9. Assessment¶
The curriculum's assessment fits Nepal's continuous, competency-based approach: frequent low-stakes checks, projects marked with a rubric, and oral checks.
The assessment toolkit¶
| Tool | When | Purpose | Marked? |
|---|---|---|---|
| Check questions (3 per lesson) | every lesson | retrieval practice; spot misconceptions | no (self-check) |
| Starter quiz | start of lesson | recall the previous lesson | no |
Unit quiz (quiz.yml) |
end of unit | check core ideas | low stakes |
| Practice exercises (★/★★/★★★) | homework | practice and stretch | selected, or peer-marked |
| Real-data exercises | homework / project | transfer to real data | as part of project |
| Unit project | end of unit | apply the whole cycle | rubric |
| Oral check (2 min) | every project | verify understanding; own the work | yes |
| Observation checklist | during activities | "I can" statements | records |
| Portfolio | across the year | growth and reflection | end of year |
The six-criterion project rubric¶
From docs/pedagogy/assessment.md, used for every project:
| Criterion | 1 Beginning | 2 Developing | 3 Secure | 4 Extending |
|---|---|---|---|---|
| Problem | vague question | stated but not answerable with data | clear, answerable, local | also anticipates limits |
| Plan & Data | source unclear | source named; bias not discussed | source, sampling and consent described | also compares alternatives |
| Analysis | wrong or missing method | right method, errors | correct, readable graph/table | also checks outliers/rates/uncertainty |
| Conclusion | no evidence | claim without evidence | claim–evidence–reasoning | also states uncertainty and who is missing |
| Communication | hard to follow | partly clear | clear poster/notebook | engaging, accurate, accessible |
| Ethics & honesty | ignored | mentioned | privacy and fairness handled | also credits all sources and AI help |
Score each criterion 1–4 (total out of 24). Share examples of "Secure" work so students know what it looks like.
The oral check¶
Two minutes per student (or group, with questions to individuals). Example questions: * "Show me where you removed the −999 values. Why did you remove them?" * "Why did you report the median and not the mean?" * "What would change your conclusion?" * "Who is missing from your data?" * "Explain this line of code."
The oral check is essential now that AI tools can produce plausible work: it rewards understanding, not polish. Keep it friendly; a student who cannot explain something gets a chance to revisit it.
Marking code¶
Mark the thinking, not the elegance: does the code do what was planned? Is it readable (names, comments)? Does it
run from top to bottom (Restart & Run All)? Are results checked (assert, totals adding up)?
Grading for growth¶
- Allow resubmission of one project per term after feedback.
- Keep a portfolio: the best project per unit with a short reflection. By Grade 12 it is evidence for university applications and jobs.
Item writing for data science¶
Good quiz questions test interpretation, not only calculation: * "A news site says cases doubled from 2 to 4. What question would you ask?" * "Which summary would you report for monthly household income in a village, and why?" * "A model has 85% accuracy. What else do you need to know?"