EM Foundation ← Publications
Architecture Document — v1.0

EM Foundation Systems Architecture

A unified architecture for ARIA, Assessment, Corroboration, and Trust Infrastructure — resolving overlaps, establishing canonical terminology, and mapping the complete EM Foundation trust stack.

Executive Summary

The EM Foundation's publications describe several related but partially overlapping systems: ARIA, ARIA Home, ARIA Network, ARIA Trust Ledger, Intelligence Assessment Framework, Corroboration Standard, and Assessment Charter. This document unifies them into one coherent institutional model, establishes canonical terminology, resolves documented contradictions, and maps the relationship between every component.

The Foundation is not proposing AI tools. It is proposing a trust infrastructure for advanced intelligence. The unified structure — from ARIA as the governing philosophy to the Assessment Charter as the governance constitution — forms the EM Foundation Trust Stack.

I. The EM Foundation Trust Stack

Seven layers from governance philosophy to public record

Every EM Foundation system occupies a defined position in a layered architecture. The layers are not independent — each depends on the layers below it for governance legitimacy and on the layers above it for operational purpose. The Trust Stack is the canonical description of how every Foundation system relates to every other.

Governance
Layer 7 — Governance Constitution
Assessment Charter
Governs independence, conflicts of interest, funding restrictions, reviewer conduct, score publication, appeals, disputes, provider response rights, correction policy, methodology updates, and public transparency. The governance layer that protects the integrity of every layer beneath it.
Answers: Why should anyone trust the Foundation's process?
Record
Layer 6 — Public Memory System
ARIA Trust Ledger
Preserves the permanent, publicly verifiable history of assessments, scores, corroborations, disputes, corrections, provider responses, reviewer actions, and methodology versions. Cryptographically committed at the time of each event. The AI Assessment Index is the storefront displaying current scores; the Trust Ledger is the archive recording everything that has ever occurred.
Answers: What happened, when, under which standard, reviewed by whom, and with what confidence?
Corroboration
Layer 5 — Human Review of Specific Outputs
Corroboration Standard (EM-CS)
Allows credentialed domain experts to assess whether specific AI-generated answers are accurate, complete, well-cited, jurisdictionally appropriate, balanced, and safe to rely upon — without creating professional-client relationships or providing professional advice. Evaluates answers, not systems.
Answers: Have qualified reviewers corroborated this answer, disputed it, or identified limitations?
Measurement
Layer 4 — System Evaluation Framework
Intelligence Assessment Framework (EM-IAF)
Evaluates AI systems across accuracy, hallucination resistance, citation integrity, consistency, fairness and viewpoint balance, uncertainty disclosure, manipulation resistance, human dignity and user agency, civic responsibility, and wisdom and tradeoff reasoning. Evaluates systems, not individual answers.
Answers: How does this AI system perform under structured evaluation against published criteria?
Knowledge
Layer 3 — Human-AI Knowledge Environment
ARIA Network
A governed platform where AI-generated answers, human input, expert corroboration, disputes, and confidence histories coexist transparently. Structured environments where questions are answered by registered AI agents, reviewed by qualified humans, verified by credentialed experts, and published with explicit epistemic status labels.
Answers: How should humans and AI exchange, challenge, and refine knowledge?
Physical
Layer 2 — Physical-World Deployment Environment
ARIA Home
Permission-based, human-controlled environments where AI assists with household safety, accessibility, continuity, automation, and resilience without uncontrolled surveillance or autonomous authority. The physical-world expression of bounded AI within human spaces.
Answers: How should AI operate inside human spaces?
Philosophy
Layer 1 — Governing Architecture and Philosophy
ARIA — Adaptive Responsible Intelligence Architecture
The parent concept and governing philosophy for all Foundation systems. ARIA is not a product — it is the operating architecture and ethical framework within which every other Foundation system exists. It establishes that advanced intelligence must be bounded, accountable, reviewable, transparent, and human-aligned to be genuinely useful.
Answers: What principles govern how intelligence operates within human systems?

II. Canonical Terminology

Standardized names to be used consistently across all Foundation publications

Inconsistent terminology across Foundation documents creates confusion for readers, weakens institutional identity, and makes cross-document references unreliable. The following table establishes the canonical name for every Foundation system and the naming to avoid.

Canonical NameAbbreviationWhat It IsAvoid Using
ARIAARIAThe umbrella architecture and governing philosophy for all Foundation systems"ARIA Framework" as a standalone publication title (this creates confusion with the architecture document)
ARIA HomePhysical-world AI deployment environment for residential and human spaces"Smart home AI," "ARIA residential"
ARIA NetworkHuman-AI knowledge verification and communication platformUsing "ARIA Network" to describe AI identity development (see Contradiction 1 below)
ARIA Trust LedgerATLPublic memory system recording all assessments, disputes, corrections, and methodology history"Trust Index," "Assessment Archive," "transparency log" (the cryptographic log is a component; the Ledger is the public-facing system)
Intelligence Assessment FrameworkEM-IAFFramework for evaluating AI systems across eleven governance dimensions"IAF" alone (use EM-IAF to distinguish from any future external frameworks also abbreviated IAF)
Corroboration StandardEM-CSExpert review protocol for specific AI-generated answers"Verification Standard," "Expert Review Standard" (corroboration has specific legal and conceptual meaning; do not substitute)
AI Assessment IndexPublic-facing display of current IAF assessment scores"AI Trust Index," "Assessment Ratings" — these imply either moral judgment or consumer ratings rather than methodology-based assessment
Assessment CharterGovernance constitution for all assessment activities"Governance Policy," "Ethics Code" — the Charter is a binding governance instrument, not a policy document

The distinction between corroboration and verification is deliberate and legally significant. Verification implies confirming objective truth. Corroboration means qualified reviewers assessed the answer against professional knowledge standards — a meaningfully different and legally safer claim. The terminology difference is not cosmetic.

III. System Relationship Map

How every Foundation system relates to every other
EM Foundation Trust Stack — System Relationship Map EM Foundation Trust Stack — System Relationship Map EM FOUNDATION ARIA — Governing Architecture ARIA HOME Physical-world deployment ARIA NETWORK Human-AI knowledge exchange AI ASSESSMENT INDEX Public score display MEASUREMENT AND REVIEW LAYER EM-IAF — Intelligence Assessment Framework Evaluates AI systems · 10 composite dimensions EM-CS — Corroboration Standard Reviews specific answers · Expert human review ARIA TRUST LEDGER Public memory · Cryptographic record · Full history ASSESSMENT CHARTER Governance constitution · Independence · Appeals · Transparency Solid lines: operational dependencies · Dashed lines: governance oversight (Charter governs all layers)

The key structural relationships: The Assessment Charter governs every other layer — it is the governance instrument that protects the integrity of the entire stack. The ARIA Trust Ledger receives records from both EM-IAF assessments and EM-CS corroboration reviews, creating a unified public record. The AI Assessment Index is a display layer over the Trust Ledger, showing current scores; it does not maintain its own records.

IV. Critical Distinctions

What each layer evaluates — and the distinctions that must not be collapsed
PairLayer ALayer BKey Distinction — Must Not Be Collapsed
EM-IAF vs. EM-CSEvaluates AI systems across dimensionsReviews specific AI-generated answersA high IAF score does not mean every answer from that system is corroborated. A corroborated answer does not mean the system that generated it scores well on IAF dimensions. Never use one score to imply the other.
AI Assessment Index vs. Trust LedgerStorefront — displays current scoresArchive — stores full history, all versions, all disputesThe Index shows what is currently published. The Ledger shows everything that has ever occurred. Score corrections, withdrawals, and superseded versions appear in the Ledger but not necessarily in the Index. Do not merge them architecturally.
ARIA Network vs. Trust LedgerWhere interactions occur — questions asked, answers given, disputes filedWhere records are preserved — what was asked, what was answered, what was assessed, who reviewed itARIA Network is the live operating environment. The Trust Ledger is the permanent record of what occurred there. Interactions happen on the Network; their history lives in the Ledger.
Methodology vs. GovernanceHow scores are calculated (IAF methodology documents)Why the process can be trusted (Assessment Charter)The methodology explains the measurement process. The Charter explains why that process is protected from corruption. Conflating them implies that following a good methodology is sufficient protection against capture — it is not.

V. Documented Contradictions and Resolutions

Inconsistencies across current publications — with canonical resolutions
Contradiction 1 — ARIA Network Dual Identity
Several Foundation documents describe ARIA Network as both an AI identity development network and a human-AI knowledge verification platform. These are meaningfully different concepts and using the same name for both creates serious confusion about what ARIA Network actually is.
Resolution — Use this language consistently
ARIA Network is the Foundation's proposed human-AI knowledge and communication infrastructure. Its initial implementation is focused on knowledge verification, corroboration, and trust records. Future research extensions may study AI identity continuity and agent registration, but these represent long-term research directions, not current platform scope.
Contradiction 2 — Certification Language
Some Foundation materials use "certified" or imply certification status. The Assessment Charter and Legal Risk Report both prohibit certification language for legal and governance reasons. Inconsistency across documents creates legal exposure and misleads readers about what Foundation assessments actually claim.
Resolution — Use this language consistently
Assessed under EM-IAF v[x.x] · Evaluated under published methodology · Corroborated under EM-CS Never: Certified · Verified Safe · Approved · Endorsed
Contradiction 3 — Corroborated vs. Verified True
Some documents imply that a corroborated answer has been confirmed as objectively true. Corroboration means qualified reviewers assessed the answer against professional knowledge standards — a meaningfully different and legally necessary distinction.
Resolution — Use this language consistently
Corroborated by qualified reviewers under EM-CS at the stated confidence level · Reviewer assessment of information quality · Not: verified true, confirmed accurate, proven correct

VI. Recommended Site Navigation

A navigation structure that reflects the Trust Stack logic

The current site navigation is a flat list that does not reflect the hierarchical relationship between Foundation systems. The recommended structure below organizes navigation around the Trust Stack, giving visitors a clear path through the logic: Why → Architecture → Measurement → Review → Governance → Deployment.

This structure reflects the Trust Stack logic: ARIA contains the architecture and deployment systems; Assessments contains the measurement and scoring infrastructure; Standards contains the governance and legal frameworks. A visitor asking "why should I trust this Foundation?" navigates: About → Start Here → Standards. A visitor asking "how was this AI evaluated?" navigates: Assessments → AI Assessment Index → EM-IAF Methodology.

VII. Publication Hierarchy

Recommended reading order from governing philosophy to operational deployment

The Foundation's publications should be presented and linked in the following logical sequence. This order reflects the trust stack logic and ensures that readers encounter the governance rationale before the operational details.

#PublicationStack LayerLogic
1ManifestoFoundation purposeWhy the Foundation exists — the motivating question
2EM Foundation Systems Architecture (this document)ArchitectureHow all the pieces fit — the map before the territory
3Intelligence Assessment FrameworkMeasurementHow AI systems are evaluated
4Corroboration StandardReviewHow specific answers are reviewed
5Assessment CharterGovernanceWhy the process can be trusted
6ARIA Trust LedgerRecordWhere the history lives
7ARIA Network ProposalKnowledge layerThe knowledge exchange environment
8ARIA Home ProposalPhysical layerThe physical deployment environment
9Pilot BenchmarkMeasurement toolThe specific assessment instrument
10Adversarial ReviewGovernance analysisHow the system was stress-tested

The EM Foundation builds governance, assessment, corroboration, and deployment frameworks that help advanced intelligence become useful, accountable, and worthy of public trust. This is not AI ethics. This is not AI ratings. This is trust infrastructure for advanced intelligence — the institutional layer that makes advanced AI safe to integrate into human systems.