The art of thinking · Philosophy and logic

Occam's razor

Also known as: Ockham's razor, Principle of parsimony, Law of parsimony, Lex parsimoniae, Principle of simplicity

German: Ockhams Rasiermesser

In philosophy and the philosophy of science, Occam's razor is the principle of parsimony: among competing explanations that account for the evidence equally well, prefer the one that posits the fewest entities or assumptions. It is named after the medieval logician William of Ockham (c. 1287–1347), although the principle is older than he is and the Latin slogans usually quoted are not found in his own texts. It is a cautionary rule for theory choice, not a proof: the leaner explanation is preferred, it is not thereby shown to be true.

Thinking models
Philosophy
Problem solving

In one sentence

Occam's razor: among explanations that fit the evidence equally well, prefer the one with the fewest assumptions — a heuristic, not a proof.

Example

Investigating intermittent dropouts on a machine network, the integrator first tests the one loose connector that would explain every symptom, instead of assuming a firmware bug, a switch fault and electromagnetic interference all began on the same day.

How it applies

What the razor cuts: The common ontological formulation is that entities are not to be multiplied beyond necessity. The Stanford Encyclopedia of Philosophy distinguishes an epistemic reading (it is rational, other things being equal, to believe the simpler theory) from a methodological reading (it is rational to adopt the simpler theory as a working theory). The methodological reading is the safer one for everyday engineering and documentation work.
Attribution, carefully: The Stanford Encyclopedia notes that modern formulations are connected only tenuously to the 14th-century Ockham and that the slogan "entities must not be multiplied beyond necessity" appears nowhere in his writings, even though the sentiment is his. The Internet Encyclopedia of Philosophy records that Ockham did not invent the principle; versions of it are found in Aristotle, Aquinas and other authors he read. Formulations such as pluralitas non est ponenda sine necessitate and frustra fit per plura quod potest fieri per pauciora circulated among the Oxford Franciscans; entia non sunt multiplicanda praeter necessitatem is a later, untraceable wording.
In science: Newton put a parsimony rule at the head of Book III of the Principia (1687) — admit no more causes of natural things than are both true and sufficient to explain the appearances. Kant defends a comparable maxim in the Critique of Pure Reason as a regulative idea of reason. The razor is therefore a long-standing norm of theory building, not a law of nature.
In analysis and troubleshooting: Use it to order hypotheses, not to close the list. Test the explanation that needs the fewest independent failures first, because it is the cheapest to falsify — and record which hypotheses you ruled out, so the next person does not repeat the work.
In modeling and machine learning: A preference for the smaller hypothesis is built into regularization, pruning and model selection. This use is contested: Pedro Domingos, in "The Role of Occam's Razor in Knowledge Discovery" (1999), separates a preference for comprehensible models from the claim that simplicity predicts lower generalization error, and argues the second reading should be restricted to the few cases where it holds.
In documentation work: Applied to an information set, the razor is a question about inherited content — which sections, parameters, legacy variants and warnings in the draft still explain or prevent something, and which only survived the last revision. Deleting an assumption that explains nothing shortens the manual; deleting information a reader needs does not.
Where it stops: Parsimony is a tie-breaker between explanations that fit the evidence equally well. It never licenses removing required safety information, warnings or instructions, and it is no substitute for a risk assessment or for the evidence a conformity assessment asks for. In clinical reasoning the counterweight is stated as Hickam's dictum: a patient may have several diseases at once — real systems, too, often fail for more than one reason.
Relation to the method tradition: The razor works on candidate explanations; first principles thinking works on the premises those explanations rest on. Both oppose carrying forward a conclusion only because it was there before, which is also what Socratic questioning probes.

Parsimony vs. elegance

The Stanford Encyclopedia separates two kinds of simplicity that discussions of the razor often merge. Ontological simplicity, or parsimony, counts the kinds of entities a theory postulates — the default reading of Occam's razor is qualitative parsimony, the number of kinds, as opposed to quantitative parsimony, the number of individual things. Syntactic simplicity, or elegance, counts the number and conciseness of a theory's basic principles. A model can be parsimonious in its entities and still be inelegant in its equations, and a short, elegant formula can presuppose a crowded ontology. In documentation the same split appears: fewer concepts in the product model is parsimony; fewer rules in the style guide or the axiomatic core of an information architecture is elegance.

External references

How to read this pool

Entries, context cards and transitional topics

The art of thinking explains ways of thinking — where a method comes from, what it asks of you and how it was adapted later. You meet four kinds of content here, each with its own job:

Glossary entry: Defines one concept or method of thinking: its origin, an example and how it relates to neighboring ideas. “X vs. Y” sections set two methods side by side.
Context card of the art of thinking: A short, self-contained answer to one question about a way of thinking — its lineage, its steps, its limits. These cards are listed on The art of thinking only.
Transitional topicBridges both pools: A context card that bridges both pools: it shows how a way of thinking was adapted into documentation, analysis or quality practice. It sits with the reference context cards and is marked as transitional wherever it appears.
Reference entry or context card: Belongs to the Glossary & Norms Space on regulation, norms and engineering. Entries here link to one only where a method is also an established tool of documentation or quality work, such as the five whys or a user’s mental model.

All context cards of the art of thinking