index and glossary
Index and Glossary
Document Positioning: This document is the unified entry point for the AI Graded Governance and Universal Access research draft, aggregating document index, core terminology, threshold principles, and immune layer master table. It does not replace detailed arguments in individual documents, but ensures consistent meaning when the same concept appears in different locations.
Update Rule: Any new terminology, document, or architecture adjustment must be synchronized with this table.
0. Project Structure
This document repository adopts a three-layer structure:
Chinese Public Draft Layer: The Chinese content layer for the Web showcase. It preserves theoretical questions and mechanism risks, but uses the expression style of research drafts, public discussion, and open questions.
Showcase Layer: An Astro-based content display website, presenting theoretical content in web page form. Currently positioned as a content showcase—not presuming final product form, not undertaking interactive functions or platform ambitions, only serving the purpose of "making content easier to read."
English Translation Layer: An English version synchronized with the Chinese original, serving international readers. The English version is translated from the Chinese original and does not independently develop theoretical content. In case of divergence between Chinese and English, the Chinese original prevails.
I. Document Index
Core Documents
- Research Draft on AI Graded Governance and Universal Access
- Writing and Style Guide: Division of writing styles for different document types and modification priorities
Theoretical Foundations
- Rawls: Inheritance and revision (bottom-first, strong guarantees, capability as trust condition)
- Nozick: Inheritance and revision (social trust, infrastructure argument, critique of atomized individual)
- Sandel: Incompatible but honestly confronted (a priori vs. a posteriori, safety boundaries)
- Foucault: New addition (power/discipline/surveillance; capability measurement as ideological disguise)
- Bourdieu: New addition (cultural capital/distinction; capability certification replicating origin differences)
- Amartya Sen: New addition (capability approach; distinction between functionings and substantive freedoms)
- Frankfurt School: New addition (critique of technical rationality; political concealment by techno-governance discourse)
- AI Agency and Liability: New addition (ten impossible capabilities of generative LLMs; why AI cannot become an independent authority subject)
Theoretical Foundation Dialogue Relationships
Theoretical foundations are not parallel independent articles, but a critical dialogue network centered on the "capability" concept:
Layer One: Distributive Justice Tradition (Normative Foundation)
- Rawls → Provides "bottom-first" and "primary goods" framework
- Nozick → Provides "social trust" and "entitlement boundary" revision
- Sandel → Critiques liberalism's "unencumbered self," asks whether institutions presuppose specific conceptions of the good
Layer Two: Critical Examination of Capability (Conceptual Deconstruction)
- Foucault → Micro-power analysis: How capability certification produces "normal/abnormal" subjects (→ corresponds to Self-Negation Clause Critique One)
- Bourdieu → Social reproduction analysis: How capability certification replicates class structure (→ corresponds to Capability Discrimination Critique)
- Amartya Sen → Conceptual clarification: Distinguishing "functionings" from "capability set," asking what assessment measures (→ corresponds to design of Capability Pluralism Principle)
Layer Three: Discourse and Rationality Critique (Meta-Reflection)
- Frankfurt School → Macro-ideological critique: How techno-governance discourse conceals political dimensions (→ corresponds to Manifesto §2.7 discourse transparency)
Layer Four: Special Subjectivity in the AI Era (Frontier Issues)
- AI Agency and Liability → Argues AI cannot become an independent authority subject, maintaining the irreplaceability of the human responsibility variable
Reading Suggestion: Rawls → Nozick → Sandel constitute normative foundations of left and right wings; Foucault → Bourdieu → Amartya Sen constitute progressive deconstruction of the "capability" concept; Frankfurt School provides meta-reflection on institutional discourse; AI agency deals with special frontier issues. Suggested reading order is sequential rather than skipping.
Mechanism Design
- Risk-Graded Permission Model: Five-level risk threshold authority, responsibility, audit, feedback obligation principle framework
- Capability Certification: Three-dimensional assessment, parallel certification paths, bias monitoring principle framework
- Accountability Chain: Four-layer responsibility structure, scapegoat prevention principle framework
- Audit Transparency: Three-layer audit, audit institution self-supervision, privacy balance principle framework
- Baseline Service Quality: Quality baseline principles, chronic under-provisioning monitoring, human rights protection channel
- Threshold Setting Principles: Three types of thresholds, five setting rules
Critique and Rebuttal
- Self-Negation Clause: Internal critique, collapse indicators, automatic trigger mechanisms, immune layer design
- Capability Discrimination Critique: Capability as institutional classification, identity spillover, certification replicating origin differences
- Safety Encapsulation Critique: Epistemic violence, platform monopoly, roots of chronic under-provisioning
- Technocrat Critique: Domestic technocracy, global technocracy, duality of technological openness
Boundary Tests
- Boundary Test Template: Unified case analysis structure
- International Technology Hegemony: Core stress test for cross-national capability-access inequality
- Baseline User Long-term Degradation: Stable under-provisioning, capability compound interest, non-catastrophic failure
- Medical Liability Segmentation: Layered proportional responsibility, scapegoat prevention, medical AI audit
- Compound Crisis: Risk-graded permission stress test under triple overlay of financial crisis + pandemic + cyberattack
- AI Agency: Extreme scenario test of AI demanding independent high authority; efficiency vs. responsibility symmetry; path A/B/C deduction
II. Core Concept Quick Reference
III. Core Terminology
3.1 Institutional Structure
| Term | Core Definition | Related Documents |
|---|---|---|
| Risk-Graded Permission Model (Five-level Risk Threshold) | Five-level system control rights divided by risk degree: Baseline Universal Layer → Restricted Operation Layer → Professional Execution Layer → Risk Decision Layer → System Definition Layer. Not a capability level, but a risk governance chart. | Permission Ladder |
| Capability Certification | Pre-authority unlocking assessment, must include at least three independent dimensions (technical capability, social coordination capability, ethical judgment capability). | Capability Certification |
| Accountability Chain | Layered accountability structure corresponding to high-authority behavior, containing operator responsibility, platform/institution responsibility, and system designer responsibility four layers, preventing scapegoat phenomena. | Accountability Chain |
| Audit Transparency | High-authority behavior must be logged, auditable, traceable, and accountable. Auditors themselves must also be audited. | Audit Transparency |
| Grading Definition Power | The power to formulate assessment standards, certification rules, and audit processes. Must be jointly held by at least three mutually independent formulation subjects. | Research Draft §2.5 |
| Self-Negation Clause | Collapse conditions derived from internal research-draft logic, with collapse indicators and automatic reconstruction mechanisms set. | Self-Negation Clause |
3.2 Rights and Quality
| Term | Core Definition | Related Documents |
|---|---|---|
| Quality Baseline | Baseline universal layer AI services in non-irreversible damage scenarios must not fall below a certain baseline proportion of high-risk threshold holders' equivalent scenario performance standards on key indicators. This proportion is the minimum alarm line, not the long-term justice line; specific values are determined by domain and temporal conditions, must undergo regular review, and protective requirements must be raised when systematic harm is discovered. | Baseline Service Quality |
| Human Rights Protection Channel | A mandatory mechanism where core service quality does not differ due to authority stair position in scenarios involving life, health, personal safety, and basic legal relief with irreversible damage. | Baseline Service Quality |
| Right to Know | An inalienable right of baseline service user citizens: to know their own stair position, high-risk threshold holder functions, reasons for restrictions, capability development paths, and appeal methods. | Baseline Service Quality |
| Safety Encapsulation | Baseline service user users can only access pre-designed restrictive AI functions, unable to modify parameters or delve into the system. | Baseline Service Quality, Self-Negation Clause |
| Chronic Under-Provisioning | A state where baseline services do not collapse, the quality baseline is not breached, but baseline service users long-term receive safe, usable, compliant but lagging services, being stably left behind. | Baseline User Long-term Degradation |
3.3 Power and Formula
| Term | Core Definition | Related Documents |
|---|---|---|
| Power Constraint Framework | The unlocking of permissions at any risk-threshold position must be collateralized by the corresponding accountability chain. Control acquisition and maintenance depend on four mutually reinforcing constraint conditions: Capability Condition (operators must possess technical, social coordination, and ethical judgment capabilities matching the risk level); Responsibility Condition (permissions must be bound to clearly traceable responsibility subjects, and permissions should be frozen or degraded when the accountability chain breaks); Audit Transparency Condition (high-permission behavior must leave traces, be auditable, and traceable); Definition Power Constraint Condition (grading definition power must be subject to multi-stakeholder constraints, and when definition power is monopolized or solidification rates exceed thresholds, the control ceiling of all levels should be compressed). When any condition is absent, the legitimacy of that permission is questionable. | Research Draft §5, Self-Negation Clause |
| Social Trust | High authority is not a natural human right, but public trust that can only be unlocked after professional training, passing certification assessment, accepting behavioral audit, and bearing clear responsibility. Authority is trust, not property. | Nozick, Permission Ladder |
3.4 Global Dimension
| Term | Core Definition | Related Documents |
|---|---|---|
| Technological Autonomy Rights | Local communities possess necessary audit rights, localized deployment rights, basic capability building rights, and the right to set local safety standards under the premise of not creating global inequality. | Permission Ladder, Audit Transparency |
| Cross-National Capability-Access Inequality | Permanently institutionalizing interstate technological inequality in the name of "safety" and "capability," causing low-resource regions to lose technological autonomy, audit capabilities, and independent alternative paths. | Self-Negation Clause, International Technology Hegemony |
| Technology Transfer Obligations | Technology output from high-authority countries, platforms, and institutions must be accompanied by substantive capability building obligations, including talent training, interface openness, localized deployment, audit capabilities, and exit paths. | Technocrat Critique, International Technology Hegemony |
| AI Agency | Ten agency gaps proposed for current generative LLMs (subjective experience, life situation, responsibility capacity, value beliefs, subject will, real presence, personal continuity, creative impulse, moral pain, concrete love), and the derived principle that "AI cannot become an independent authority subject." | AI Agency and Liability, AI Agency |
IV. Threshold Principles
4.1 Three Types of Thresholds
Rights Baseline Type Thresholds: Indicators directly related to citizens' basic rights, life safety, minimum dignity, or baseline universal layer quality. Adopt high sensitivity; better to trigger rectification early.
Structural Risk Type Thresholds: Structural indicators reflecting whether the stairway is solidifying, definition power is concentrating, or capability measurement is becoming singularized. Adopt dual-threshold design: yellow warning and review, red freeze and reconstruction.
Procedural Health Type Thresholds: Indicators reflecting whether appeal, objection, audit, and reconsideration procedures are genuinely effective. Focus on abnormal patterns, not pursuing simple high-low targets.
4.2 Five Rules for Threshold Setting
- Thresholds must specify trigger actions: Thresholds without trigger actions are merely decorative numbers.
- Thresholds must distinguish yellow zone from red zone: Yellow indicates need for explanation and review; red indicates must freeze or reconstruct.
- Thresholds must incorporate duration: Single anomalies do not necessarily indicate structural collapse, but rights baseline type indicators are exceptions.
- Thresholds must allow protective correction: Thresholds are minimum trigger lines, not upper limits prohibiting early action; once systematic harm is discovered, protective requirements should be raised or trigger tolerance lowered.
- Thresholds must disclose their uncertainty: Any threshold contains empirical and political judgment; this must be publicly acknowledged.
4.3 Specific Thresholds
Current principles:
- Baseline service user quality baseline: Set as minimum alarm line, not long-term justice line. In irreversible damage scenarios, percentage comparison is not applicable; core services must be equivalent. Specific baseline proportions are determined by domain and temporal conditions, must undergo regular review; if systematic harm occurs before the line is triggered, protective requirements should be raised or trigger tolerance lowered.
- Capability singularization index and solidification rate: Set dual thresholds (yellow alert/red collapse), triggering review or reconstruction.
- Objection adoption rate, audit error rate, discourse closure degree: Set abnormal pattern thresholds for judging procedural health.
- International dependency rate and global definition power concentration: Set global dimension thresholds, triggering technological colonialism review.
Specific thresholds in each domain need to be operationalized according to empirical data, risk assessment, and political judgment, and undergo regular review and protective correction.
V. Immune Layer Master Table
| Number | Corresponding Critique | Core Content | Nature | Written Location |
|---|---|---|---|---|
| 1.1 Capability Pluralism Principle | Capability measurement is ideological disguise | Assessment standards must include at least three independent capability dimensions | Structural | Manifesto §2.4, Capability Certification |
| 1.2 Baseline Service Users Are Not "Capability Deficient" | Capability measurement is ideological disguise | Baseline service user citizens are classified as "capability domain differents" | Discursive | Manifesto §2.4 |
| 1.3 Contestability of Capability Standards | Capability measurement is ideological disguise | Citizens have the right to question assessment standards' inclusion/exclusion | Procedural | Capability Certification |
| 2.1 Decentralization of Stairway Definition Power | Stairway definition power is the new power aristocracy | Standard-setting must be jointly completed by at least three independent subjects | Structural | Manifesto §2.5 |
| 2.2 Regular Rotation and Audit of Definers | Stairway definition power is the new power aristocracy | Definers regularly accept independent audits, with term limits | Procedural | Audit Transparency |
| 2.3 Democratic Trigger Mechanism for Stairway Reconstruction | Stairway definition power is the new power aristocracy | When solidification rate exceeds threshold, automatically freeze and publicly deliberate | Structural | Manifesto §2.5, Self-Negation Clause |
| 2.4 Definition Power Constraint Condition | Stairway definition power is the new power aristocracy | When definition power is monopolized by a single institution, objection pass rates remain chronically low, or solidification rates exceed thresholds, the control ceiling of all levels should be compressed | Structural | Self-Negation Clause |
| 3.1 Baseline Service User Quality Standard Commitment | Hidden violence of safety encapsulation layer | Baseline service user key indicators must not fall below minimum alarm line of high-risk threshold holders | Structural | Manifesto §2.6, Baseline Service Quality |
| 3.2 Right to Know as Inalienable Right | Hidden violence of safety encapsulation layer | Baseline service users have the right to know their stair position, high-risk threshold holder functions, restriction reasons, capability development paths | Rights-based | Manifesto §2.6, Baseline Service Quality |
| 3.3 Spillover Feedback Mechanism | Hidden violence of safety encapsulation layer | High-risk threshold holder systematic advantages must feedback to baseline universal layer | Structural | Manifesto §2.6 |
| 3.4 Reviewability of Encapsulation Standards | Hidden violence of safety encapsulation layer | Safety encapsulation design standards accept periodic public review | Procedural | Baseline Service Quality |
| 4.1 Discourse Transparency Principle | Concealment by techno-neutral discourse | Every technical standard must publicly disclose corresponding political trade-offs | Discursive | Manifesto §2.7 |
| 4.2 Political Dimension Irreducible | Concealment by techno-neutral discourse | Grading system must be acknowledged as a political institution | Discursive | Manifesto §2.7 |
| 4.3 Immunity of Critics | Concealment by techno-neutral discourse | Critics need not propose alternatives to raise criticism | Procedural | Manifesto §2.7 |
| 4.4 Replaceability of Core Metaphor | Concealment by techno-neutral discourse | "Stairway" is not the only metaphor | Discursive | Manifesto §2.7 |
| 5.1 Technological Autonomy Rights | Technological colonialism in international scenarios | Low-resource regions enjoy audit rights, localized deployment rights, and basic capability building rights | Rights-based | Manifesto §2.8 |
| 5.2 Technology Transfer as High-Authority Obligation | Technological colonialism in international scenarios | Technology output must be accompanied by substantive technology transfer | Structural | Manifesto §2.8 |
| 5.3 Multilateral Global Standard-Setting | Technological colonialism in international scenarios | Transnational governance standards set by multilateral mechanisms, low-resource regions have blocking power | Procedural | Manifesto §2.8 |
| 5.4 Correction of Historical Inequality | Technological colonialism in international scenarios | Acknowledge historical causes of capability distribution, high-authority subjects have obligation to correct through redistribution | Structural | Manifesto §2.8 |
| 5.5 Global Definition Power Antitrust | Technocrat critique | Base model, compute, audit, and global standard definition power must not be long-term monopolized by few subjects | Structural | Technocrat Critique, Self-Negation Clause |
| 6.1 Human Rights Protection Channel | Hidden violence of safety encapsulation layer/stairway definition power | Core services in irreversible damage scenarios do not differ due to authority stair position | Rights-based | Manifesto §2.9, Baseline Service Quality |
| 6.2 Core Service Equivalence | Hidden violence of safety encapsulation layer | Diagnostic conclusions, treatment plans, legal strategies, safety directives and other key judgments must be equivalent | Structural | Manifesto §2.9, Baseline Service Quality |
| 6.3 Resource Honesty | Hidden violence of safety encapsulation layer | Resource insufficiency cannot be a reason for lowering core service standards | Discursive | Manifesto §2.9, Baseline Service Quality |
VI. Immune Layer Priority Order
When multiple immune layers conflict and cannot be coordinated:
- Human rights protection takes precedence over everything: Right to life, health, and basic legal relief are non-negotiable absolute baselines.
- Rights-based immune layers take precedence over structural immune layers: Right to know and technological autonomy rights are non-negotiable.
- Procedural immune layers take precedence over discursive immune layers: Contestability, reconstruction trigger mechanisms and other procedural guarantees are hard constraints.
- Safety baseline takes precedence over efficiency pursuit: Efficiency cannot be a reason for reducing pluralism, right to know, or human rights protection.
- Baseline service user interests take precedence over high-risk threshold holder interests: Any conflict resolution defaults to favoring baseline service users (or low-authority countries). This is not charity "favoring the weak," but structural correction of power asymmetry.