4. Anatomy of a good data-science lesson¶
Every lesson in this book follows the same five blocks (docs/pedagogy/lesson-anatomy.md). This chapter explains why,
and how to run each block well.
| Block | Time | Purpose |
|---|---|---|
| Hook | 5 min | curiosity and a prediction |
| Concept | ≤ 5–10 min | one idea, explained clearly with an example |
| Activity | 20 min | students do data work |
| Check | 5 min | retrieval: three short questions |
| Reflect | 2 min | "What surprised me? Who is missing from this data?" |
Hook: start with a question and a prediction¶
A good hook is local ("Is Biratnagar hotter than Jumla — by how much?"), concrete (a real number, a photo of a notice board, a news headline) and asks students to predict before they know. Predictions matter: when a student has committed to a guess, they care about the answer, and a wrong guess is a powerful learning moment ("I thought most of us walked 10 minutes — the median is 29.5!").
Techniques: * Show of fingers (1–5) or mini-whiteboards for predictions, so every student commits, not only the confident ones. * Write the class predictions on the board and return to them at the end.
Concept: one idea, briefly¶
The Explanation sections in this book are longer than you need to say aloud: they are written for students to read. In class, pick the one core idea and one worked example. Talk for at most 5–10 minutes. Then let students work.
- Use the board for one clear example (a table of five rows is enough).
- Ask "why?" questions: "Why did the mean go up so much when we added the rich family?"
- Use the Words to know with their Nepali equivalents; let students explain terms in their own languages.
Activity: students do the data work¶
- Groups of three with roles — data steward (keeps the table clean and the log), analyst/driver (computes or types), storyteller (writes the sentence and presents). Rotate roles each lesson so everyone does everything.
- Unplugged first. Shuffle slips before running a permutation test; sort cards on a grid before k-NN; be the "database" holding up cards before writing SQL. Physical activities make abstract ideas concrete and work without power.
- PRIMM for code (chapter 6): Predict → Run → Investigate → Modify → Make.
- Circulate and listen. Ask groups to explain what they are doing; misconceptions show up in explanations, not in answers.
Check: three quick questions¶
Every lesson has three Check questions, with answers in the teacher guide. Use them as retrieval practice, not as a test: * students answer from memory, on paper or mini-whiteboards; * one question recalls a previous lesson (spacing); * go over answers immediately; it takes 5 minutes.
Retrieval practice is one of the most robust findings in learning research: remembering something makes it easier to remember later.
Reflect: what surprised me? who is missing?¶
Two minutes, every lesson, the same two questions. Over a year they build two habits: noticing (surprise means a prediction was wrong — learning happened) and asking about the edges of data. Collect a few answers in a class "surprise wall" or "missing people list".
Questioning in data lessons¶
Good questions to keep asking: * "How do you know?" / "What in the data shows that?" * "Compared with what?" * "Could that happen by chance?" * "What else could explain it?" * "Who collected this, and why?" * "Who is not in this table?" * "What would change your mind?"
Managing discussions about sensitive data¶
Data about caste, ethnicity, religion, disability, income or health can hurt. Guidelines: * Use synthetic data for practice; real sensitive data only in aggregate. * Describe groups; never rank people. * Agree ground rules (Grade 8 Unit 8's class data charter). * Allow students to opt out of a dataset that touches their life (for example, the 2015 earthquake), and always offer an alternative.