Data Science Judgment Lab

Data Science Judgment Lab is a static, case-based web lab for practicing data science judgment under uncertainty. Learners inspect messy evidence, form a working theory, make a decision, assign confidence, and then compare their reasoning with an expert replay.

The public version currently includes:

• 25 interactive cases
• a filterable Case Finder
• guided Pathways and a short Intake route
• local progress tracking in the browser
• certainty-based marking for confidence calibration
• a self-attested Judgment Record after enough reviewed replays
• instructor-facing answer maps and facilitation notes

Purpose

The lab is built for the part of data science that is hard to assess with a coding quiz: deciding what imperfect evidence can support, what it cannot support, and how confident a responsible analyst should be when pressure is in the room.

It is not a proctored credential, statistics refresher, or certification exam. It is a practice environment for evidence discipline, causal caution, model deployment judgment, fairness and governance review, and uncertainty communication.

Stack

• Astro static site
• MDX-authored case pages
• React islands for interactive case work
• browser-local progress and scoring
• GitHub Pages deployment

Local Development

cd site
npm install
npm run dev
npm run build

GitHub Pages

The repository includes a GitHub Actions workflow at .github/workflows/deploy.yml. It builds the Astro app from the site/ subdirectory and deploys through GitHub Pages.

The current deployment target is the custom domain https://dsjlab.org. The workflow sets PUBLIC_SITE_URL=https://dsjlab.org, and site/public/CNAME contains the same domain so the Pages artifact carries the custom-domain marker.

Analytics

Google Analytics is optional and controlled by the public build-time environment variable PUBLIC_GA_MEASUREMENT_ID. In GitHub, set the repository variable GA_MEASUREMENT_ID to the GA4 web stream Measurement ID for dsjlab.org. If the variable is absent or not shaped like G-XXXXXXXXXX, the analytics tag is not rendered.

License

Data Science Judgment Lab uses a split open license:

• site source code, scripts, styles, scoring logic, and interactive components are licensed under MIT
• curriculum content, case content, evidence packets, replay text, instructor notes, documentation, learner-facing text, audio scripts, and research notes are licensed under CC BY-NC 4.0

See LICENSE.md, LICENSE-CODE, and LICENSE-CONTENT.md for the full terms.

Documentation

• Product Blueprint
• UX and Interaction Model
• Content Model
• Scoring and Calibration
• Core Case Map
• Build Plan
• Future Case Library Plan
• GitHub Pages Deployment
• Case Authoring Template

Some design docs preserve earlier planning language; the site itself is the source of truth for the current 25-case release.