Science is not a collection of facts — it's a method for being less wrong over time. The power of science lies not in certainty but in systematic self-correction: hypothesis → test → revision → retest. Every finding is provisional, subject to revision when better evidence emerges.

This is science's greatest strength and the source of its most common misunderstanding. When "the science changes," that's not failure — it's the process working. COVID guidance evolved because evidence evolved. Dietary recommendations change because research accumulates. The alternative — rigid certainty in the face of new evidence — is dogma, not science.

Falsifiability (Karl Popper): a claim is scientific only if it can be proven wrong. "The universe was created by an unfalsifiable being" is not a scientific claim — not because it's wrong, but because no possible evidence could disprove it. "This drug reduces blood pressure by 10mmHg" IS scientific because it can be tested and potentially falsified.

The hierarchy of evidence matters: a single study is a data point. A replicated finding is evidence. A systematic review of all available evidence is the closest we get to scientific confidence. Most "science says" claims in media reference single studies — the weakest level of evidence.

The scientific method is sound. The institutions of science are fallible.

Publish-or-perish: Academic careers depend on publishing in prestigious journals. This incentivizes: positive results (journals prefer them), novel findings (replications aren't rewarded), and quantity over quality. The incentive structure rewards publishable findings, not reliable findings.

Funding bias: Industry-funded research is 4-8x more likely to produce results favorable to the funder. Not always through fraud — more often through: favorable study designs, selective endpoint reporting, and publication of positive results while burying null findings.

Peer review limitations: Peer review catches some errors but misses many. Reviewers are unpaid volunteers who may have conflicts of interest, limited time, and expertise gaps. They evaluate methodology and logic — they don't replicate the study. Peer review is a minimum quality filter, not a guarantee of truth.

The key distinction: "this is bad science" vs "science is unreliable." The method is robust. The institutions implementing it are subject to the same incentive problems as any institution. Understanding both — the power of the method and the vulnerability of the institutions — is scientific literacy.