Method and claim boundary

Evidence begins with the job.

TENX ARENA is designed as a structured work sample, not an interview chatbot or a universal score. The current instrument connects a small versioned occupational subset and a fictional employer brief to scenario opportunities, accepted typed events, artifacts, rubric indicators, and bounded claims. Each event retains its server-measurement, client-report, or deterministic-fixture basis. That chain is a design hypothesis awaiting independent job analysis and subject matter expert evidence.

TENX doctrine source statusThe governing TENX_Human_AI_Systems_Master_FA.html doctrine is bundled and versioned with this instrument. Its formulas for synergy, regret, causal effects, optimization, marginal candidate value, and configuration fit remain inactive until the required observations and validation evidence exist.

Traceability chain

From real work to a reviewable claim

01Critical workRole evidence and employer confirmation

02OpportunityBranch, object availability, workspace access-path report, delivery, object render, response, timing, and accommodation kept separate

03Observable actionDecision, check, delegation, or artifact

04Rubric indicatorScenario specific behavioral anchor

05Evidence claimScoped statement with limits and unknowns

Mixed systems stay explicit

Actual humans, candidates, configured collaborator roles, simulated human roles, and software tools are separate actor kinds. The present package uses deterministic scripted outputs, not a live model. Communication, access, authority, approval, and delegation remain separate relations.

Challenge states stay separate

The taxonomy separates a legitimate disturbance from fault injection, exposure, corrupted state, failure, discrepancy classification, containment, correction, verification, residual harm, recurrence, and adaptation. Northstar establishes only injection, exposure, a frozen oracle match for an exact accepted source-comparison request, containment, a correction-action record, and a narrow recovery-oracle check. The comparison result does not establish participant recognition or comprehension; the other stages remain inactive or unknown.

Reports preserve uncertainty

Observed, measured, derived, interpreted, and unknown are never collapsed. One episode does not become a personality claim or proof of future performance.

People make the decision

Measurement, recommendation, and the employment decision remain separate. Hard constraints do not disappear inside a weighted average.

Source registry

Every source has a permitted use.

Versions and limits are first class data. Sources are not blended into an unsupported claim of scientific certainty.

O*NET DatabaseUSDOL/ETA

31.0 / Aug 2026

SupportsThe bundled subset contains three occupation records and eight exact task statements

Does not supportOther role fields are design hypotheses. The subset does not prove employer relevance or scenario validity

Ingested subset

ESCO ClassificationEuropean Commission

1.2.1 / Dec 2025

SupportsMultilingual occupation and skill vocabulary with stable identifiers

Does not supportDoes not provide task importance or employer specific work evidence

Registry only

SIOP PrinciplesSociety for Industrial and Organizational Psychology

5th edition / 2018

SupportsJob linkage, validation, reliability, fairness, documentation, and intended use

Does not supportProfessional guidance, not product validation or legal clearance

Design basis

NIST AI RMFNIST

1.0 / 2023

SupportsGovern, map, measure, and manage controls for AI risk

Does not supportVoluntary guidance, not evidence of hiring validity or compliance

Governance basis

This product includes a transformed subset of the O*NET 31.0 Database by the U.S. Department of Labor, Employment and Training Administration, used under CC BY 4.0. USDOL/ETA has not approved, endorsed, or tested TENX ARENA.

Validation ladder

No generic validated badge.

Evidence attaches to a specific interpretation, use, role, population, version, and decision context.

Source subset registeredThree occupation records and eight task statements are versioned
2
Job linked and SME reviewedEmployer relevance, design hypotheses, and challenge fairness require accountable review
3
Reliability evaluatedAcross tasks, raters, occasions, and versions
4
Criterion evidence obtainedRelationship with relevant later outcomes
5
Local use evaluatedPopulation, setting, fairness, and transportability

Inspect the evidence model

Replay accepted typed events.

An authorized reviewer can follow a session from package and typed event through artifact revision, rating, derivation, contradiction, and bounded claim. Server measurements, client reports, and deterministic fixtures retain different evidence classes. Public ground truth is not exposed.

Open Evidence Studio boundary