Occam's Razor

Occam's Razor

When several explanations fit, bet on the one with the fewest assumptions. Occam's Razor is a centuries-old rule of thumb for cutting through needless complexity.

How can it be useful to you?Whenever you're choosing between explanations or solutions, especially when some of them are convoluted or need a lot of things to be true at once.

The Missing Keys

Your car keys aren't where you always leave them. Either a) someone in your family moved them, or b) a burglar broke in, took only the keys and left everything else untouched. Occam's Razor backs option a: it needs far fewer assumptions.

The Mysterious Noise

A loud bang outside. It could be a) a car backfiring, or b) a small meteor exploding in the atmosphere. The razor points to the car, the simpler explanation that doesn't depend on a rare astronomical event.

The Vanishing Celebrity

A famous celebrity suddenly drops out of public view. Theory a): they're taking a break for personal reasons. Theory b): they've been kidnapped as part of a global conspiracy. Occam's Razor favours the first, which needs no elaborate plot.

The razor won't always hand you the truth, but it's a great first filter. Its sibling, Hanlon's Razor, applies the same instinct to people: don't assume a plot when a mistake explains it. Start with the explanation that asks the fewest leaps, and only reach for the complicated one when the simple one fails.

Occam's Razor is a rule of thumb, not a law of nature, and philosophers of science have spent a long time arguing about what it really says and why anyone should trust it.

  1. Why should simpler be truer? This is the big one. Nobody has produced a widely accepted argument that the world must be simple. Some defences rest on intuition, some on theology, and inductive defences ("simple theories have worked before") risk arguing in a circle. Richard Swinburne (1997) defends simplicity as a basic principle of reasoning, but sceptics aren't persuaded.

  2. What counts as "simple"? Simplicity is surprisingly hard to define. Fewer kinds of things? Fewer things overall? Fewer adjustable parameters? Shorter equations? These can point in opposite directions: postulating the planet Neptune added an object but simplified celestial mechanics. Nelson Goodman's famous "grue" puzzle showed that what looks simple can depend on the language you describe things in.

  3. Simplicity has to be traded against fitIn statistics, a simpler model guards against overfitting, but a model that's too simple misses real patterns. Different methods (such as AIC and BIC) weigh that trade-off differently, and Thomas Kuhn (1977) argued the weighting comes down partly to taste.

  4. Its domain may be narrowThe razor only breaks ties between explanations that fit the evidence equally well. Kenneth Holsinger (1980) argued such ties are rare in real science. Most of the time the evidence, not simplicity, decides. History also has cases where the more complicated answer won, such as oxygen molecules turning out to contain two atoms rather than one.

  5. It's context-dependentElliott Sober argues there's no single, global justification for parsimony. Whether simplicity is a good guide depends on the specific subject and the background assumptions in play. The medical maxim about horses and zebras works because horses really are more common, not because simplicity is magic, and doctors are also taught that patients can have more than one thing wrong at once.

  6. Biology isn't obliged to be tidyFrancis Crick warned that the razor can mislead in biology, because evolution builds systems by tinkering, not by optimal design. The simplest mechanism you can imagine may not be the one nature actually uses.

  7. A loose link to OckhamThe usual Latin slogan, "entities should not be multiplied beyond necessity", isn't found in William of Ockham's surviving works. That wording comes from John Punch in 1639, and modern versions of the razor are only tenuously connected to Ockham's ideas. Even in his own day it had critics: his contemporary Walter Chatton proposed an "anti-razor", arguing that if fewer things can't explain something, you must add more.

  8. "Simple" can mean "familiar": In everyday reasoning, the explanation that feels simplest is often just the one we've heard most, a trap the availability bias sets. And reaching for the tidy explanation can stop you looking for evidence that complicates it.

None of this means you should throw the razor away. As a first filter, preferring explanations that need fewer unsupported leaps is sound advice, and it guards against conspiracy thinking. Just remember it tells you where to start looking, not what the answer is.

Resources

Occam's Razor, also called the principle of parsimony or simplicity, is one of the oldest rules of thumb in science and philosophy: all else being equal, the explanation with the fewest assumptions is more likely to be right. It's named after William of Ockham, the 14th-century English philosopher and Franciscan friar who championed the idea in his writings.

The razor rarely works alone. It pairs naturally with Karl Popper's principle of falsifiability, which says a scientific theory must be testable and capable of being proven wrong. Simpler theories tend to be easier to test and easier to falsify, which makes Occam's Razor a handy guide when choosing between them.

In Bayesian inference, the statistical approach to testing hypotheses and selecting models, the razor shows up as a preference for models with fewer parameters. Simpler models are less prone to overfitting and usually easier to interpret, very much in Ockham's spirit.

The principle has spread well beyond philosophy, and has inspired other "razors" such as Hanlon's Razor. Medical students learn a version of it: "When you hear hoofbeats, think of horses, not zebras." In other words, rule out the common diagnoses before chasing rare ones.

References

  • Ockham, William of (c. 1287-1347). In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge.
  • Popper, K. (1959). The Logic of Scientific Discovery. London: Hutchinson.
  • Jefferys, W. H., & Berger, J. O. (1992). Ockham's Razor and Bayesian Analysis. American Scientist, 80(1), 64-72.
  • Senn, S. (2001). Two cheers for P-values?. Journal of Epidemiology and Biostatistics, 6(2), 193-204.