Session 4 · Design

Designing courses and research programs.

Everything so far was to get here: a real course, a real assignment, a real research practice — and a defensible decision about where AI belongs in each.

Working session Interactive: Design tool
The method, made operable

Run an assignment through the questions.

This is Session 2's “what are we trying to do?” turned into a tool. Answer for one concrete assignment you teach. It will weigh the answers and recommend a modality — analog, hybrid, or fully-AI — with the design moves and the evidence behind it. It is a structured prompt for your judgment, not a verdict from on high.

Now do it for real — courses

The group-work prompts.

Three versions of one course

Take a course you actually teach. Sketch it three ways — fully analog, hybrid (AI as co-worker), fully-AI. For each version: what does it preserve, and what does it forfeit?

Concrete changes

Pick the modality the tool (and your gut) point to. Name two or three specific changes to assignments or assessment you would make on Monday — not aspirations, edits.

Back to fundamentals

What human skills is this course really trying to impart? Has AI changed which of those still need protecting — or which are even worth teaching now?

The honest trade

Every modality loses something. Say the loss out loud and decide whether it's worth the gain. A decision you can defend beats a rule you can enforce.

Now do it for real — research

Your own use of AI.

The ethics of AI in higher education is not just student plagiarism. It includes how we use AI — in the classroom, in research, and in how we communicate with each other. The same questions apply to your research program: which uses is it important to try, which to adopt, and which are ambiguous either ethically or in real utility?

Reasonable

Worth adopting

Ideation and counter-argument; literature-search prompting with fact-checking; language polishing; turning analysis into visualizations. Disclose where it shaped the work.

Handle with care

Ambiguous

AI as analyst or modeller — beware the illusion of explanatory depth, the illusion of exploratory breadth, and ML “leakage.” Train the method, not just the prompt.

Don't

Off-limits

Letting AI autonomously write substantial or integral parts of the work; uploading confidential or third-party data; using AI to evaluate others' performance.

The research parallel The classroom lesson and the research lesson are the same lesson: decide where the cognitive work must remain yours, protect that on purpose, and be transparent about the rest. See how institutions converged on exactly this →
Where this leaves us

What to carry out of the workshop.

Not a policy. A practice: understand the machine honestly (Session 1), decide what you believe it is and what you're protecting (Session 2), reason from the real evidence and its limits (Session 3), and redesign on purpose — course by course, assignment by assignment (Session 4). Then revise, because the system will change and so should you.

The question was never “how much AI?” It is: what learning are we trying to protect — and have we designed for it on purpose?