The history of science is not a parade of discoveries. It is a history of checking — of claims made, propagated, tested, and in many cases withdrawn. This course teaches that history by making the student do the checking. Every module carries one short computation the student executes personally. Not a demonstration watched on video: an arithmetic they run themselves, with a published check value they can compare their result against.
A claim is not knowledge until someone checks it. The people in this course are worth studying not because they were right, but because they built the machinery by which being wrong becomes discoverable.
CUNY SPS is fully online, and the delivery norm is asynchronous work inside Brightspace, completed on the student's own schedule within a module window. The student population is working adults, frequently on phones, in different time zones, with no campus, no lab, and no scheduled class hour. Every design decision below follows from that.
The obvious casualty is the live classroom moment — the room computing something together and reacting to it at the same instant. That is gone and cannot be faked with a webcam.
What replaces it is better for this particular course, and the reason is specific. The pedagogy here depends on a predict-then-check sequence: the student must commit to what they expect before they compute, or the discrepancy teaches nothing. In a live room that sequence is fragile — one student says the answer out loud and it collapses for everyone. Brightspace enforces it structurally. A discussion forum set to "must post before viewing others' posts" makes the commitment mandatory and private, and the reveal cannot leak. The gate that a classroom cannot hold, the LMS holds automatically.
Part A — Predict (gated). Before opening the data, the student posts a one-sentence prediction to a forum they cannot read until they have posted.
Part B — Compute (solo). The student runs the computation in a spreadsheet or on paper and compares against the published check value. They know immediately whether they got it right — no waiting on a grader.
Part C — Reconcile (open). The forum unlocks. The student reads what the cohort predicted, and replies to two peers on where the prediction and the result diverged, and why.
Predictability matters more online than in person; a working adult opening Brightspace at 11pm needs to know exactly what is in front of them. Every module is the same six items:
Collaborative in-session computation and paired work are gone — replaced by gated prediction plus peer reconciliation, which is asynchronous by construction. Any activity requiring shared physical apparatus is gone. The cross-staff build survives only because it needs a ruler and a piece of card.
Two modules were rebuilt rather than trimmed: the Faraday document analysis (was a group reading, now an individual annotation task) and the Escher classification (was a live sorting exercise, now a decision-key worksheet with a self-check).
Students work on whatever machine they own — typically an inexpensive laptop and Chrome. Before any content is taught, Week 0 puts them through a five-minute setup that ends with a working AI tutor in their browser sidebar. This is deliberately the first thing that happens, not a resource mentioned in passing on the syllabus, because a tool introduced in week six is a tool nobody adopts.
The mechanism is a tutor deck: a dropdown of fifteen historical figures, one per module. Selecting one builds a full persona prompt and hands it to Claude — copy to the Chrome sidebar, or open in a browser tab for machines where the extension will not install. No account beyond a free one, no software to buy.
An AI tutor that answers the question destroys the predict-then-check mechanic and the grade attached to it. So each tutor ships in two stages, and the student selects which one they are in.
Before I compute — the generated prompt contains no check value anywhere in it, and instructs the tutor to refuse the result outright, however it is asked. It opens by asking the student what they expect. It helps freely with method, units, setup, and vocabulary.
After I compute — the check value is included and the tutor discusses it openly, starting by asking what the student predicted and where the gap opened up.
Nothing stops a student from switching stages early. That is stated plainly on the deck, and it is the point: the course is about what you do when nobody is checking. The grading removes the incentive to cheat rather than trying to police it — predictions are marked on punctuality and specificity, never on being right, so an honest wrong guess already scores full marks.
One further use, and it is not incidental. The assigned Book 7 chapters sit at graduate register on purpose. The tutor is what makes reaching for them feasible for a working adult at 11pm — and because it is fluent, plausible, and occasionally wrong, it doubles as the course's most realistic training instrument. Students are told in week one that the chapter outranks the tutor, and that catching the tutor in an error is worth a discussion post. A course about verifying claims should not hand students an oracle it asks them to trust.
Week 15 has no tutor, by design. The final audit is the student applying the method alone.
The audit framework itself: WP-24 through WP-63. How a claim is checked, what an overclaim looks like, how a correction is recorded.
The Scientist Gallery. Who did the work, who caught the errors, what it cost them, and what they understood about the stakes.
The original two-course sketch treated the seven domains as generic placeholders to be swapped out. They are not placeholders — they are Book 3, sitting in the repository, and they are what gives the history layer something concrete to be a history of. Modules 4, 7, 10 and 14 reach into them for their verification cases: Hamilton into ρ, Pasteur into 5, Dirac into W, Escher into T.
HIST 201 sits in the Flexible Core, Scientific World area. Every Flexible Core course must meet three general outcomes; Scientific World requires at least three of five area-specific outcomes. This course meets all five.
| Required outcome | Where it is met |
|---|---|
| Gather, interpret, and assess information from a variety of sources and points of view | Every module pairs a primary-source claim with a modern audit |
| Evaluate evidence and arguments critically or analytically | The predict–compute–reconcile cycle, all 15 modules |
| Produce well-reasoned written or oral arguments using evidence | Peer reconciliation posts weekly; written audits M5, M13; final audit |
| Apply concepts and methods of a field to exploring the scientific world | Φ audit method applied to historical claims throughout |
| Demonstrate how problems can be analyzed and solved using tools of science, math, technology, or formal analysis | Modules 1, 2, 5, 8, 9, 11 |
| Articulate and evaluate the empirical evidence that supports a scientific or formal theory | Modules 1, 7, 8, 9 — the falsification modules |
| Articulate and evaluate the impact of technologies and scientific discoveries on today's world | Modules 11, 12, 14, 15 |
| Understand the scientific principles that underlie science-related matters of policy or public concern | Module 0 (foundations), Module 15 (closing) |
real chapter written and substantial · write new text required · gate gated prediction is load-bearing this week · 7wk survives the accelerated cut
| Wk | Module & assigned reading | What the student does, alone, on their own schedule | Published check value | Status |
|---|---|---|---|---|
| 0 | Foundations: why checking is the story new essay, from book6/wp32-forced-urgency-gap.html |
Set up the tutor first. Five minutes: install the Claude extension or open a tab, then load a persona from the tutor deck and answer its opening question. Then the Brightspace quiz: sort four one-sentence claims into true, false, unfalsifiable, merely unchecked. Tutor for this week is Primo Levi. | withheld until Wk 15 — by design. The one module where the answer is never released early. |
write 7wk |
| 1 | Kepler — the overreach and what survived it book7/ch-keplers-correspondents.html |
The anchor module. Predict: Kepler published two models in the same decade — the Platonic-solid nesting and T²∝a³. Which will fit better, and by how much? Post before seeing the data. Compute: a supplied spreadsheet with orbital period and radius for six planets; fill two columns. | T²/a³ → 0.986 to 1.000 (within 1.4% over a 40× range) Platonic solids → errors of +6.3%, +7.9%, +9.9%, +13.8%, +21.1% |
real gate 7wk |
| 2 | The cross-staff and the ledger book7/ch-cross-staff-and-ledger.html |
Build a cross-staff from a ruler and a strip of card at home. Sight any fixed object, photograph the setup, and compute the error budget: how far off is your latitude if your sighting is off by one degree? | 1° of latitude = 111.3 km = 60 nautical miles, exactly |
real |
| 3 | Faraday — checking without algebra book7/ch-faraday.html |
Faraday had almost no mathematics. Individual annotation task: mark up his 29 August 1831 induction-ring notebook entry and identify, in his own words, the sentence that constitutes the control. Rigour is not the same as formalism. | rubric-scored, not numeric — model annotation released after due date |
real |
| 4 | Hamilton — when the rules break book7/ch-hamilton.html |
Multiply two quaternions on paper in both orders; photograph the worked page. Five minutes, no prerequisites, and the student has personally produced a failure of commutativity — the property every arithmetic they have ever done silently assumed. | ij = k but ji = −k | real |
| 5 | Maxwell — the four bridges book7/ch-maxwell.html |
Predict: here are two constants measured in a laboratory with no light involved anywhere. Guess what 1/√(ε₀μ₀) will come out to, and what its units are. Post before computing. Compute: one line in a spreadsheet. Written audit assignment attached. | 299,792,458 m/s the speed of light, to nine digits |
real gate 7wk |
| 6 | Cantor — the Absolute and the checkable new text; comparative research in ch-d2-academic.html |
Run the diagonal argument on a supplied 5×5 grid of binary digits and construct the missing row. Then extend to 8×8 with your own grid. The construction is self-verifying: the new row differs from row n in position n, every time. | constructed row differs from every listed row — verify all 5 |
write 7wk |
| 7 | Pasteur — designing the control book7/ch-pasteur.html |
Given the swan-neck flask setup with the control removed, redesign it: what exactly must be held constant for the result to mean anything? Submit the design, then compare against what Pasteur actually did. The control is the whole experiment. | rubric-scored — Pasteur's own design released after due date |
real |
| 8 | Curie — measurement as the claim book7/ch-curie.html |
Compute the activity of one gram of radium-226 from its half-life and Avogadro's number. Compare against the defined value of the curie — the unit named for her. The 1.1% gap is the lesson: the unit was fixed by convention before the constants were known precisely. | computed: 3.658 × 10¹⁰ Bq defined 1 Ci: 3.700 × 10¹⁰ Bq gap: 1.1% |
real gate 7wk |
| 9 | Einstein — the 43 arcseconds book7/ch-einstein.html |
Mercury's perihelion had been drifting more than Newton allowed, unexplained from Le Verrier's 1859 measurement until 1915. Plug Mercury's orbital numbers into the relativistic correction in a supplied spreadsheet. Nothing is tuned; there is no free parameter. | 43.0 arcseconds per century observed anomaly: 43 |
real |
| 10 | Dirac — the answer you did not want book7/ch-dirac.html |
Solve E² = p²c² + m²c⁴ for E and notice there are two roots. Dirac declined to discard the negative one and predicted the positron four years before it was found. The discipline is in what you refuse to throw away. | E = ±√(p²c² + m²c⁴) — two roots, not one |
real |
| 11 | Ada Lovelace — the first program book7/ch-ada.html |
Hand-execute the first steps of Note G's Bernoulli-number loop, keeping a register table in a supplied spreadsheet. Find and document the point where the notation is ambiguous. Executing a program is a form of checking it — and this one had never been run. | B₁ = 1/6 | real 7wk |
| 12 | Ramanujan — pattern, and the bar for proof book7/ch-ramanujan.html |
Predict: count the partitions of 1 through 5 by hand, then post your guess for 6 and 7 before continuing. Compute: count 6 and 7. Then verify 1729 = 1³+12³ = 9³+10³ yourself. The module's question: which of these is a proof, and how would you know? | p(1..7) = 1, 2, 3, 5, 7, 11, 15 1729 = 1+1728 = 729+1000 |
real gate |
| 13 | Erdős — proving a thing exists without building it book7/ch-erdos.html |
Count 2-colourings of a small graph and show that if the bad colourings number fewer than the total, a good one must exist — without exhibiting one. A proof that produces certainty and no example. Written audit assignment. | R(k,k) > 2^(k/2): k=3 → 2.83, k=6 → 8.00 |
real |
| 14 | Escher — the mathematician who used graphite book7/ch-escher.html |
Classify the symmetry group of two Escher tilings using a supplied decision key, then verify by locating the generators and marking them on the image. Escher derived the tilings empirically, not knowing the classification had been proved complete in 1891. | 17 wallpaper groups, proved complete by Fedorov, 1891 |
real |
| 15 | Closing: how science corrects itself new text, narrative retelling of book6/wp29-numerology-sweep.html |
A dated correction record told as a story: a claim made, propagated, caught, and fixed, with timestamps. The Week 0 quiz reopens — students re-sort the same four claims and write on what changed. Final claim audit due. | Week 0 answers released here | write 7wk |
SPS runs accelerated terms alongside the full session. Seven modules, keeping one figure per era and keeping every module whose computation the student cannot get anywhere else:
In seven weeks the Monday-predict / Sunday-compute rhythm has to compress to a single week per module with a mid-week gate. Cut lecture-only modules before cutting any module with a computation. Weeks 2, 3, 7 and 14 go first; 4 and 9 go last, because Hamilton and Einstein are the two places a non-science student most reliably experiences having checked something difficult and found it true.
Participation cannot mean attendance in a course with no class hour. It means posting on time, which in an async course is the actual behaviour that keeps a cohort alive.
| Component | Weight | What it measures |
|---|---|---|
| Weekly computation upload | 30% | The work, not the answer. Full credit for a correctly-executed computation that reveals a discrepancy, including when the student's own prediction was wrong. |
| Gated prediction posts | 15% | Graded for being on time and specific, never for being right. A wrong prediction posted honestly scores full marks. |
| Peer reconciliation replies | 15% | Two substantive replies per module. This is the async substitute for classroom discussion and it is weighted to match. |
| Two written audits | 20% | Weeks 5 and 13. Given a historical claim and its evidence, assign a certainty tag and defend it. |
| Final claim audit | 20% | Any claim, any domain. Apply the method, produce a dated record with a defended tag. |
The entire mechanic collapses if students believe a wrong prediction costs them. They will stall, search for the answer, and post a hedge. Grading the post for specificity and punctuality — and saying so explicitly in the syllabus, in week one — is what makes an honest wrong guess safe. The wrong guesses are the pedagogically valuable ones.
A sweep of all 39 Scientist Gallery chapters, checked against the presence of a stub-note marker rather than against the index:
ada (63 KB) · curie (68) · dirac (68) · einstein (47) · escher (54) · faraday (72) · hamilton (29) · hawking (63) · hopfield (20) · huh (28) · keplers-correspondents (16) · kitagawa (38) · lattes (58) · levi (19) · maxwell (29) · metchnikoff (28) · pasteur (25) · ramanujan (19) · ramos (19) · tatiana (66) · thom (18) · thoreau (56) · connes (12) · erdos (9) · anjos (21) · cross-staff-and-ledger (15) · waddington (16) · symplectic (13) · tropical (13)
bak · dyson · katherine-johnson · klein-thymus · kovalevskaya · mirzakhani · mitchison-kirschner · noether · turing · victora-nussenzweig
Four of the outline's headline picks are stubs: Noether and Kovalevskaya, Turing and Katherine Johnson, and Mirzakhani. As originally sequenced, the birth-of-computing unit was two-thirds unwritten text. The 15-week map above is built entirely on real chapters, so the course is teachable from the repository as it stands. The ten stubs become a writing queue, not a blocker.
ch-d2-academic.html.book7/. The deck works without them; the cards are the visual layer, not a dependency.The three cards in book7/ implement their "Ask" button as a keyword-matched lookup table of about eight canned replies. That pattern will not survive contact with a real cohort — the ninth question falls through to a generic refusal, and students notice immediately. The deck replaces the mechanism with a generated prompt handed to Claude, and keeps the cards as the visual identity. Their persona text, quotes and operator diagrams are worth preserving and were the source for three of the fifteen deck personas.
HIST 201 and SCI 200 both sit in the Flexible Core, Scientific World area — students take one course from that area, not both. The two are therefore alternatives competing for the same slot, not a sequence. The zero-overlap design still matters, but for a different reason than originally stated: it means both can be offered in the same term without redundancy, and a student who takes either one leaves with the full verification methodology rather than half of it.
The natural third course is PHIL 110 — Critical Thinking, which appears twice in the SPS catalogue: once in the Flexible Core under Individual and Society, and again in the SPS College Option Core. It is the only course in the Gen Ed curriculum countable in two places, and it is the direct home for Book 6 — Φ, the audit method taught as method. That is where "verify, do not trust" stops being the frame and becomes the syllabus.