祝融说。

What This Institutional Model Can Show

How common rules expand into an overall evaluation, and whether errors can return to the decision-makers who hold authority. This volume belongs to philosophical argument and thought experiment. Terminology, normative requirements, and empirical material play different roles: normative requirements need reasons and counterexamples; empirical conclusions need corresponding research or records; the setting of a story shows only how conditions combine. Writing an empirical claim in the form of "possibly" does not discharge the responsibility to state one's grounds.

The Load-Bearing Structure of the Five-Layer Cage

This volume employs three layers of argument. The first layer is the definitions RC gives of limited order, power hierarchy, and evaluation alienation. The second layer is the bridging hypothesis of this volume: when the same evaluation spans multiple domains of life and connects to resources of high consequence, those under evaluation adjust their visible behavior, and the system continues to train the evaluation on these behaviors. The third layer is the empirical judgment: whether a real institution forms total obedience must be demonstrated through the actual connections among standards, data, resources, prediction, and appeals, and cannot be derived directly from the existence of a score or a platform.

The loop the book tests is: a local standard crossing domains → the binding of life chances → behavioral adaptation → predictive shaping → information sorting and lateral isolation → executors sustaining themselves by indicators. Task relevance, the continuity of basic resources, independent review, lateral comparison, and appeals entering decisions are the positions where the loop can be interrupted. The institution can still evaluate and execute; the difference lies in whether it can acknowledge that the object, the scope, and the results of an evaluation were once wrong.

Key Components of the Institutional Cross-Section

RC: the author's philosophical framework adopted by this book, denoting the construction of reality through observational convergence. Citations of the theory indicate the provenance of ideas; they do not automatically prove social or psychological mechanisms.

Limited order: RC 3.1 understands order as a set of certainties formed in interaction, coordinable and operable. This volume therefore does not treat rules themselves as a cage; it asks after the concrete steps by which a rule moves from solving a finite task to evaluating the complete subject.

Power hierarchy: the structural differences described in RC 3.2, appearing as differences among positions in the capacity to lock in possibility, to define problems, and to transmit the consequences of evaluation. This volume identifies it through chains of authorization, resource connections, and the authority to correct errors, rather than inferring the magnitude of power directly from job titles.

Evaluation alienation: RC 3.3 describes the process by which dynamic consensus degenerates into one-way certification and the evaluation power and the bearing of outcomes fall out of alignment. This volume grounds it in a checkable institutional cross-section: when a local standard crosses domains, when a score becomes bound to resources, and when prediction begins to shape the predicted.

Conceptual consensus: in RC 2.4, consensus arises from the provisional stabilization of interpretive frameworks in comparison. On this basis the volume distinguishes shared understanding, procedural majority, and no one continuing to object; all three can support action, yet none can impersonate the others.

Proxy indicator: a recordable object used to answer a finite question; it cannot stand for the whole of value merely because it lends itself to comparison.

Authorization: the scope within which a position is permitted to decide on particular matters; success in implementation cannot be used to retroactively infer the original authorization.

Punitive backflow: turning the filing of an appeal itself into an unfavorable label or an unrelated restriction; it does not include the backflow of data needed to carry corrections out.

Value unification / predictive discipline: the former denotes a local indicator crossing domains to become the overall scale of value and qualification; the latter denotes prediction shaping its objects through resources, surveillance, and sanctions, and then vindicating the judgment by its own consequences. The two are carried by Chapters 7 and 10 respectively, and neither characterizes common standards or risk prediction as such as technologies of obedience.

The RC terms above and the analytic vocabulary of this volume are working definitions within the author's framework. Words such as labor, care, statistics, and feedback carry specialized uses in their respective disciplines, and the inferences of this book cannot be treated as disciplinary conclusions merely because the names coincide.

The Five-Layer Index from Spreadsheet to Appeal

• Chapter 1: How a Spreadsheet Becomes Order: the spreadsheet first solved a real problem; exceptions are not merely a nuisance, nor are they naturally right.
• Chapter 2: Why Common Rules Are Indispensable: that no one has laid down rules does not mean that no one dominates; the entry of exceptions and the entry of privilege.
• Chapter 3: Unanimity and Observational Consensus: four kinds of the common that look alike; silence has no single translation.
• Chapter 4: The Scope of Application of Standards: what question a standard must first answer; the evidential basis of a standard can expire.
• Chapter 5: Why Delegation Cannot Cancel Accountability: delegation transfers a finite right of action; executors are not without judgment.
• Chapter 6: What Silence Actually Expresses: silence and procedural acceptance; how silence is processed into consensus.
• Chapter 7: When a Single Score Governs All Value: how an indicator oversteps its original question; the consequences of combining scores with resources.
• Chapter 8: The Convenience of Classification and the Attrition of Persons: how categories arrange subsequent handling; exceptions, identity, and self-classification.
• Chapter 9: Life Chances and Evaluation Binding: evaluations of high consequence and rightful standing; the loop of error, appeal, and consensus.
• Chapter 10: How Prediction Changes the Predicted: how prediction intervenes in its object; changes in disclosure, surveillance, and enforcement.
• Chapter 11: Why Executors Too Lose Their Available Margin: the tasks, time, and authority of the front line; failed escalation and the transfer of hardship.
• Chapter 12: Why Institutions Can Never Hear Themselves: from on-site information to institutional summaries; how messages lose their consequences and their owners.
• Chapter 13: Openness, Surveillance, and the Boundaries of Privacy: public reasons and continuous surveillance; privacy, use, and the authority to observe.
• Chapter 14: Can Repeated Voices Stand for the Majority: the subjects who precede counting; how visibility manufactures a majority.
• Chapter 15: When Each Person Sees Only Their Own Version: information sorting and the common master copy; the conditions of exchange and the power to sort.
• Chapter 16: When Members Can Speak Only to the Center: what lateral ties provide; the pressure of isolation and power within the network.
• Chapter 17: Procedural Participation and Substantive Influence: the authority, the influence, and the timing of participation; the conditions under which experience enters decisions.
• Chapter 18: No One Responsible, Everyone Obedient: error across interfaces, categories, and hierarchies; outsourcing, indicators, and the allocation of responsibility.
• Chapter 19: Standards That Never Expire: how a standard outgrows the conditions of its validity; expiry, triggers, and stability.
• Chapter 20: Appeals Are Not Another Obedience Test: from acceptance of a case to a reasoned decision; once error is established, how far the correction must travel.
• Chapter 21: Bringing Consequences Back to the Decision Table: why decisions and consequences come apart; attribution, evaluation, and the boundaries of retrospection.
• Chapter 22: Preserving Multiple Scales of Value: the question of scales after the total score withdraws; parallel scales and the conditions of conversion.
• Chapter 23: How Exceptions Become Knowledge for Improvement: how exceptions become material for rules; recording and the handling of recurring problems.
• Chapter 24: Common Life with Disagreement: the disagreement that remains after procedure; the place of disagreement, scales, and responsibility.

Notes on Sources by Chapter

The theoretical entries are for checking the concepts used in this volume. The chapter numbers below list only the places where the text explicitly discusses the terms in question; they do not automatically mark the whole theory as evidence for every chapter. The causal and normative inferences of the thought experiments that follow remain the responsibility of each chapter's own conditions and counterexamples.

• For "observational consensus" as involved in Chapters 3, 4, and 6: see RC 1.3.

• For "processual completeness" as involved in Chapters 1, 2, and 3: see RC 2.1.

• For the "bidirectional representation of language" as involved in Chapters 1, 2, and 3: see RC 2.2.

• For "theoretical dimensional reduction" as involved in Chapters 1, 2, 3, and 5: see RC 2.3.

• For "conceptual consensus" as involved in Chapters 1, 2, 3, 5, and 6: see RC 2.4.

• For "limited order" as involved in Chapters 1, 2, and 4: see RC 3.1.

• For "power hierarchy" as involved in Chapter 1: see RC 3.2.

• For "evaluation alienation" as involved in Chapters 1, 2, and 4: see RC 3.3.

• For the "continuation of stability" as involved in Chapter 2: see RC 4.2.

• For "sustainable decision" as involved in Chapter 1: see RC 4.3.

• For the survey methodology of Chapter 14: AAPOR, Best Practices for Survey Research; see the parts on sampling frames, question design, nonresponse, and disclosure of methods. It is used to limit the inference from repeated voices to overall opinion, and does not directly furnish proof of political representation or relations of authorization.

• For the governance of models in Chapter 10: NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0), see the four core functions of Govern, Map, Measure, and Manage; U.S. Government Accountability Office, Artificial Intelligence: An Accountability Framework for Federal Agencies and Other Entities, see the four principles of governance, data, performance, and monitoring. The two frameworks support only the claim that continuous documentation, role responsibility, monitoring, and review belong to the system lifecycle; they do not prove that the predictions in the synthetic cases have already produced discipline, nor do they replace the legal obligations of any concrete institution.

• For the uses of data in Chapter 20: OECD, Guidelines on the Protection of Privacy and Transborder Flows of Personal Data, Part Two, paragraphs 9–10 (purpose specification and use limitation). The chapter's distinction among raw material, the results of correction, and statistical material is a concrete application; it does not treat the Guidelines as the current adjudicative rule of any jurisdiction. Substantive review may uphold the original decision, and the normative analysis of burden of proof and delay does not impersonate an introduction to law.

Bibliography

The following materials provide specific sources; the scope of their support is governed by the notes above and by the main text.

• Zhu Rong, RC, version checked on September 8, 2026. Section numbers and uses are given in "Notes on Sources by Chapter" above.
• Zhu Rong, Processual Realism, version checked on September 8, 2026. See especially 3.4 on ideas and consensus, 4.1 on limited order, 4.2 on power hierarchy, and 4.3 on evaluation alienation; it is used for the institutional cross-section, not as a substitute for material on actual authority and consequences.
• Elham Tabassi, 2023, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1. Here we adopt only the framework in which risk governance runs through design, deployment, use, and evaluation, and in which Govern, Map, Measure, and Manage require continuous connection; the voluntary general framework is not written up as the compliance conclusion of any particular institution. Used in: Chapter 10: How Prediction Changes the Predicted.
• U.S. Government Accountability Office, 2021, Artificial Intelligence: An Accountability Framework for Federal Agencies and Other Entities, GAO-21-519SP. Here we adopt only the limited requirements that governance, data, performance, and monitoring must jointly enter accountability and that deployed systems require continued observation; we do not judge on this basis that any fictional model has already produced discrimination or control. Used in: Chapter 10: How Prediction Changes the Predicted.
• American Association for Public Opinion Research, Best Practices for Survey Research, accessed September 8, 2026. See question wording, sampling frames, nonresponse, and disclosure of methods; the boundaries of authorization in the chapter are normative argument. Used in: Chapter 14: Can Repeated Voices Stand for the Majority.
• OECD, Guidelines on the Protection of Privacy and Transborder Flows of Personal Data, the principles of purpose specification and use limitation in Part Two, paragraphs 9–10, accessed September 8, 2026. Here we borrow the distinction of data uses; we do not adjudicate on this basis whether any concrete institution is lawful. Used in: Chapter 20: Appeals Are Not Another Obedience Test.

Note on Theoretical Versions

The theoretical section numbers in this volume follow the author's texts as checked on September 8, 2026. The website may continue to be revised; when citing, please record the title, the section number, and the date of consultation together, and when checking sources, readers should locate them by these readable details rather than substituting repository hashes for them.