Case B — the assignment as students see it

Written in plain language. Every stage shows its AI-use level as icon + text. The example disclosure line shows what logging looks like in practice.

STAT 201 — Weeks 1–6: practice sets, two in-class checks, one class dataset

This course has two kinds of statistics work, and they have different AI rules:

Practice sets (weekly, 10% total): where you learn the methods. AI is allowed here as a tutor — that’s the point.
In-class checks (weeks 3 and 6, 10% each): where you show what stuck. No devices, 20 minutes, four questions on the same skills you practised.

The weights tell you what we care about: the secure checks count for more than all five practice sets together. Practice is practice; the checks are the evidence.

The weekly practice sets — AI level 4 · Full AI

Use AI however it helps you learn: ask it to explain, to check your setup, to quiz you — even to solve. The only requirement is the log: next to each question you used AI on, write one line:

Q3: asked AI why we use t not z here; its first answer skipped df; redid it with the book’s method.

You get 2 points per set for submitting it with a real log. We read logs, not answers — a log that says “AI got it wrong and here’s why” scores just as well as one where AI was right. That’s the skill that matters.

The in-class checks — AI level 1 · No AI

Twenty minutes in your tutorial, in weeks 3 and 6. Four questions, same skills as the practice sets: one computation, one interpretation of the class dataset, one “find the error”, one concept question. Pen and paper (or laptop with a blank document if you have an accommodation plan — see below). No notes beyond your one-page formula sheet.

The class dataset — AI level 3 · AI Collaboration

In week 2 everyone answers two quick questions (commute time, sleep hours) in a 3-minute form. From week 3, the anonymised class dataset goes up on the LMS, and interpretation questions run on our real numbers. You may use a spreadsheet or AI to help you compute — but the sentence that interprets the result in context must be yours, and the in-class checks ask exactly that skill.

“Find and fix” items — AI level 5 · AI Exploration

From week 3, each practice set includes one AI-written solution with an error in it (we use real mistakes from last year’s class, anonymised). Find it, fix it, explain the fix in two sentences. Using AI to help you hunt the error is allowed and expected — log it. Finding nothing is fine; explaining why you checked what you checked is what earns the point.

Why these rules (the honest version)

Practice with AI as a tutor is closer to how you’ll actually use statistics after uni. Practice sets count little because copying answers teaches nothing — the log is what turns AI use into learning.
The in-class checks are the only things that certify what you can do without help. That’s why they’re weighted higher, why they’re short, and why there are only two.
We don’t use AI detectors. They misfire — especially against multilingual writers — so nobody gets investigated over a detector score. The check marks say what needs saying.
If the tutorial time doesn’t work for you one week (illness, care duties, sport), tell your TA before — there’s a make-up slot; you don’t need to explain why.

Example log entries

Q1: no AI — did it from the worked example in §8.2. Q3: asked AI to check my CI arithmetic; it matched mine, and I understand why we use 1.96 here. Q7 (find-and-fix): AI helped me test the suspicious df value; the error was in the variance formula, not the arithmetic.

What good “AI Collaboration” looks like on interpretation questions

You may compute with AI. The context sentence is yours:

“The mean commute time in our class data (n = 213) is 27.4 min, SE 1.6, so the 95% CI is [24.3, 30.5] — we’re 95% confident the interval procedure captures our class’s true average commute, not the city’s.”

Two sentences like that in your own voice, tied to our data, beats five pages of generic textbook phrasing — and it’s exactly what the checks ask for.

Checklist before each set

Log line for every question where I used AI (or “no AI” on the ones I didn’t)
Interpretation sentences written in my own words, about our data
Find-and-fix: did I say why I checked what I checked?
For check weeks: formula sheet updated, make-up slot booked if needed