One of the dumbest sentences intelligent people keep repeating is: “The simplest explanation is usually the correct one.”
You have probably seen how it gets used. You are arguing with someone, both explanations are still on the table, evidence has not settled the matter, and then they say it: “Occam’s razor.” Usually with the calm satisfaction of someone who has just moved a rook onto your back rank. That’s it. Game over. Your explanation has four moving parts, theirs has three, William of Occam has apparently descended from the fourteenth century to referee the discussion, and you are expected to pack your things.
This has bothered me for years because the objection always seemed almost embarrassingly obvious. Why on Earth should reality be simple? I understand why we prefer simple explanations. I come from science; I like clean models, minimal assumptions, elegant mechanisms and equations that make twenty observations collapse into one idea. Simplicity is enormously useful. But I could never see how our preference for it became evidence about the structure of the world. Still, when enough smart people invoke a principle with a Latin-adjacent name and several centuries of intellectual prestige behind it, you start wondering whether you are missing something profound. I have spent enough time giving the razor that benefit of the doubt. The careful version deserves it. The version people throw across the table to end arguments does not. That one isn’t a principle, it’s a mic drop with nothing inside.
Common sense in a medieval costume
Before even getting to the serious version of Occam’s razor, it is worth noticing how often we invoke it when simplicity is not really doing the work. If you see a bird land on a rooftop in the middle of a city, you might guess the pigeon before the peregrine falcon and call that “Occam’s razor.” But a pigeon is not a simpler animal than a falcon. It is just far more common in that setting. You are reasoning from what is likely to be there, not from what is simpler. A lot of casual Occam starts this way, with perfectly sensible judgments about what usually happens getting quietly relabeled as simplicity.
Fine, the pigeon is a cheap shot. So ask a defender what the razor is actually good for and you’ll get examples like these: If you hear hoofbeats, checking for a horse before investigating whether a genetically modified zebra escaped from a secret laboratory is a good use of your afternoon. If your laptop stops connecting to Wi-Fi, restarting the router before blaming a geomagnetic storm is reasonable. This is all perfectly sensible.
It is also, frankly, a little “no shit, Sherlock.”
Of course we should investigate ordinary explanations before extraordinary ones. Of course we should avoid inventing ten invisible mechanisms when one known mechanism already explains the observation. Did we really need to attach a medieval philosopher’s name to this? If this qualifies as a profound law, I have one too: Items are always found in the last place you check. Aissam’s Law. Please cite me accordingly.
So yeah, fine, keep the razor, it stops every unexplained light in the sky from immediately becoming aliens. But the trouble starts when this harmless advice is promoted into something stronger. “Start with the simplest adequate explanation” becomes “the simplest explanation is more likely to be true.” Those are not the same claim. The first tells us how a finite mind should organize a search through an absurd number of possibilities. The second supposedly tells us something about what sort of universe we inhabit. Somewhere between the two, administrative efficiency for human reasoning quietly becomes a statement about reality.
Suppose you walk into your kitchen and find a glass shattered on the floor. Your cat is sitting nearby, looking guilty in the particular way cats look guilty while feeling absolutely nothing. The simplest explanation is obvious: the cat knocked it over. Then you check the security camera. At 3:14 a.m., a burglar entered through the window, tripped over a chair, hit the glass with his elbow, panicked, stole nothing and left.
At that moment, the stupid explanation becomes the true one.
Reality does not care that the cat theory had fewer moving parts.
That is the problem with treating simplicity as evidence. A complicated event does not become less real because a simpler story was available. Evidence can force us toward ugly explanations, unlikely explanations and explanations we would never have invented voluntarily. Reality is under no obligation to respect our preference for clean narratives.
Still, maybe I am being unfair. Maybe behind this suspiciously convenient little rule there is a deeper reason why simplicity should guide us toward truth. That is the version worth arguing with. So rather than keep making fun of a fourteenth-century friar, let’s see how far the razor actually survives once we start asking it for evidence.
What does “simple” even mean?
Fewer causes? Fewer objects? Fewer assumptions? A shorter equation? Something easier to picture? These can point in completely different directions. “The Sun moves around the Earth” is easier to picture from ordinary human experience than a rotating Earth orbiting the Sun while the whole solar system travels through a galaxy. Nobody gives geocentrism extra truth points for that. A theory can be simpler in one sense while being worse in every sense that matters.
More deeply, simplicity always lives inside a description. “DNA carries hereditary information” feels wonderfully compact today because centuries of biology are hiding behind four words. What looks elegant depends on the vocabulary we use, the knowledge we already assume and which complexity we decide to count as free. Scientists can define precise forms of complexity when they need to. But casually moving from “this feels simpler to me” to “this is probably closer to reality” asks far more of simplicity than the word can deliver.
A careful defender will therefore insist that the razor means something more disciplined: all else being equal, prefer the simpler explanation.
Fair enough, William
Suppose two astronomers explain a planet using exactly the same model, except one also insists that an invisible object follows the planet around while exerting no force, emitting no light and affecting absolutely nothing. Throw it away. Obviously. But what exactly has Occam taught us? We already defined the additional assumption as useless. Removing something that changes nothing is not evidence that reality loves simplicity; it is scientific decluttering. The phrase “do not multiply entities unnecessarily” becomes considerably less profound once we already know which entities are unnecessary.
There is a stronger argument. Suppose I say, “a ball broke the window,” and you say, “a ball broke the window, thrown by a left-handed Moroccan pharmacist wearing red shoes.” Mine is at least as probable as yours: A cannot be less probable than A plus several additional claims. I concede that completely. But both statements can be true at once. Mine leaves the details open; yours commits to them. The arithmetic rewards me for claiming less. It does not establish that less happened. The pharmacist could have been there, red shoes and all. I just need evidence before putting him in the story.
Overfitting gives simplicity another legitimate job. A model flexible enough to fit every observation can end up fitting random noise and predicting badly. Restraining that flexibility can help it make better predictions. That is a substantial reason to be cautious about complexity, but it concerns what we can reliably learn from limited data. A simpler model can be more useful without being a complete account of what produced the data.
So yes, remove idle assumptions, demand support for extra claims, and be cautious of a model that can fit anything.
The difficulty begins when we leave those clean examples behind. In a real investigation, the extra assumption may explain something, the competing theories may make different claims altogether, and we may not yet know which complexity is unnecessary. That is the point at which the razor is supposed to help us choose. It is also the point at which things get messy.
Careful what you shave off
Take the extinction of the dinosaurs. The clean story is an asteroid hitting Earth and collapsing the climate, and the evidence for its central role is overwhelming. But enormous volcanic eruptions in India were occurring around the same geological period and affecting the climate. An asteroid-only explanation is plainly simpler than one in which the asteroid and significant volcanic stress both contributed. Nature is nevertheless perfectly capable of doing two things at once.
We cannot erase the volcanoes because they make the story less elegant. We have to investigate how much they mattered. If they contributed substantially, the messier history was simply the history that happened. Here simplicity is not ambiguous. It simply cannot provide the missing information.
This is also where the pharmacist example reaches its limit. Saying that volcanism contributed is not like adding “and somewhere in the Pacific, a bacterium died that morning.” Both are technically extra facts, but only one is a live causal candidate. Massive eruptions were happening at the right time, were capable of disturbing climate, and could plausibly have affected ecosystems. The question is whether that additional mechanism earns its place in the explanation. Counting it as one more claim tells us to demand support; it does not tell us what the rocks will show.
The warning about overfitting still matters here too. A theory with more ways to explain the outcome deserves scrutiny, because flexibility makes apparent success easier. But that scrutiny has to lead back to evidence. If the rocks show that volcanism mattered, then it mattered. If they do not support that contribution, leave it out of the explanation. Extra mechanisms deserve extra evidence; they do not become less real merely because they make the story messier.
We can now see exactly where the razor helps and where its work runs out. When an added assumption is known to be useless, removing it is straightforward. When different kinds of simplicity compete, we need more than a preference for “simple” to choose between them. When extra complexity could be part of what happened, we have to investigate it.
“Usually”? Show me the numbers.
Return to the kitchen. The cat was a reasonable first suspect; the camera revealed the burglar. Our friend the defender objects: “But the razor says the simplest explanation is right more often than not. It doesn’t promise it will be right every time. You’ve shown us an exception. That doesn’t disprove the rule.” Fair enough. A good bet can lose, and finding the burglar does not establish that suspecting the cat was bad reasoning. The question is what made it a good bet in the first place.
We have an answer: cats knock things over, and burglars rarely break into kitchens just to destroy a glass and leave empty-handed. That background knowledge gives us a reason to favor the cat. But describing the favored explanation as “simpler” does not establish that simplicity earned the preference. The pigeon was more common, too; it was never a less complicated bird.
This matters because the slogan travels much further than the evidence behind those examples. When we encounter a new problem, the competing explanations might both involve familiar mechanisms, or both involve unfamiliar ones. The cleaner story might omit a contributing cause. Knowing that cats are a sensible bet in kitchens does not establish that simpler explanations are better bets elsewhere. We still need a reason to think simplicity helps us choose in this new case.
“Usually” means the simpler explanation wins more often than it loses. That is a claim about a success rate, so where is the evidence for it? To establish that pattern empirically, we would need to define what counts as “simpler,” compare explanations across a representative set of cases, and check how often the simpler one actually turns out to be right. Showing that simplicity helps in particular circumstances is a start. What still needs support is the leap from those circumstances to a rule we can confidently throw into any argument.
Nothing we have examined establishes that broader result. We have reasons why particular explanations deserve our confidence, but no general finding that fewer moving parts deserve better odds. Saying “usually” does not fill that gap. It makes room for the cases where the rule fails without establishing how often it succeeds. You have made the claim more cautious. Now make it convincing.
So what is left of the razor?
What remains is useful, but much smaller than the legend around it. Do not carry assumptions you do not need. Be suspicious of explanations flexible enough to accommodate anything. When evidence is thin, avoid committing to details you have no reason to believe. Use what you already know to decide where to look first. These are good habits because we are limited: limited in time, limited in evidence, limited in how many possibilities we can seriously entertain. The razor disciplines the investigator. It does not discipline reality.
That distinction changes what we should expect from it. It can help us organize the investigation, make more cautious claims and avoid fooling ourselves with an explanation that fits everything. It cannot spare us the work of finding out whether the cat did it, or whether two disasters contributed to the same extinction. A good starting point earns its value by helping us reach the answer, including when the answer is uglier than the place we started.
Finally, maybe the deepest explanation of the universe really will turn out to be breathtakingly simple. I hope it does. There would be something beautiful about discovering that all this apparent complexity grows from a tiny set of principles. But if that happens, simplicity will be something we discovered about reality, not something we were entitled to demand from it in advance.
I hope the universe is that considerate. I just don’t see why it should care what I hope.