Belief Correction via Costly-Accurate Signals

A student's belief in a false explanation is modeled as a probability tracked in log-odds form, updated Bayesian-style by every piece of coursework encountered. Because formal science education is expensive to produce and consume — derivations, lab work, peer-reviewed sources — each exposure carries a large, mostly truth-pointing likelihood ratio, so the cohort's log-odds drift steadily toward the correct explanation over a semester timescale, distinct from the fast, noisy contagion dynamics of casual social exposure. Tune the course-exposure rate, evidence quality and starting belief to see how quickly — and how reliably — a whole cohort converges.