Democratic Accountability for Government AI
"No Decision Without Justification" — A decision without explanation is an enemy of democracy.
Your loan application was denied. Why?
Your welfare benefits were terminated. Why?
Your parole was rejected. Why?
"Because the AI decided so" is not an answer democracy can accept.
Every citizen has the right to know the reasons behind decisions that affect them.
Every decision must be subject to appeal and judicial review.
PAP cryptographically records every AI decision in public administration —
documenting when, what data was used, which rules were applied, and why that conclusion was reached —
in a tamper-proof format.
Current challenges in public AI systems and how PAP addresses them
AI declares "ineligible" without explaining why. Citizens cannot even begin to appeal without understanding the basis for the decision.
AI trained on historically discriminatory data perpetuates and reproduces unfair decisions systematically.
Unelected algorithms make decisions that affect citizens' fundamental rights, without democratic oversight or accountability.
Benefits eligibility determination
[Applicant] → [AI Decision: "Ineligible"] → [Appeal] → [Request for Decision Rationale]
↓
"The AI's reasoning is proprietary information."
"Details of the scoring model cannot be disclosed."
↓
The citizen loses all means to contest the decision.
Criminal justice risk assessment
[Defendant] → [AI Risk Score: "High Risk"] → [Bail Denied / Harsher Sentence]
↓
Post-hoc analysis revealed:
"ZIP code" and "arrest history" were primary factors
→ Effectively using proxies for race and socioeconomic status
↓
No mechanism existed to verify this at the time of the decision.
Visa application processing
[Visa Applicant] → [AI Review: "Denied"] → [Reason: "Comprehensive Assessment"]
↓
"Which factors were problematic?" Unknown
"What should be improved?" Not indicated
"Will reapplication yield the same result?" Unpredictable
Target systems and recorded events
Data flow and system structure
┌─────────────────────────────────────────────────────────────┐
│ Public Administration System │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Citizen │──▶│ AI/ML │──▶│ Decision │ │
│ │ Data │ │ Scoring │ │ Output │ │
│ └──────────┘ └────┬─────┘ └──────────┘ │
│ │ │
│ ┌───────▼───────┐ │
│ │ PAP Logger │ ← Records all decision events│
│ │ (Sidecar) │ │
│ └───────┬───────┘ │
│ │ │
└──────────────────────┼───────────────────────────────────────┘
│
┌───────▼───────┐
│ Hash Chain │ ← Cryptographically chained
│ + Signature │ ← Digitally signed
│ + Factor Log │ ← Complete decision factor record
└───────┬───────┘
│
┌───────▼───────┐
│ Audit Archive │ ← Long-term storage
│ (GDPR Compliant)│ ← Access control
└───────┬───────┘
│
┌───────▼───────┐
│Citizen Portal │ ← Self-information disclosure
│(Accountability)│ ← Appeal support
└───────────────┘{
"event_id": "019234ab-7c8d-7def-8123-456789abcdef",
"timestamp_ns": 1734567890123456789,
"event_type": "ELIGIBILITY_DECISION",
"agency_id": "AGENCY_XXXXX",
"provenance": {
"actor": {
"type": "AI_MODEL",
"identifier": "benefit_eligibility_v2.3.1",
"model_hash": "sha256:abc123...",
"training_date": "2024-03-15",
"bias_audit_date": "2024-06-01",
"bias_audit_result": "PASSED"
},
"input": {
"applicant_id_hash": "sha256:applicant_anonymized...",
"data_sources": [
{"source": "TAX_AUTHORITY", "freshness_days": 30},
{"source": "EMPLOYMENT_REGISTRY", "freshness_days": 7},
{"source": "RESIDENCE_REGISTRY", "freshness_days": 1}
],
"input_hash": "sha256:def456...",
"missing_fields": ["PRIOR_BENEFIT_HISTORY"],
"imputation_method": "CONSERVATIVE_DEFAULT"
},
"context": {
"applicable_law": "SOCIAL_WELFARE_ACT_2023_SEC_12",
"policy_version": "2024-Q2",
"budget_constraint_active": false,
"human_override_available": true
},
"action": {
"decision": "ELIGIBLE",
"confidence": 0.89,
"eligibility_score": 72.5,
"threshold_applied": 65.0,
"explainability": {
"method": "SHAP",
"top_contributing_factors": [
{"factor": "income_below_threshold", "contribution": 0.35, "direction": "positive"},
{"factor": "dependent_children", "contribution": 0.25, "direction": "positive"},
{"factor": "employment_status", "contribution": 0.20, "direction": "positive"},
{"factor": "asset_level", "contribution": -0.10, "direction": "negative"}
],
"counterfactual": {
"if_income_increased_by": 15000,
"then_decision_would_be": "INELIGIBLE"
}
},
"appeal_rights": {
"deadline_days": 30,
"appeal_authority": "ADMINISTRATIVE_TRIBUNAL",
"required_documentation": ["INCOME_PROOF", "EXPENSE_RECORDS"]
}
}
},
"prev_hash": "sha256:789xyz...",
"signature": "ed25519:...",
"citizen_disclosure": {
"simplified_explanation": "Your application has been approved. Main reason: You meet income criteria.",
"full_explanation_available": true,
"explanation_request_url": "/citizen/explanation/019234ab..."
}
}PAP alignment with international regulations
| Regulation | Jurisdiction | Requirement | PAP Support |
|---|---|---|---|
| EU AI Act Article 6(2) | EU | AI affecting fundamental rights classified as high-risk | ✅ Full Support |
| EU AI Act Article 13 | EU | Transparency and explainability requirements | ✅ Full Support |
| EU AI Act Article 14 | EU | Human oversight obligation | ✅ Full Support |
| GDPR Article 22 | EU | Right to contest automated decision-making | ✅ Full Support |
| GDPR Article 15 | EU | Right of access (self-information disclosure) | ✅ Full Support |
| US Executive Order 14110 | USA | AI safety and fairness requirements | ✅ Aligned |
| NYC Local Law 144 | NYC | Audit obligation for hiring AI | ✅ Full Support |
| Illinois BIPA | Illinois | Biometric information use restrictions | ✅ Supported |
| Administrative Procedure Act | Japan | Obligation to provide reasons for administrative decisions | ✅ Planned |
| Annex III Category | Examples | PAP Support |
|---|---|---|
| Employment and Worker Management | Hiring AI, Termination Risk Assessment | ✅ Required |
| Access to Essential Public Services | Welfare Benefits, Public Housing | ✅ Required |
| Creditworthiness Assessment | Loan Review, Credit Score | ✅ Required |
| Emergency Services Dispatch | Ambulance/Fire Priority | ✅ Required |
| Migration, Asylum, Border Control | Visa Review, Refugee Status | ✅ Required |
| Justice and Democratic Processes | Sentencing Support, Recidivism Risk | ✅ Required |
Detection, recording, and mitigation of bias
Transparency portal and multi-layered explanation model
| Layer | Target Audience | Explanation Detail |
|---|---|---|
| Citizen | General applicants | Plain language summary (simplified_explanation) |
| Representative | Lawyers, advocates | Detailed decision factors and contributions |
| Auditor | Regulators, audit firms | Model specs, training data, bias audit results |
| Technical | AI developers, researchers | Complete logs, model hashes, reproduction steps |
Practical scenarios for PAP application
| Phase | Without PAP | With PAP |
|---|---|---|
| Denial Notice | "Ineligible based on comprehensive assessment" | "Income criteria met, but assets exceeded threshold" |
| Reason Inquiry | "Details cannot be disclosed" | All decision factors viewable via citizen portal |
| Appeal Preparation | Unclear what to contest | Can specifically point out "asset valuation error" |
| Administrative Review | Dismissed due to insufficient evidence | PAP evidence proves asset valuation logic issue |
| Outcome | Resignation | Re-review approval + system improvement contribution |
[Company] → [Annual hiring AI audit obligation]
↓
↓
[Publication of audit report] → [Compliance certification]
[Defendant] → [AI Risk Score: "High Risk"] → [Harsher sentence]
↓
[5-year follow-up study]
↓
↓
[System improvement / Basis for reviewing past judgments]
Balancing transparency and privacy
"We want to record everything for accountability"
↕
"We want to minimize data for GDPR/privacy"
Key technical parameters
| Item | PAP Specification |
|---|---|
| Timestamp Precision | Seconds to milliseconds (depending on use case) |
| Event Recording Frequency | Per decision event |
| Hash Algorithm | SHA-256 |
| Signature Algorithm | Ed25519 (Dilithium support planned) |
| Personal Data Protection | AES-256-GCM encryption + Crypto-Shredding |
| Retention Period | Per legal requirements (typically 5-10 years, longer for judicial) |
| Citizen Access | Self-information disclosure via portal |
| Bias Audit | Periodic fairness metric calculation and recording |
PAP development timeline
Framework hierarchy and profile positioning
┌────────────────────────────────────────┐
│ VAP (Verifiable AI Provenance) │
│ Cross-domain parent framework │
└──────────────────┬─────────────────────┘
│
┌──────────────────▼─────────────────────┐
│ VSO (VeritasChain Standards Org) │
│ Standards body │
└──────────────────┬─────────────────────┘
│
┌──────┬──────┼──────┬──────┐
▼ ▼ ▼ ▼ ▼
[VCP] [DVP] [MAP] [EIP] [PAP]
Finance Auto Medical Energy Public
↓ ↓
v1.0 Planned
Released| Characteristic | VCP (Finance) | MAP (Medical) | DVP (Automotive) | EIP (Energy) | PAP (Public) |
|---|---|---|---|---|---|
| Impact Speed | Milliseconds | Minutes-Hours | Seconds | Minutes-Days | Months-Years |
| Reversibility | Partial | Irreversible | Irreversible | Slow Recovery | Difficult |
| Primary Victim | Investors | Patients | Occupants/Pedestrians | Citizens | Individual Rights |
| Regulatory Focus | Market Integrity | Patient Safety | Physical Safety | Continuity | Democracy |
| Appeal Rights | Limited | Limited | Limited | Limited | Required |
PAP is currently under development. We welcome participation from government agencies, municipalities, human rights organizations, and AI ethics researchers.
"Power without explanation is the enemy of democracy. AI is no exception."
— VeritasChain Standards Organization
"The right to an explanation is the foundation of due process in the algorithmic age."
This specification is licensed under CC BY 4.0 International
Provider: VeritasChain Standards Organization (VSO) | https://veritaschain.org