Course Synthesis
The 16-point course synthesis, a 14-point individual submission plus 2 points for the Week 16 presentation sessions.
The question it answers: you have a real decision, a model you built, and a recommendation. Can you hand all three to somebody who has to act on Monday, in ten minutes, and have them actually use it?
The course ends with two linked pieces and no final exam. Together they are worth 16 points:
| What | Points | When | |
|---|---|---|---|
| Presentation | Ten minutes, live, if you’re drawn | 2 | Mon Dec 7 and Wed Dec 9, both sessions |
| Submission | The individual workbook and write-up | 14 | Sun Dec 13 |
Presentations come before the submission deadline, so the room’s questions can still improve your final version. Your analysis is graded once, in the submission; the 2 presentation points are for being in the room and ready on both days, which the next section explains.
The submission (14 points)
This is the term’s largest take-home, worth 14 points (regular take-homes are 5 points; see Grading). It was announced in Week 12 so you could identify and develop a suitable problem well before the end of term. Like every take-home, it is individual.
Pick a real problem, choose a method from the course that fits its structure, build a model somebody else could re-run, and give a recommendation with a caveat and a what-if check attached.
AI use. Tier 2: AI is expected here, and your critique of it is part of what is graded. Log what you asked, what it produced, what was wrong or unsuitable, and what you changed, then confirm you can reproduce, modify, and explain the model unaided. See Using AI.
What to submit, and how it is scored
| Part | What it needs | Points |
|---|---|---|
| Problem in plain language | The decision, its context, and why it matters. | 1 |
| Method and why it fits | The structural signal that points to break-even, decision analysis, linear programming, transportation/assignment, multi-criteria or goal programming, CPM/PERT, forecasting, or regression. | 2 |
| Re-runnable model | Inputs, choices, formulas, constraints, and results, organized so somebody else can open and run it. | 4 |
| Results and recommendation | The key output translated into an action a non-expert can take. | 3 |
| Sensitivity or validity check | One consequential input changed, with what happens; for a forecast, an error or validity check. | 2 |
| Caveats and assumptions | Weak data, simplifying assumptions, and what you left out. | 2 |
| Total | 14 |
The AI-use log is required on top of that: the disclosure, plus what the tool produced, what was wrong or unsuitable, and what you changed. If you didn’t use AI, say so and describe how you checked the model instead. A submission without it comes back for resubmission before it is graded.
Due Sun Dec 13, submitted in Canvas. The 3-3-3 rule covers late work here, but a late pass can never cover the live presentation points.
The presentation (and the draw)
With 34 students and two 75-minute sessions, there is no way to give everyone a presentation long enough to be worth giving. So we do what the course already does with the checkpoints: draw at random. About ten students present, five per session, drawn in the room on the day.
Come to both sessions ready to present. The draw is not announced in advance, so “ready” means ready either day. That is a lighter ask than it sounds: you are building the synthesis anyway, and being drawn adds no grade risk.
The 2 presentation points
Only about ten of you are drawn, so the points cannot depend on being picked. Everyone present earns them, drawn or not: 1 point per session, for being in the room and taking part on Dec 7 and Dec 9. Coming ready is the whole ask, and the draw adds no grade risk.
Like every in-class item these can’t be made up, so missing a session costs that point. The only exceptions are a documented AES accommodation or religious observance arranged in advance; see Policies.
If you’re drawn: ten minutes
Ten minutes is enough to actually show the work rather than gesture at it. Walk us through the five beats, open the model itself where it helps, and expect a question or two at the end.
- Problem. State the real situation, decision, and who is affected.
- Method and rationale. Name the method and the structural signals that made it appropriate.
- Result. Show the one or two outputs the audience needs.
- Recommendation. Translate the model output into a concrete action.
- Caveat and what-if. Name an important limitation and show how one plausible input change affects the recommendation.
Use the live workbook and at most a slide or two. A clear chain of reasoning beats a polished deck, and at ten minutes there is time to show a formula or a Solver constraint and say why it is there.
Before you present
Rehearse your story in Week 15, then open the model you actually built rather than the one you meant to build and ask the one question a rubric can’t ask for you: could another person find the decision, the inputs, the formulas, and the result without you standing next to them? If not, that is the work left to do. Do this whether or not you expect to be drawn.
Bring anything still unresolved to the Wed Dec 9 close, while we are all in the room.
What you’re actually certifying
Every method this term did the same thing to a messy decision: it forced the assumptions into the open where somebody could argue with them. That is the part worth keeping. You will forget which menu Solver’s Answer Report hides under; you will not forget that a confident number with an unexamined probability behind it is worth less than an honest range.