Skip to content
@node-and-norm

Node & Norm

Independent AI Assurance & Evaluation Lab

NODE & NORM

AI assurance & evaluation

What evidence establishes that governance had practical force?

Node & Norm develops research methods, evidence records, and reference implementations for examining human authority and institutional control in AI-mediated decisions. The work asks what reviewers can reconstruct, which conclusions the evidence supports, and where uncertainty remains.

Human Control & Decision Authority

Project Question or function Start here
Trust, Autonomy & Evidence (TAE) What supports justified reliance and practical human control? Research repository · Releases
Human Influence Telemetry (HIT) What do records establish about the exercise of human authority? Method and implementation · Releases

Decision Evidence & Governed Action

Project Role Repository
Governed Decision Intelligence (GDI) General decision-record specification and reference implementation governed-decision-intelligence
AI Assistance Governance Method/profile for bounded AI participation, assigned human authority, and dependency assessment rgds-ai-governance
Regulated Gate Decision Support (RGDS) Regulated phase-gate reference implementation rgds
RGDS independent study Historical and exploratory research supporting the decision-governance program Study repository · Read the study

The hierarchy is GDI's general decision architecture, the AI Assistance Governance profile, and RGDS's regulated reference implementation. The independent study retains its own research record.

AI Evaluation & Assurance

Catholic Doctrinal Fidelity Index (CDFI) provides evaluation-governance methodology and a reference implementation for domain-specific doctrinal assessment. Scoring-contract reconciliation remains open. SAICRED is an external collaborative project with its own leadership and publication authority.

Emerging research and shared infrastructure

Work Question and status
Institutional Responsibility & Repair What does an institution owe after AI-mediated harm? Poenitentia Institutionum develops the normative argument; empirical outcome validation remains open. Its research repository is private.
Incident Reconstruction & Assurance What can the surviving record establish about an incident and the controls available? An emerging question across the three programs.
Control Evidence Corpus (CEC) Can independent reviewers reconstruct control operation from surviving evidence? Private development infrastructure; independent empirical validation remains pending.

These emerging areas preserve the website's three permanent research programs. The website Registry owns presentation identities and admitted summaries. Each research repository retains its own source and release authority.

Research workflow

Applied AI Research Translator turns research inputs into traceable claims, bounded tasks, and decision records subject to human review.

How the repositories relate

Each project maintains its own scope, versioning, validation status, and releases. The developing Control Evidence Corpus is intended to supply canonical empirical records for downstream work in TAE, HIT, RGDS, and future studies. That integration remains prospective and must preserve source dependence, missingness, and exact-version provenance.

Shared organizational ownership supplies no additional validity to a method or result. A claim about one project's validation must be assessed against that project's reported evidence and limitations.

Read, reproduce, and contribute

Start with a project's README and research-status statement, then inspect its versioned release, source provenance, and validation instructions. Use that repository's contribution and correction process. Cite its authors and exact artifact version; organizational stewardship preserves those credits.

Research led by Mark Julius Banasihan. Public repositories retain their individual licenses and attribution requirements.

Pinned Loading

  1. governed-decision-intelligence governed-decision-intelligence Public

    Defines the decision layer for AI systems where deployment outcomes are governed, recorded, and reconstructable. 5th conforming implementation of draft-farley-acta-signed-receipts (IETF).

    Python 3

  2. applied-ai-research-translator applied-ai-research-translator Public

    Governed research-to-decision translator for applied AI: safety-policy intake, claim extraction, bounded tasks, human gate, and audit-ready decision records.

    Python 1

  3. rgds rgds Public

    A reference implementation for human-governed, defensible phase-gate decisions in regulated environments (non-agentic, schema-validated).

    Python 1

  4. rgds-ai-governance rgds-ai-governance Public

    Working AI Assistance Governance method/profile for bounded AI participation, assigned human authority, and structural/substantive dependency assessment.

    Python 1

  5. human-influence-telemetry human-influence-telemetry Public

    Open documentary assurance method for evaluating whether human authority retained practical force in AI-mediated institutional decisions.

    Python 1

  6. trust-autonomy-evidence trust-autonomy-evidence Public

    Evidence architecture for justified reliance and practical human control in AI-mediated decisions.

    Python 1

Repositories

Showing 9 of 9 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…