Glossary
Definitions for the political-science, public-reasoning, institutional, and AI-governance concepts used throughout the book.
The use of AI for research, analysis, deliberation, administration, monitoring, or decision support inside institutions whose legitimate human authority remains intact.
The ability to inspect important evidence, transformations, disagreements, procedural steps, authorizations, and outputs after the fact.
The capacity for affected people, reviewers, or institutions to challenge claims, evidence, procedures, classifications, or recommendations.
Power over the classifications and definitions that determine which safeguards, procedures, and authority boundaries apply.
An analytical assessment, research result, forecast, normative map, or AI-panel recommendation that contributes knowledge without itself constituting an authorized public decision.
A bounded description of what has actually been established about a claim or mechanism, such as proposed, implemented, tested, externally reviewed, validated, or contested.
A controlled process for determining impact, materiality, urgency, and safeguard triggers rather than leaving those labels to ad hoc model judgment.
An identifiable, legitimate human authority accepting, rejecting, modifying, or overriding a recommendation where consequential action requires authorization.
A staged description of the assurance, enforcement, independence, legitimacy, and operational readiness an implementation can demonstrate for a defined use.
A NousPolis reference-implementation concept for selecting task-appropriate models while considering capability, reliability, cost, privacy, resilience, and epistemic diversity.
A governance condition in which review, assurance, witnessing, or authorization is controlled by a genuinely separate principal rather than the same actor whose work is being evaluated.
Power over framing, routing, panel construction, aggregation, stopping conditions, workflow order, and what gets recorded.
Recorded information showing where evidence, claims, model outputs, transformations, and decisions came from and how they changed through the workflow.
A political and institutional order that deliberately improves collective reasoning with advanced computational tools while keeping legitimate political authority, rights, and accountability human-governed.
The ability of later reviewers to reconstruct what problem was posed, what evidence was admitted, which assumptions and participants shaped the process, what dissent existed, who had authority, and what happened afterward.
A concrete implementation used to operationalize, test, and expose the implications of a more general framework without becoming the definition of that framework.
Power over the actual answer, recommendation, or policy position produced by a governance process.