
Fourth, the masking effect of dividends. For decades, the dividends of globalisation and demography and the infrastructure accumulated by previous generations allowed systems to bear the cost of symbol production for long periods without visible consequence. This gave incumbents the illusion that the system was functioning well. Only when dividends are exhausted and external constraints tighten does the gap between symbol and reality surface in the form of crisis.
Together these mechanisms explain why reform documents proliferate while behaviour does not change, and why genuine correction tends to come not from internal awakening but from the compulsion of external reality—technological blockade, fiscal exhaustion, employment crisis. This is not an optimistic conclusion, but it is an honest one.
This essay has sought to show that from the popularisation of late-Ming mind-learning, through mid-twentieth-century voluntarism, through the reform-era pendulum that replaced competence with credentials, to the contemporary industrialisation of scholarship that replaces results with metrics and the social sciences that replace experience with mathematics, there runs an identifiable structural thread: the substitution of completeness in some internal state for the investigation of external reality. The thread is not uniquely Chinese; Western financialisation, the inward turn of the humanities, and voluntarism in policy constitute a parallel path. Their shared consequence is the collective blindness of governing and intellectual classes before non-linear crises of the material world, and a resulting decline in cognitive capacity that crosses the boundaries of political systems.
That decline is becoming visible now because the "reality credits" that sustained symbol production—infrastructure accumulated by earlier generations, the dividends of globalisation and demography—are running out. The laws of thermodynamics, the scarcity of critical resources, the inertia of demographic structure, and the entropy of economic systems do not read policy documents and do not respond to econometric models.
If the diagnosis holds, the way out lies not in more refined metrics but in restoring the feedback loop between knowledge and reality. This implies:
First, shifting the centre of evaluation from symbolic output to verifiable real-world effect—whether a study has solved a genuine engineering or social problem, whether a policy has produced observable improvement in implementation, whether a model has withstood out-of-sample reality. Such evaluation is harder, demands professional judgement, and requires evaluators to bear responsibility; but it is the only standard whose "rules" cannot be worked out and gamed.
Second, rebuilding the channel between practitioners and decision-makers, so that those who directly confront physical and market laws—engineers, entrepreneurs, front-line researchers—hold substantive voice in resource allocation and agenda-setting rather than serving merely as objects of evaluation.
Third, distinguishing "virtual" from "real" by epistemological orientation rather than disciplinary label. A philosophy that generalises the human condition in ways historical experience can test is "real learning"; an engineering programme that produces only publishable, non-functioning results is "virtual learning." Abolition by discipline may preserve what is wrong and discard what is right.
Fourth, keeping will in its proper place. Will is necessary, but it must operate within the recognition of law. Governance that substitutes resolve for research and slogans for argument, under whatever system it appears, has had its ending rehearsed repeatedly by history.
For individuals living through this era, the analysis also has consequences. In an environment that over-values credentials, rankings, and discourse, the core method for preserving cognitive independence is to insist on real-world effect rather than symbolic status as the basis of judgement: look at what a person does rather than says; at the practical consequences of a claim rather than the elegance of its expression; at what an institution has solved rather than where it ranks. This is the empiricism-compatible core of the evidential reading of "unity of knowledge and action," and it is what "seeking truth from facts" and "investigating things to extend knowledge" mean at the level of the person. To keep one's intellect continuously and unconditionally tied to the real, rough, rhetoric-indifferent material world is not only an epistemological stance; in an age of symbolic over-expansion it is also the most reliable strategy for survival.
Finally, the boundaries of the argument should be stated. First, this is an interpretive theoretical analysis; its historical lineage is an explanatory framework that does not exclude others, and on the historical role of the School of Mind and the causes behind the Needham question there is substantive scholarly disagreement, of which this essay's position is one. Second, the characterisation of three stages of talent is a typological generalisation and should not be read as a judgement on every individual in any group; each stage contains many exceptions. Third, the descriptions of contemporary Western scholarship and policy rest on public reporting and secondary literature, not first-hand empirical work. Fourth, the institutional proposals are clear in direction but their operational detail remains to be developed; in particular, the concrete form of "evaluation by real-world effect" across different disciplines is itself a difficult problem requiring long exploration. These limits do not weaken the central thesis, but they remind the reader that the test of this essay, too, should be empirical rather than merely logical.