CADET: Pilot Candidate Assessment

Online Stage 1 assessment for pilot candidates, built as two single-file web pages on Firebase with no server and no build step. The candidate page opens with an access code and runs eleven modules by default: seven canvas-based cognitive tasks (dual task, tank resource management, 2-back, command chain, spatial, numerical and multiple object tracking), a Situational Judgment Test, a balanced CBTA forced-choice scale (pick most and least), a hazardous attitudes Likert scale and a crew resource management (CRM) scale. Proctoring relies on behavior logs (fullscreen exits, tab and focus changes, blocked copy and paste) and never on a camera or microphone. The item pool the candidate sees and the answer keys only the admin can read are kept apart; an access code is locked in a transaction that checks expiry and session ownership, and the response and the code's used flag are written in one atomic batch. A psychometric audit raised 105 findings across 11 dimensions, each logged as a decision; fairness checks (the four-fifths rule, Cohen's d) run in the browser, and color bands stay hidden until 50 candidates have taken part. GitHub Actions publishes changes that pass its checks. Live, and awaiting an expert panel of at least five before the first real candidate.