Each student holds a belief expressed as a probability p that a pre-scientific (e.g. extraterrestrial-craft) explanation is correct, tracked internally as a log-odds L = ln(p / (1−p)). A formal undergraduate curriculum — labs, primary literature, worked derivations — is a stream of costly-to-produce, high-quality evidence, unlike a casual social-media post. Sequential Bayesian updating on many small, mostly-truth-pointing signals is well approximated by a drift-diffusion process on the log-odds:
λ = exposureRate / semester
ΔL = −q·λ·dt + σ·√(λ·dt)·ξ, ξ ~ N(0,1)
p = 1 / (1 + e^−L)
The drift term −q·λ·dt is the deterministic pull toward truth: every signal's expected likelihood ratio favors the correct explanation, and stronger, better-produced evidence (higher q) pulls harder per signal. The diffusion term is the noise from any single signal's ambiguity or from background low-quality exposure — it can push belief the wrong way for a while, but it doesn't grow the way the drift does, so it eventually loses out to accumulated coursework. This is the same log-likelihood-ratio random walk behind Wald's sequential probability ratio test.
- Course-exposure rate — how many distinct evidence-bearing signals (lectures, labs, readings) the cohort encounters per semester; more signals means faster accumulated drift.
- Evidence quality — the average log-odds pull of a single signal; a rigorous derivation corrects belief harder than a hand-wavy lecture aside.
- Initial belief level — the cohort's starting probability of holding the false explanation, before any coursework.
- Each student has their own engagement multiplier on the drift (attendance, study habits), which is why the individual traces fan out even though every student sees the same curriculum.
Contrast with the other simulations in this family: those model belief spreading through a social network via contagion/herding. This one models belief correcting through a single high-cost, high-quality information channel — a semester-timescale process, not a person-to-person one.