Case B — the original: STAT 201 problem set
Context (fictional): Lakeside State University — a fictional US public university preparing for ADA Title II digital accessibility compliance. STAT 201 Introduction to Statistics is a first-year service course with ~250 students across four tutorial sections. The brief below is realistic but fictional; no real course or student work is used.
The original assignment brief
STAT 201 — Problem Set 4 (10% of course grade)
Task. Answer all ten questions below. Show your working. Units and interpretations matter: a bare number is not an answer.
- A dataset of 40 quiz scores has mean 72.4 and standard deviation 9.1. Compute the standard error of the mean and a 95% confidence interval, and state in one sentence what the interval means.
- Explain, with a sketch, what happens to the width of a 95% CI when the sample size quadruples.
- For the scenario in Q1, test H₀: μ = 75 at α = .05. State the test statistic, p-value logic and your conclusion in context.
- Define a Type I and Type II error in this scenario. 5–10. (Six further calculation/interpretation items on two-sample t-tests, chi-square, correlation vs causation, and reading a regression output.)
Submission. One PDF via the LMS, 11:59 pm Sunday of week 6.
Academic integrity. You may use your notes and the textbook. Working together is fine; identical submissions are not. Use of ChatGPT or similar tools to solve problems is not permitted.
Marking. Each question is worth 1 point for a correct method + 1 point for a correct answer/interpretation (some 2-point, some 3-point; 20 points total).
The original rubric (20 points)
| Points | Standard |
|---|---|
| Method (per item) | Correct statistical procedure chosen and set up |
| Execution (per item) | Correct arithmetic, correct units, correct context sentence |
| Interpretation items | One-mark items: correct statistical concept in context |
What the teaching team actually does with it
- Four TAs, ~65 scripts each, ~10–12 minutes per set with an answer key; partial credit is constant calibration work.
- The “no AI” line is unenforceable and everyone knows it — take-home answers arrive looking suspiciously uniform, with occasional answers in a different register than the student’s in-class work.
- The course team’s real worry: students pass the problem sets by copying (human or AI) and then fail the midterm.
Why this case matters
Calculation-and-interpretation problem sets are the second most common assignment shape in quantitative service courses. The interesting design question is different from Case A: here AI is arguably good for learning as a practice tutor, and terrible as the thing being certified. The redesign separates those two roles.
- 1. Original assignment ← you are here
- 2. Stress test
- 3. Redesign
- 4. Student brief
- 5. New rubric
- 6. Faculty notes