I am proposing to develop HIST 201 as a fully asynchronous online course built on a single pedagogical mechanic — students predict, then compute, then reconcile — with the AI tutor layer gated to protect that sequence rather than dissolve it. Roughly 60% of the required reading already exists in publishable form. The remaining build is scripted, scoped and costed below.
A history-of-science survey that only narrates leaves the student with a gallery of admirable people and no transferable skill. They can recount that Kepler was wrong about the Platonic solids; they cannot tell you how anyone established that, and they have not performed the operation that would let them establish anything themselves.
The second problem is newer and more urgent. In an asynchronous online course, any assessment that can be answered by a language model in fifteen seconds is now, functionally, ungraded. Banning the tools does not work and cannot be enforced. Ignoring them produces a course whose grades measure nothing.
Every module runs the same three-beat sequence, and the sequence is what does the teaching:
The anchor example is Week 1. Students compute T²/a³ across six planets and find it holds to within 1.4% over a fortyfold range of orbital radius. They then compute Kepler's own Platonic-solid prediction for the same six spacings and find errors of 6% to 21%. Same man, same data, same decade — one model survived checking and one did not, and the student is the one who found out.
Predict-then-check is fragile in a physical classroom: one student says the answer aloud and it collapses for everyone. Brightspace enforces the gate structurally and privately. This is a case where the asynchronous format is not a compromise against an in-person ideal — it is the better delivery vehicle for this particular mechanic.
Students are issued an AI tutor in Week 0 — deliberately the first thing that happens, since a tool introduced in Week 6 is a tool nobody adopts. Fifteen historical personas, one per module, delivered through a browser-based deck that builds a full prompt and hands it to Claude in the Chrome sidebar. No purchase, no install beyond a free account.
The tutor is gated to match the assessment. In before mode the generated prompt contains no check value anywhere in its text and instructs the tutor to refuse the result however it is asked. It cannot leak what it was never given. In after mode the value is included and the tutor discusses it freely. The student selects the mode, and nothing prevents switching early — which is stated plainly to them, because the grading already removes the incentive. A prediction graded on punctuality rather than correctness gives a student nothing to gain by having a machine guess for them.
The pattern is: do not ban the tool, gate it, and make the gate structural rather than honour-based. It requires no proctoring, no detection software, and no assumption of good faith. It works in any course with a computable or predictable result, which at SPS includes MATH 102, MATH 215, QUAN 201, the Life and Physical Sciences core, and PHE 200.
HIST 201 is a good first instance precisely because it is not a mathematics course. If the mechanic holds for adult non-specialists in a Flexible Core humanities-adjacent elective, it will hold in the quantitative courses.
Students are also told, in Week 1, that the tutor is fluent and will sometimes be confidently wrong; that the assigned chapter outranks it; and that catching it in an error is worth a discussion post. A course about verifying claims should not hand students an oracle and ask them to trust it. The tutor's unreliability is the most realistic training instrument in the course.
HIST 201 sits in the Flexible Core, Scientific World area. Flexible Core courses must meet three general outcomes; Scientific World requires at least three of five area outcomes. This design meets all five, with the mapping documented module by module in the syllabus. The outcome most often thin in a history-of-science course — demonstrate how problems can be analyzed and solved using tools of science, math, technology, or formal analysis — is here carried by the weekly computation rather than asserted.
The content base is an existing corpus of 637 documents. The relevant portion is Book VII, a gallery of 39 named-scientist chapters, of which 29 are written and substantial (12–72 KB each) and 10 are placeholder stubs. The 16-module map is built entirely on the 29, so the course is teachable from existing material and the stubs are not on the critical path.
| Component | Status | Note |
|---|---|---|
| Assigned reading — 13 chapters | exists | Written, at graduate register, used as-is |
| 16-module map, outcome mapping, assessment scheme | exists | course-hist201.html |
| Tutor deck — 15 gated personas | exists | tutor-deck.html, working, gate verified |
| 7-week accelerated cut list | exists | For the compressed term |
| 3 new textbook-tier chapters | build | Week 0 foundations, Week 6 Cantor, Week 15 closing case study |
| 16 computation worksheets | build | Google Sheets template + printable PDF each, check value published |
| 16 lecture scripts and recordings | build | 2–3 clips of 4–6 min per module, captioned. The largest line item. |
| 16 gated forum prompts + instructor guide | build | Documents the one Brightspace setting the design depends on |
| Rubrics — prediction, computation, two audits, final | build | |
| Brightspace shell assembly + accessibility pass | build | Captions, alt text, phone-first layout check |
The video is the dominant cost: 6–10 hours per module once scripting, recording, re-recording, editing and caption cleanup are counted — 100 to 160 hours across sixteen. Two decisions cut it materially. Recording narrated slides rather than talking head roughly halves it and produces something better for students listening on a commute. Splitting each module into 2–3 short clips means a four-minute segment that goes wrong is cheap to redo, where a twelve-minute one is not. Worksheets, rubrics, forum prompts and shell assembly add roughly 60–80 hours. The three new chapters are the only genuinely open-ended item.
All of it is capital rather than labour — recorded once, reused each term, with only the discussion facilitation recurring.
The last item in that list is a genuine dependency, not a formality. If that Brightspace setting is unavailable, the gate becomes honour-based and the design needs rethinking before any video is recorded. I would want it confirmed before Week 1 is built.