Organizational Psychology Survey Designer

What it does

Converts a measurement goal into a psychometrically defensible survey plus its scoring and analysis plan. The output is not a shallow question list but a complete instrument that builds the construct → dimension → item → scale → scoring → reliability gate → analysis plan → action threshold chain.

It first reduces the goal to an operational definition (e.g. engagement → vigor/dedication/absorption). Then, instead of writing items from scratch, it selects a validated base scale: UWES-9 (Schaufeli & Bakker) for engagement, MBI (Maslach & Jackson; personal accomplishment reverse-scored) for burnout, the Edmondson (1999) 7-item Team Psychological Safety scale for safety. It collects ≥3 items per construct (single-item measurement forbidden), fixes the response scale to the construct family (Likert vs 0–6 frequency), and runs every item through a 6-criterion quality rubric.

When to use it

For requests like "design an engagement survey", "let's measure burnout", "psychological safety pulse survey", "climate survey", "inclusion measurement for neurodivergent employees", or "psychometrically audit my existing survey". Whenever a defensible measurement system is wanted, not a single item. Built for HR analysts, OD consultants, and people-analytics teams.

Method / frameworks

• Classical Test Theory (CTT) + reliability: Cronbach α / McDonald ω; internal-consistency threshold ≥0.70 (≥0.80 for critical decisions). UWES-9 literature benchmark: vigor α≈.84, dedication ≈.89, absorption ≈.79, total ≈.90–.93.
• Validation chain: content (Lawshe CVR + expert panel) → construct (EFA/CFA) → criterion (absenteeism/turnover correlation).
• Response-style control: reverse-coded items + anonymity statement against acquiescence, central tendency, social desirability, common-method variance.
• Universal Design + reasonable accommodation: literal, jargon-free, no double-negatives; "Don't know/N-A" escape; screen-reader compatibility — a validity (noise) issue, not mere courtesy.
• Sampling & power: N ≥ 10×items (min ~150); below that no statistical claim is made, only descriptive reading.
• Cross-cultural adaptation: Beaton forward-back translation / ISPOR; "TR validation present/absent" always logged in gaps.
• Standards & ethics: AERA/APA/NCME testing standards; KVKK + default anonymization.