This page defines the authoritative vocabulary used across this research program — in articles, the book, talks, and reference architectures. Definitions here supersede earlier informal usage.

Status legend

Status Meaning
Foundational Core principles defining the initiative’s long-term vision. Expected to remain stable for years.
Canonical Official concepts, frameworks, or terminology formally introduced and recommended for use.
Emerging Publicly introduced concepts that continue to evolve through research and feedback.
Working Active research ideas under refinement — open for discussion, not yet stable.
Experimental Early hypotheses or prototypes shared to stimulate discussion.
Deprecated Historical terminology retained for context but no longer recommended.

Foundational concepts

Engineering Trustworthy Industrial Decisions — Foundational An engineering discipline focused on designing Industrial AI systems that earn trust through engineering knowledge, grounded reasoning, decision governance, human accountability, and operational rigor.

Engineering Knowledge — Foundational The body of engineering expertise — standards, procedures, physical principles, operational practices, asset knowledge, failure mechanisms, and institutional experience — that provides the foundation for trustworthy Industrial AI.

Grounded Reasoning — Foundational Reasoning constrained by engineering knowledge, standards, physical principles, and operational context, rather than statistical correlation alone. Grounded reasoning precedes trustworthy operational decisions.

Canonical frameworks

The Governed Decision Layer — Canonical · first introduced in Volume I The architectural layer between reasoning and operational action, responsible for validating, governing, explaining, authorizing, and recording industrial decisions.

ASSET-grade — Canonical A benchmark describing Industrial AI outputs that are Auditable, Specific, Standards-aligned, Explainable, and Tested — tested meaning the output’s grounds can be interrogated and the decision rehearsed before anyone commits. ASSET-grade is what makes an output deserve trust; trusted is the outcome it earns, not a letter in the benchmark. The acronym is fixed — additional letters should not be appended. (Revised 2026-07-10: T previously read “Trusted”; changed to remove the circularity of defining trust by trust.)

Trustworthy Decision — Canonical A decision that is sufficiently grounded, explainable, governed, auditable, and operationally appropriate for industrial use. Trustworthiness is an emergent engineering property, not a model characteristic.

“Who Signs?” — Canonical A recurring design test asking whether responsibility for an AI-assisted operational decision has been clearly assigned. If nobody signs, the decision is not operationally ready.

Emerging concepts

Earned Autonomy — Emerging The principle that autonomous capability should be granted progressively through demonstrated evidence of trustworthy performance. Autonomy is earned — not configured.

Working concepts

Deterministic Decision Envelope — Working A deterministic operational boundary surrounding non-deterministic AI reasoning, ensuring predictable and accountable operational commitments. This term refers to operational decisions — not deterministic AI.

Reserved terms

The following are reserved for future research; definitions will be added when formally introduced.

Engineering Trust · Industrial Decision Intelligence · Decision Provenance · Decision Ledger · Engineering Memory · Trust Calibration · Industrial Cognitive Architecture · Operational Confidence · Engineering Assurance

Changes to canonical terminology follow the process described in GOVERNANCE.md. Please cite the originating article or framework when referencing these terms.