qualtrics-codex

Demo Artifacts

These downloads show the discrimination-beliefs teaching demo using synthetic fixture data only. They do not contain real Qualtrics responses or private survey metadata.

discrimination-beliefs-demo-slides.pdfDownload

discrimination-beliefs-demo-slides.htmlDownload

discrimination-beliefs-demo-figures.zipDownload

discrimination-beliefs-demo-tables.zipDownload

Codex Loop

Start by asking Codex to inspect local readiness and ask the needed preference questions before it builds anything. Then run the main live teaching demo.

Open prompts/configure-local-preferences.md and follow it as instructions for this Codex session. Do not summarize it. First inspect Python, dependencies, Stata, LaTeX/Beamer, and whether QUALTRICS_DATACENTER and QUALTRICS_API_TOKEN are set without printing values. If Stata, LaTeX, or Qualtrics secrets are missing, tell me where to configure them or which fallback to use. Then ask the needed follow-up questions and save my answers in ignored AGENTS.override.md without secrets.

Create a public opinion survey on beliefs about discrimination in hiring in Qualtrics. Then generate 100 synthetic responses on Qualtrics to test the survey before it is launched to human subjects, download and clean the generated data, create figures, and compile slides that summarize the workflow, survey design, synthetic response patterns, and main figures. Include the survey link in the slides.

Because this prompt asks for work on Qualtrics, Codex should verify credentials without printing them and ask before creating the draft survey, submitting synthetic responses, or exporting responses.

The reusable preference prompt is prompts/configure-local-preferences.md. Stata workflows use SPSS/SAV exports; Python workflows use CSV exports.

For the full conversational workflow, see docs/intended-codex-loop.md.

For the worked prompt, see prompts/discrimination-beliefs-example.md.

Run Locally

git clone https://github.com/Ingar30/qualtrics-codex.git
cd qualtrics-codex
.\scripts\setup.ps1
codex

After opening Codex, paste the opening prompt above so Codex follows prompts/configure-local-preferences.md instead of summarizing it.

For shorter plain-language variants, see docs/codex-prompt-alternatives.md.

If virtual environment setup fails or Stata is not found, see docs/setup-troubleshooting.md.

Private by default. Raw exports, processed real data, metadata, survey IDs, reusable links, and API tokens should stay local unless you explicitly decide otherwise.

Live Qualtrics Loop

Store credentials outside the repository, start with check-auth, save reusable links only to ignored local files, submit generated synthetic rows to the test survey, export once, analyze once, and build slides. See the walkthrough and the validation notes.

Validated Full Loop

A local live run validated the lean command loop with 100 synthetic Qualtrics submissions, one export, one analysis pass, and slide output. Public Pages artifacts remain synthetic-only; live validation runs locally with user secrets.

python scripts/run_live_validation.py --survey-key <survey_key> --survey-name "<survey name>" --spec-file code/<survey_key>/survey_spec.json --n 100 --dry-run