Add survey psychometric validity assistant

[drsteinerdj]

Summary

/claim #16

Adds a distinct survey psychometric validity assistant for the AI-Powered Research Assistant Suite. This slice evaluates synthetic survey/manuscript packets before AI peer-review output is released, with gates for:

• reverse-coded item handling drift
• low or missing internal-consistency evidence
• weak factor loadings and high cross-loadings
• confirmatory factor analysis sample-size shortfalls
• thin Likert anchor sets
• construct-claim mismatch between manuscript claims and validated scales

The implementation is self-contained under survey-psychometric-validity-assistant/, uses no external AI APIs, credentials, network calls, private manuscripts, or live datasets, and emits deterministic RELEASE, REVIEW, or HOLD decisions.

Demo artifact

• survey-psychometric-validity-assistant/reports/demo.mp4
• survey-psychometric-validity-assistant/reports/summary.svg
• JSON and Markdown reports under survey-psychometric-validity-assistant/reports/

Validation

npm run check
npm test
npm run demo
npm run demo:video
git diff --check
ffprobe -v error -select_streams v:0 -show_entries stream=codec_name,width,height,r_frame_rate,duration -show_entries format=size -of json survey-psychometric-validity-assistant/reports/demo.mp4

Results:

• syntax check passed
• tests passed: survey psychometric validity assistant tests passed
• demo emitted clean RELEASE and risky HOLD
• demo video generated successfully
• git diff --check passed
• ffprobe: H.264, 960x540, 12 fps, 5 seconds, 29032 bytes

Non-overlap note

This is not another broad assistant suite, model-assumption diagnostic, missing-data sensitivity assistant, reviewer calibration module, preregistration consistency guard, citation recency guard, external-validity transfer assistant, data fabrication anomaly assistant, or assistant-output disclosure guard. It is scoped to psychometric survey scale validity before trusting AI-generated peer-review output.