OCKHAM Ockham
Glossary中文
Glossary

Inference glossary

The concepts behind limited-clue inference, in plain language. For the method itself, see Method; to try it, go to the bench.

Reasoning methods

Occam's razor

Entities should not be multiplied beyond necessity: when two explanations fit the available clues, prefer the one with fewer assumptions and better testability. It is not "the simplest answer is always right".

Abductive reasoning

Inferring the most likely explanation for an observation — the formal name for detective-style reasoning. Unlike deduction it only gives "most likely", and it can always be overturned by new evidence.

Deductive reasoning

From general premises to a necessary conclusion: if the premises are true and the reasoning valid, the conclusion must be true. It guarantees validity but creates no new knowledge.

Inductive reasoning

From specific observations to a general rule. The conclusion is "probably true", never guaranteed, and is sensitive to sample bias.

Testable hypothesis

A statement that can be supported or refuted by observation or experiment. Good inference tells you how to test it, rather than just asserting a conclusion.

Falsifiability

A claim that cannot be refuted under any circumstance carries little information. Science favours statements that could be shown false.

Probability and judgement

Bayesian reasoning

Update a judgement with new evidence: first ask how common the thing is, then how strong the evidence is. Formally, posterior ∝ prior × likelihood.

Prior probability

How likely an explanation was before seeing the current evidence. Usually driven by base rates.

Posterior probability

The updated probability after seeing the evidence — the prior adjusted by the strength of the evidence.

Likelihood

How probable the observed clues would be if the explanation were true. High likelihood means the evidence supports it strongly.

Base-rate fallacy

Being carried away by a vivid story while ignoring how rare the thing actually is. A plausible-sounding explanation can still be very unlikely.

Probability range

Expressing uncertainty as "roughly 60–75%" rather than a single number — more honest when clues are limited.

Biases and heuristics

Anchoring

The first explanation that comes to mind drags later judgement toward it. Counter it by enumerating possibilities before evaluating any of them.

Confirmation bias

Seeking evidence that supports what you already believe and ignoring the rest. Counter it by asking what would refute the hypothesis.

Survivorship bias

Seeing only the samples that survived and missing the silent failures, which makes a method look better than it is.

Correlation is not causation

Two things occurring together does not mean one causes the other: it may be coincidence, reversed, or driven by a third factor.

Hanlon's razor

Attribute to oversight or misunderstanding what can be explained that way, rather than to malice. Often used alongside Occam's razor.

Premature convergence

Stopping at the first plausible explanation and ignoring alternatives — the most common error in abductive reasoning.