Blank-Page, Classification & Changed-Fact Drills
Five-minute blank-page reconstruction
B1 — Chapter 2 mental skeleton
Draw five branches: 1. why ethics matters; 2. ethical principles; 3. privacy/online principles; 4. unethical-practice risks; 5. ethical culture + Governance.
Add at least three sub-branches each.
B2 — Ethical decision path
Start with a proposed data use and build a flow asking: legal? → ethical purpose? → people/harm? → fairness? → privacy themes? → source/quality/lineage? → disclosure/re-identification? → Governance/oversight? → controls/evidence?
B3 — Privacy without law-cramming
Rebuild recurring privacy-management ideas and connect each to a handling decision. Keep it source-bound.
B4 — Unethical-practice map
Put the six §3.4 patterns around “misleading or harmful decision.” Under Bias add the five mechanisms. Add one countermeasure clue to every branch.
B5 — Ethical-culture operating model
Draw values/principles → risk → practice → control → training/comms → monitoring/audit → escalation/no retaliation → Governance/legal oversight. Show how a roadmap stages implementation.
B6 — Five-minute teach-back
Teach Chapter 2 to a new analyst, connecting ethics beyond law, Belmont/Menlo, privacy themes, six risks, culture, four checkpoints and Governance.
Where-does-this-belong classification
Choose the most direct Chapter 2 concept.
| ID | Situation |
|---|---|
| C1 | People have no meaningful choice about sensitive secondary use. |
| C2 | Overall benefit is high, but avoidable harm is unnecessary. |
| C3 | Comparable people receive systematically different outcomes by group. |
| C4 | A chart scale makes a tiny change look dramatic. |
| C5 | Two true statistics use different populations but are presented as comparable. |
| C6 | Filters change repeatedly until a preexisting hunch is confirmed. |
| C7 | A merged dataset has unknown origins, conflicting definitions and no change history. |
| C8 | Names are removed, but external data reconstructs identity. |
| C9 | Ethics statement exists with no training, controls, monitoring or safe escalation. |
| C10 | Legal sees no prohibition, but high-impact use has no Governance review. |
Compression teach-back
- 5 minutes: whole chapter including Governance handoff.
- 2 minutes: ethics beyond law; people/misuse/value; Belmont/Menlo; privacy themes; six risks; culture; Governance.
- 30 seconds: one professional rule that remains accurate when compressed.
Changed-fact discrimination
- Model issue changes from group disparity to lack of meaningful choice. Which principle now leads?
- Misleading result changes from distorted axis to selected observation window. Which risk now leads?
- “Anonymous” data becomes aggregated enough that people cannot realistically be traced. What risk is reduced, and what responsibility remains?
- Integrated data gets complete lineage/Metadata/history but source values are known inaccurate. What remains strongest?
- Training/escalation are excellent but no testable controls exist. What link is missing?
- Governance review occurs, but practitioners knowingly hide a fairness concern. Has ethical responsibility been satisfied?