# The AI Trust & Fairness Compendium

> The AI Trust & Fairness Compendium by validant.ai: 12 standalone parts with 115 fixed chapter numbers and three appendix volumes on AI fairness, non-discrimination, explainability and assurance, written from identified sources, reviewed and fact-checked. One notation and one vocabulary for the whole work. Version 1.1 of 2026-11-06. Public, Creative Commons Attribution-NonCommercial 4.0 International (https://creativecommons.org/licenses/by-nc/4.0/).

The compendium informs professional judgment; it is not legal advice for a specific case and does not certify any AI system. Law is stated as of November 2026. Cite a chapter with its number, its part and the part's version and date, as printed on the part's first page. Commercial use, including training commercial models on the text, needs a word with hello@validant.ai; the machine-readable terms are at https://validant.ai/license.xml (RSL 1.0).

Checked before publication: 2268 reference entries looked up, 2867 claims and 861 formulas rechecked. 2750 pages in 19 printed volumes. The chapter text as pages and Markdown files follows; until then each part's PDF is the published text. Downloads: 5 files per visitor are free (the front matter is not counted); an email address at https://validant.ai/compendium unlocks every file.

## Front matter

- [Front Matter and Master Table of Contents (PDF, 38 pages)](https://validant.ai/compendium/pdf/Validant_Compendium_Front_Matter_and_Contents.pdf): The way in: preface, executive overview of every part, scope and limitations, the editorial method and reading paths by task, with the full table of contents.

## Part I: Foundations of Trust, Justice and Non-Discrimination

Fairness in AI is a normative and historical question before a technical one: trust must follow evidence of trustworthiness, the choice of fairness conception must be explicit, and documented past discrimination shows which risks to test for.

- [Part I, Chapters 1 to 11 (PDF, 181 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_I_Foundations.pdf): Chapter 1: The Nature of Trust; Chapter 2: Trust in Artificial Intelligence; Chapter 3: Conceptions of Fairness and Justice; Chapter 4: Ethical Frameworks for Fairness Evaluation; Chapter 5: Equality, Equity and Human Dignity; Chapter 6: Historical Patterns of Discrimination in Technology; Chapter 7: Data Representation and Social Context; Chapter 8: Modern Manifestations of Historical Biases; Chapter 9: Human Rights, Human Agency and Contestability; Chapter 10: Fairness, Explainability and Accountability; Chapter 11: The Historical Context Assessment Method

## Part II: Formal Foundations of Algorithmic Fairness

Fairness criteria are precise statistical statements about predictions, outcomes and groups; the mathematics shows which of them can hold together, but choosing among them remains a normative decision.

- [Part II, Chapters 12 to 22 (PDF, 185 pages, version 1.4 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_II_Formal_Foundations.pdf): Chapter 12: Formalizing Algorithmic Decision-Making; Chapter 13: Probability and Statistical Foundations; Chapter 14: Independence: Demographic Parity and Related Criteria; Chapter 15: Separation: Equal Opportunity and Equalized Odds; Chapter 16: Sufficiency: Predictive Parity and Calibration; Chapter 17: Individual Fairness; Chapter 18: Causal and Counterfactual Fairness Criteria; Chapter 19: Impossibility Theorems; Chapter 20: Fairness Trade-offs; Chapter 21: Fairness Beyond Binary Classification: Scores, Regression and Ranking; Chapter 22: Contextual Selection of Fairness Criteria

## Part III: Bias, Discrimination and Causality

Bias enters an AI system at identifiable points, from the data and the model to deployment and later shift; group metrics show that a disparity exists, while causal models explain why and where an intervention can remove it.

- [Part III, Volume 1: Chapters 23 to 29 (PDF, 139 pages, version 1.2 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_III_Bias_Discrimination_Causality_Vol1.pdf): Chapter 23: Direct, Indirect and Systemic Discrimination; Chapter 24: Protected Characteristics, Sensitive Attributes and Proxies; Chapter 25: Data Collection and Representation Biases; Chapter 26: Algorithm Design and Implementation Biases; Chapter 27: Feedback Loops and Amplification Effects; Chapter 28: System Interaction and Deployment Biases; Chapter 29: Intersectionality and Compound Discrimination
- [Part III, Volume 2: Chapters 30 to 35 (PDF, 119 pages, version 1.2 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_III_Bias_Discrimination_Causality_Vol2.pdf): Chapter 30: From Correlation to Causation; Chapter 31: Building Causal Models for Bias Detection; Chapter 32: The Counterfactual Fairness Framework; Chapter 33: Causal Inference with Limited Data; Chapter 34: Distribution Shift and Emerging Discrimination; Chapter 35: The Bias Source Identification Method

## Part IV: Measuring Fairness: Metrics and Statistical Evidence

A fairness metric becomes evidence only when it is computed correctly, reported with its denominators and uncertainty, tested for robustness across every group including small ones, and scoped to what quantitative measurement can establish.

- [Part IV, Chapters 36 to 45 (PDF, 200 pages, version 1.1 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_IV_Measuring_Fairness.pdf): Chapter 36: Fairness Assessment Methodology; Chapter 37: Comprehensive Data Auditing; Chapter 38: Bias Detection in Raw Data; Chapter 39: Group Fairness Metrics; Chapter 40: Implementing Fairness Metrics; Chapter 41: Statistical Validation and Confidence Intervals; Chapter 42: Statistical Significance and Robustness; Chapter 43: Small Samples and Sparse Subgroups; Chapter 44: The Fairness Metrics Selection Method; Chapter 45: Limits of Quantitative Fairness Assessment

## Part V: Fairness Interventions

Fairness interventions act on the training data, on training itself or on a model's outputs; the right point follows from where a disparity enters, and the price of each method must be measured on the specific system.

- [Part V, Volume 1: Chapters 46 to 52 (PDF, 141 pages, version 1.1 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_V_Interventions_Vol1.pdf): Chapter 46: Choosing an Intervention Strategy; Chapter 47: Reweighting and Resampling; Chapter 48: Distribution Transformation; Chapter 49: Feature Engineering and Transformation for Fairness; Chapter 50: Fairness-Aware Data Generation; Chapter 51: Fairness Objectives and Constraints; Chapter 52: Fairness-Aware Loss Functions and Regularization
- [Part V, Volume 2: Chapters 53 to 57 (PDF, 110 pages, version 1.1 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_V_Interventions_Vol2.pdf): Chapter 53: Adversarial Debiasing; Chapter 54: Multi-Objective Optimization; Chapter 55: Threshold Optimization; Chapter 56: Calibration Across Groups; Chapter 57: Prediction Transformation Methods

## Part VI: Fairness Engineering and Operations

Fairness becomes dependable only when it is executable: measured with well-chosen tools, tested automatically before release, monitored per group in production, reported to the people who decide, and changed only after controlled experiments.

- [Part VI, Chapters 58 to 64 (PDF, 135 pages, version 1.1 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_VI_Engineering_Operations.pdf): Chapter 58: The Fairness Tooling Ecosystem; Chapter 59: Fairness in the Development Workflow; Chapter 60: Automated Bias Checks in CI/CD; Chapter 61: Metric Tracking and Real-Time Bias Detection; Chapter 62: Drift Detection and Alerting; Chapter 63: Fairness Dashboards and Reporting; Chapter 64: A/B Testing for Fairness

## Part VII: Fairness in Advanced AI Architectures

In advanced architectures unfairness lives in learned representations, exposure, generated content, fused modalities and multi-step actions, so fairness must be assessed inside the system, over time and across the ecosystem, not only on final decisions.

- [Part VII, Chapters 65 to 69 (PDF, 107 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_VII_Advanced_Architectures.pdf): Chapter 65: Deep Learning Systems; Chapter 66: Recommendation and Ranking Systems; Chapter 67: Large Language Models; Chapter 68: Multi-Modal and Vision Models; Chapter 69: Generative and Agentic Systems

## Part VIII: Explainability, Interpretability and Non-Discrimination

Explanations can help detect, contest and remedy unfairness, but they are evidence only when faithful to the model, robust, tested by group and understood by their audience; they show how a model works, not that it is fair.

- [Part VIII, Volume 1: Chapters 70 to 76 (PDF, 140 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_VIII_Explainability_Vol1.pdf): Chapter 70: Foundations of AI Explainability; Chapter 71: Explainability and Fairness; Chapter 72: Taxonomy of Explainability Methods; Chapter 73: Feature Attribution Methods; Chapter 74: Counterfactual Explanations and Algorithmic Recourse; Chapter 75: Causal Explainability; Chapter 76: Group-Level and Intersectional Explanations
- [Part VIII, Volume 2: Chapters 77 to 82 (PDF, 106 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_VIII_Explainability_Vol2.pdf): Chapter 77: Explanation Quality, Fidelity and Robustness; Chapter 78: Human-Centered Explainability; Chapter 79: Explainability and Legal Rights; Chapter 80: Explainability in Generative and Agentic AI; Chapter 81: Explainability as Assurance Evidence; Chapter 82: Limits, Risks and Fairness of Explanations

## Part IX: Legal, Ethical and Regulatory Foundations

Law sets a mandatory but moving floor for AI fairness: it names protected grounds, defines direct and indirect discrimination and imposes risk-based and sectoral duties, yet no metric proves compliance and compliance alone does not establish fairness.

- [Part IX, Chapters 83 to 92 (PDF, 195 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_IX_Legal_Regulatory.pdf): Chapter 83: Legal Standards for Algorithmic Fairness; Chapter 84: International Human Rights and Non-Discrimination Law; Chapter 85: The Global Regulatory Landscape; Chapter 86: The EU AI Act and GDPR Article 22; Chapter 87: Swiss Legal and Regulatory Framework; Chapter 88: United States Anti-Discrimination Frameworks; Chapter 89: International Standards and Governance Principles; Chapter 90: Risk Classification and Assessment Systems; Chapter 91: Sector-Specific Obligations; Chapter 92: Ethics Beyond Compliance and Regulatory Change

## Part X: AI Assurance, Evidence and Digital Trust

A fairness metric becomes assurance only inside a scoped, argued claim backed by preserved evidence and independent checking; credentials, registries and status checks make such claims verifiable over time, but no certificate shows that a system is fair in general.

- [Part X, Chapters 93 to 101 (PDF, 176 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_X_Assurance_Digital_Trust.pdf): Chapter 93: From Fairness Measurement to Assurance; Chapter 94: Claims, Evidence and Assurance Arguments; Chapter 95: Evidence Collection and Audit Trails; Chapter 96: Independent Assessment and Reproducibility; Chapter 97: Verifiable Credentials and Digital Attestations; Chapter 98: Identity, Provenance and Evidence Integrity; Chapter 99: Validity, Revocation and Continuous Assurance; Chapter 100: Trust Registries and Distributed Trust Infrastructures; Chapter 101: Limits of Certification and Assurance

## Part XI: Fairness in Organizational Practice

Fairness becomes organizational practice only when delivery, roles, documentation, governance and incident handling carry explicit fairness content, so that bias work is planned, owned, recorded, decided and repaired rather than left to goodwill.

- [Part XI, Chapters 102 to 111 (PDF, 186 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_XI_Organizational_Practice.pdf): Chapter 102: Integrating Fairness into Agile Principles; Chapter 103: Fairness User Stories and Acceptance Criteria; Chapter 104: Sprint Planning and Execution for Fairness; Chapter 105: Fairness Ceremonies and Checkpoints; Chapter 106: Roles and Responsibilities for Fairness; Chapter 107: Documentation and Communication Frameworks; Chapter 108: Governance Mechanisms and Decision Processes; Chapter 109: Incident Management, Remediation and Redress; Chapter 110: The Fairness Audit Playbook: an Integrated Workflow; Chapter 111: Applied Case Studies

## Part XII: Emerging Challenges and Research Frontiers

At the frontier, fairness is no longer a property of one model's decisions but of evidence about open-ended outputs, delegated action and systems acting over time and across ecosystems; valid methods for producing that evidence are still being built.

- [Part XII, Chapters 112 to 115 (PDF, 77 pages, version 1.0 of 2026-11-05)](https://validant.ai/compendium/pdf/Validant_Compendium_Part_XII_Research_Frontiers.pdf): Chapter 112: Fairness in Large Language Models and Generative AI; Chapter 113: Autonomous Agents and Delegated Decisions; Chapter 114: Adaptive Systems and Collective Impacts; Chapter 115: An Integrated Research Agenda

## Appendix volumes

- [Appendix Volume 1: Glossary, Notation and Index (PDF, 139 pages)](https://validant.ai/compendium/pdf/Validant_Compendium_Appendices_Vol1_Glossary_Notation_Index.pdf): The controlled glossary (Appendix A), the mathematical notation (B) and the index of every definition with its chapters (L).
- [Appendix Volume 2: Metrics, Methods and Worked Examples (PDF, 158 pages)](https://validant.ai/compendium/pdf/Validant_Compendium_Appendices_Vol2_Metrics_Methods_Examples.pdf): The catalogs of fairness metrics (C) and explainability methods (D), the statistical methods (E) and worked numerical examples to recompute (H).
- [Appendix Volume 3: Law, Standards, Templates, Evidence and Bibliography (PDF, 218 pages)](https://validant.ai/compendium/pdf/Validant_Compendium_Appendices_Vol3_Law_Standards_Templates_Bibliography.pdf): The legal and regulatory matrix (F), standards (G), method worksheets and templates (I), the evidence taxonomy (J) and the bibliography of every source (K).

## Reading paths by task

- Start a fairness assessment of a system: start with Chapter 36: Fairness Assessment Methodology; continue with Chapter 110: The Fairness Audit Playbook: an Integrated Workflow, and its four methods.
- Identify historical risks for a use case: start with Chapter 11: The Historical Context Assessment Method; continue with Chapter 6: Historical Patterns of Discrimination in Technology; Chapter 7: Data Representation and Social Context; Chapter 8: Modern Manifestations of Historical Biases.
- Choose fairness criteria: start with Chapter 22: Contextual Selection of Fairness Criteria; continue with Chapters 14 to 20.
- Locate where bias enters: start with Chapter 35: The Bias Source Identification Method; continue with Chapter 25: Data Collection and Representation Biases; Chapter 26: Algorithm Design and Implementation Biases; Chapter 27: Feedback Loops and Amplification Effects; Chapter 28: System Interaction and Deployment Biases; Chapter 29: Intersectionality and Compound Discrimination.
- Audit data before training: start with Chapter 37: Comprehensive Data Auditing and Chapter 38: Bias Detection in Raw Data; continue with Chapter 24: Protected Characteristics, Sensitive Attributes and Proxies; Chapter 43: Small Samples and Sparse Subgroups.
- Choose and specify metrics: start with Chapter 44: The Fairness Metrics Selection Method; continue with Chapter 39: Group Fairness Metrics; Chapter 40: Implementing Fairness Metrics; Chapter 41: Statistical Validation and Confidence Intervals; Chapter 42: Statistical Significance and Robustness; Chapter 43: Small Samples and Sparse Subgroups, and Appendix C.
- Report a quantitative finding: start with Chapter 45: Limits of Quantitative Fairness Assessment; continue with Chapter 41: Statistical Validation and Confidence Intervals; Chapter 42: Statistical Significance and Robustness.
- Choose an intervention: start with Chapter 46: Choosing an Intervention Strategy; continue with Chapters 47 to 57, by intervention point.
- Fix a model that cannot be retrained: start with Chapter 55: Threshold Optimization and Chapter 56: Calibration Across Groups and Chapter 57: Prediction Transformation Methods; continue with Chapter 46: Choosing an Intervention Strategy.
- Automate fairness checks: start with Chapter 59: Fairness in the Development Workflow and Chapter 60: Automated Bias Checks in CI/CD; continue with Chapter 58: The Fairness Tooling Ecosystem.
- Monitor a deployed system: start with Chapter 61: Metric Tracking and Real-Time Bias Detection and Chapter 62: Drift Detection and Alerting; continue with Chapter 63: Fairness Dashboards and Reporting; Chapter 64: A/B Testing for Fairness.
- Assess a large language model or agent: start with Chapter 67: Large Language Models and Chapter 69: Generative and Agentic Systems; continue with Chapter 112: Fairness in Large Language Models and Generative AI; Chapter 113: Autonomous Agents and Delegated Decisions.
- Use explanations in an assessment: start with Chapter 76: Group-Level and Intersectional Explanations and Chapter 77: Explanation Quality, Fidelity and Robustness; continue with Chapter 71: Explainability and Fairness; Chapter 81: Explainability as Assurance Evidence.
- Meet a legal duty to explain: start with Chapter 79: Explainability and Legal Rights; continue with Chapter 86: The EU AI Act and GDPR Article 22.
- Decide which obligations apply: start with Chapter 90: Risk Classification and Assessment Systems and Chapter 91: Sector-Specific Obligations; continue with Chapter 85: The Global Regulatory Landscape; Chapter 86: The EU AI Act and GDPR Article 22; Chapter 87: Swiss Legal and Regulatory Framework; Chapter 88: United States Anti-Discrimination Frameworks.
- Make a claim others can rely on: start with Chapter 93: From Fairness Measurement to Assurance; continue with Chapter 94: Claims, Evidence and Assurance Arguments; Chapter 95: Evidence Collection and Audit Trails; Chapter 96: Independent Assessment and Reproducibility.
- Judge a certificate or audit report: start with Chapter 101: Limits of Certification and Assurance; continue with Chapter 97: Verifiable Credentials and Digital Attestations; Chapter 98: Identity, Provenance and Evidence Integrity; Chapter 99: Validity, Revocation and Continuous Assurance; Chapter 100: Trust Registries and Distributed Trust Infrastructures.
- Organize fairness work in a team: start with Chapter 102: Integrating Fairness into Agile Principles; continue with Chapter 103: Fairness User Stories and Acceptance Criteria; Chapter 104: Sprint Planning and Execution for Fairness; Chapter 105: Fairness Ceremonies and Checkpoints.
- Set up fairness governance: start with Chapter 106: Roles and Responsibilities for Fairness; continue with Chapter 107: Documentation and Communication Frameworks; Chapter 108: Governance Mechanisms and Decision Processes; Chapter 109: Incident Management, Remediation and Redress.

## Optional

- [The start page as Markdown](https://validant.ai/compendium/index.md): the front matter's "Why this compendium", the ways in and the contents
- [Structure as JSON](https://validant.ai/compendium/index.json): parts, chapters, reading paths, questions, readers, documents
- [Controlled glossary as JSON](https://validant.ai/compendium/glossary.json): 933 terms, each with its definition and the chapters that define it
- [Licence terms for crawlers (RSL)](https://validant.ai/license.xml)
- [validant.ai](https://validant.ai/llms.txt): the site's own index