The models in this group are the philosophy and sociology of knowledge, from Bacon in 1620 to terms coined in the last decade. They answer two questions. How does knowledge change over time? And by what routes does a careful inquirer end up believing something false? The ledger itself rests on their answers.
How knowledge changes
The classic models disagree about how science advances. Hume's Problem of Induction shows that no number of past observations logically guarantees the next. The Duhem–Quine Thesis adds that no hypothesis is tested alone, so a failed prediction can always be blamed on an auxiliary assumption. Kuhn's Paradigm Shift describes science moving through long periods of normal work punctuated by revolutions. Lakatos's Research Programmes offer a middle course: judge a body of theory by whether its revisions predict new facts or merely patch old failures. Toulmin's Conceptual Evolution treats concepts as varying and being selected within a discipline.
How inquiry goes wrong
A second cluster maps the failure modes. Bacon's Idols identify four standing sources of error: those of human nature, of the individual, of language and of received systems. Galton's Problem warns that cases in comparative research are often not independent, since cultures borrow from each other. The Garden of Forking Paths, from Gelman and Loken, shows how choices made after seeing the data inflate false positives even when nobody intends to cheat. Ioannidis's Corollary calculates that under realistic conditions many published findings will be false. Tuchman's Law observes that the historical record overstates disorder, because disaster is recorded and ordinary life is not.
How reasoners fool themselves
Several recent entries, most from the rationalist writing community, name ways an individual's reasoning stops tracking the world. The Bottom Line holds that if a conclusion was fixed before the arguments were assembled, the arguments carry no evidence. A Semantic Stopsign is an answer, such as "God" or "emergence", that ends inquiry without explaining anything. Belief in Belief describes holding a belief one never acts on. Epistemic Learned Helplessness defends the opposite reflex: someone repeatedly persuaded by clever arguments that proved wrong may reasonably stop being moved by argument alone.
Where it applies
These models apply wherever conclusions are drawn from evidence. The garden of forking paths explains why pre-registration, in which researchers commit to their analysis in advance, has spread across the sciences. Duhem–Quine explains why a single failed experiment rarely overturns a theory. Scott's Metis versus Techne warns planners that practical local knowledge cannot be fully written down, which is why formal schemes imposed from above so often fail. And Mill's Handle, the ledger's own principle, explains why it exists at all: naming a pattern makes it easier to notice and to test.
Where it is argued over
The philosophy of science entries carry long-running debates. Kuhn's claim that rival paradigms cannot be compared is disputed, and Lakatos wrote partly to answer it. Ioannidis's argument is mathematics from assumed conditions rather than a measured rate, and fields differ enough that its headline does not transfer everywhere. Predictive Processing is criticised for accommodating almost any result. Mill's Handle carries its own contested flag, since naming helps bad claims travel as well as good ones.
Why it is worth knowing
This group supplies the ledger's method. Every other group reports claims with their evidence; this one explains why evidence has to be weighed rather than counted, why a failed test does not settle everything, and why careful people can reason their way to error. Knowing these models does not make anyone infallible, but it makes the routes to error visible, which is the most that any method can do.
The entries behind this essay: Model in Science & Epistemology. Every subject has its essays on the essays page.