Problem Framing, Ethics & Judgment
Mission. Turn business questions into data questions, and know when not to use ML.
Scope
- Problem framing and metric design
- Tradeoff analysis and the cost of being wrong
- Bias and fairness
- Privacy and responsible AI
- Build vs. buy, and the boring solution
Learning objectives
- Turn a vague stakeholder request into a testable, scoped data question
- Name the failure modes of a proposed model before it ships
- Argue for the boring solution when the boring solution wins
Lessons
This track is scaffolded and waiting for content. See the curriculum dashboard for the roadmap and progress across all tracks.
Planned lessons (12)
Wave 3 — depth (provisional roadmap) — 12 lessons
framing
- From Business Question to Data Question Turning vague stakeholder requests into testable, scoped questions.
- Metrics That Matter Designing measures that capture the thing you actually care about.
- Goodhart's Law When a measure becomes a target, it stops measuring.
- The Cost of Being Wrong Asymmetric errors, and decisions that follow from them.
- Baselines Before Models The dumb solution that makes the smart one earn its place.
- Build vs Buy The total-cost math of building, buying, and borrowing.
- When Not to Use ML Rules, heuristics, and the cases ML only decorates.
ethics
- Bias: Where It Comes From Data, labels, measurement, and deployment as sources of bias.
- Fairness: Measurement and Tradeoffs Fairness definitions that conflict, and choosing anyway.
- Privacy and Data Minimization PII, consent, and collecting less on purpose.
practice
- Working with Stakeholders Expectations, updates, and pushing back without friction.
- The One-Page Proposal Question, approach, cost, risk: the document that starts the work.