What does the instrument record, and who reads the record?
What the official instrument keeps from an Assessment, what is never used to score, and who may read which part of the record, in the words of AIRS-J/1.
Judgment is how well a person decides once AI has given the answer. AIRS-J/1, the standard AIR APAC publishes for judgment in AI-enabled work, reads it from what a person did during an Assessment. Before they start, candidates are told who receives their result and what it may be used for (Section 12.6).
Built by Exeter Labs
The official instrument
StepSim is the official instrument of AIRS-J/1. Section 8.1 describes it as "a scenario simulation with retained decision trace". Its specification, version 2.0.0, is incorporated in the Standard by reference (Annex F). Exeter Labs builds and operates StepSim, AIR APAC's official assessment instrument. The owner of AIR APAC also owns Exeter Labs (Section 4.9).
What the record keeps
The transcript of an Assessment retains "the full decision trace: decisions, artifacts, sources consulted, stated rationale, revisions and timings" (Section 8.7).
Where a scenario shows the consequence of a decision, the candidate's position is recorded before the consequence is shown and cannot be silently overwritten (Section 8.6).
The Standard requires every transcript to state the scenario version, the rubric version, the instrument version and the scaffolding level. A result that cannot state these is void (Section 9.6).
The assessment log records how the candidate's identity was checked at the start of each certifying session (Section 9.1). An Assessment for a hiring decision is sat in person, under supervision. Remote sittings for hiring open only when the approved instrument records the session (screen and camera) for its duration. Such a recording is integrity evidence and is never a scoring input (Section 9.1).
What is never used to score, and what never stands alone
"Keystroke capture, affect inference, video analysis and biometric signals are never scoring inputs" (Section 8.7). Every observation must trace to identified evidence in the transcript (Section 8.8). A model-generated score "is not an observation and never stands alone": a qualified human assessor confirms or overrides it, with reasons recorded (Section 11.4).
Who reads which part
- The candidate. Every candidate receives the full report, including the evidence cited for each observation, whether or not they certify (Section 12.5). A candidate may obtain their own transcript and result, and may ask for factual errors to be corrected (Section 14.5).
- An employer that commissions an assessment of its own people. It receives each result by name: dimension and domain scores, band, confidence and the evidence behind each observation. Its facilitator receives every result, and each line manager receives the results of their own people (Section 12.6, Employees).
- An employer hiring for a role. The people deciding on the role receive each applicant's result by name, for that hiring decision. Nobody else in the organisation does (Section 12.6, Applicants). Where a staffing or placement firm commissions the Assessment, a result passes to a client only if the applicant notice named that client. It then goes only to the people deciding on that placement. AIR APAC provides no rank order of applicants.
- An independent or sponsored candidate. The result goes only to the parties the candidate names. A sponsor learns a result only if the candidate names it. Where ten or more sponsored candidates from one sponsor have sat, the sponsor may receive the number of credentials issued, and nothing else (Section 12.6).
- An appeals reviewer. On an appeal, a reviewer who was not involved in the original decision considers the transcript, the published anchors and any access-support record (Section 18.5).
- AIR APAC and its instrument provider. They use assessment data to operate, check and improve the Standard and the instrument. The data is not sold (Section 14.7). Section 12.9 says the data is held by AIR APAC and Exeter Labs. A record kept only for the evidence programme and improvement carries no name or contact details, unless the person agreed to a named record (Section 14.6).
- The public register. It shows the result in summary, not the record. It holds the credential code, track, role edition where used, band, confidence, version, dates, status and scope. It does not hold the employer, the dimension scores, the stakes profiles or any transcript content (Section 14.1).
An employer, whether it commissions an assessment of its own people or hires, is never told whether an access support was used (Section 12.6).
Before a training session starts, participants are told which roles receive a development reading by name (Section 12.6). A session gives a development reading; it must not be used in a hiring, promotion or appraisal decision (Section 12.8).
What this does not claim
- That a retention period is published yet. Section 14.6 requires each period to be published and stated in the notices; this page states none.
- That an employer sees a candidate's full transcript. Section 12.6 gives the employer the evidence behind each observation.
- That the instrument watches or listens to a person in order to score them.
- That StepSim records every item on this page today.
- That every sitting is recorded on screen and camera. Section 9.1 requires a recording for remote sittings for hiring.
AIR APAC publishes the standard for judgment in AI-enabled work, certifies people against it through its official instrument, and stands behind every result.