What Makes Something the Best Explanation? (And Should We Believe It?)

You walk into the kitchen one morning. On the table there’s a plate with breadcrumbs on it, a butter knife, an empty glass, and an open jar of jam. Nobody else is up yet.

You don’t need to think hard. You know what happened: someone had a midnight snack and was too tired to clean up. You didn’t see it happen. You don’t have any statistics about how often midnight snacks lead to dirty dishes. You just know that this is the best explanation for what you’re looking at.

But here’s the strange thing: philosophers have been arguing for over a hundred years about whether this kind of reasoning is actually trustworthy. When you decide something is the “best explanation,” are you really learning something about the world? Or are you just telling yourself a satisfying story?

How We Actually Think

Let’s start with something obvious: most of what you believe, you didn’t figure out by logic alone.

If someone tells you they saw Tim and Harry jogging together, and you know they had a huge fight last week and weren’t speaking, you’ll probably conclude they made up. But that doesn’t follow logically. Maybe they’re former business partners who still have money stuff to discuss and decided to combine it with exercise. That’s also possible. You just think “they made up” is the better explanation.

This kind of reasoning is called abduction, or Inference to the Best Explanation. It’s different from deduction, where if your premises are true, the conclusion must be true. (Like: all humans are mortal. Socrates is human. So Socrates is mortal.) It’s also different from induction, where you use statistics or patterns. (Like: 96% of students at this school can speak two languages. Maria goes there. So Maria probably speaks two languages.)

Abduction doesn’t give you certainty, and it doesn’t give you probabilities. It gives you an explanation that feels right. And here’s the thing: we use it constantly. Not just for midnight snacks and reconciled friends, but for almost everything.

When you trust what someone tells you, you’re using abduction. The best explanation for why they’re saying “there’s a test tomorrow” is that they believe it’s true and want you to know. When you figure out what a sarcastic remark means, you’re using abduction. When a doctor looks at your symptoms and says “it’s just a cold,” that’s abduction too. The symptoms could be lots of things, but the cold explanation is the best one.

Scientists do this all the time. In the 1800s, astronomers noticed that the planet Uranus wasn’t moving the way Newton’s laws predicted. They could have concluded that Newton was wrong. But instead, two astronomers independently suggested that there must be another planet nobody had discovered yet, pulling Uranus off course. That was the best explanation. And they were right—they found Neptune exactly where they predicted. When J.J. Thompson figured out that cathode rays were made of tiny negatively charged particles (electrons), he wasn’t using deduction. He was saying: the best explanation for what I’m seeing in my experiments is that these are particles. I can’t see any other plausible way to make sense of this.

Abduction is everywhere. The question is: should we trust it?

The Problem with “Best”

Here’s where it gets tricky. Most textbooks describe abduction like this:

Given some evidence, and a bunch of possible explanations for it, pick the one that explains it best and believe that one is true.

Sounds reasonable. But there’s a problem that philosophers call the bad lot objection.

Imagine you’re trying to figure out why your phone died. You think of three explanations: (1) the battery ran out, (2) the charger is broken, (3) there’s a software glitch. You pick number 1 because it’s simplest. But what if the actual explanation—say, a tiny manufacturing defect that only 0.001% of phones have—never even occurred to you? You just picked the best from a bad lot. You might have missed the real explanation entirely.

This isn’t some weird edge case. Think about it: when you come up with possible explanations for something, you’re almost certainly not thinking of all of them. You’re thinking of the ones that come to mind easily. Why should the best explanation you can think of be the same as the actual best explanation? That would require some kind of special luck—as if humans were somehow designed to always hit on the right explanations.

Some philosophers think this is a fatal problem for abduction. Others think it can be fixed by changing how we state the rule. Instead of saying “infer that the best explanation is true,” maybe we should say something more modest: “infer that the best explanation is closer to the truth than the other ones you thought of.” Or maybe we should only infer that something is true if the best explanation is actually good enough, not just better than the alternatives.

But nobody has fully solved this. It’s still a live debate.

Are We Just Tricking Ourselves?

There’s another worry, raised by philosopher Bas van Fraassen. He points out that the more detailed and informative an explanation is, the more ways it has of being wrong. If I say “someone ate toast,” that’s pretty safe. If I say “my brother ate toast with strawberry jam at 2:17 AM while wearing his blue pajamas,” that’s much more informative—but also much more likely to be false if any detail is off. So why should we think that the most explanatory theory (which tends to be more detailed and specific) is the most likely to be true?

This is a serious challenge. But defenders of abduction respond that it’s not always clear that better explanations have “more ways of being wrong.” Take the discovery of Neptune. The explanation “there’s another planet” and the explanation “Newton’s laws are wrong” both make the same predictions about where Uranus should be. They’re equally specific. But one is a much better explanation than the other. So explanatory power isn’t just about being detailed—it’s about making sense of things in a way that fits with everything else we know.

Can You Prove That Abduction Works?

Some philosophers have tried to argue that abduction must be reliable because science keeps working. Look at all the successful predictions science has made. Those predictions came from theories that were accepted partly because they were the best explanations. So abduction must be a good method, right?

The problem is that this argument uses abduction itself. It says: “The best explanation for why science is so successful is that our methods (including abduction) are reliable.” But if you’re trying to prove that abduction is trustworthy, you can’t use abduction in your proof—that’s circular. It would be like trying to prove that a witness always tells the truth by asking the witness to swear they’re telling the truth.

Philosophers argue about whether this kind of circularity is actually a problem. Some say it’s fine, as long as it’s not the kind of circularity where you assume the very thing you’re trying to prove. Others think it’s hopeless. It’s one of those philosophical debates that doesn’t seem close to being settled.

What About Probability?

Here’s another way this debate shows up. There’s a very popular theory called Bayesianism that says we should update our beliefs using probability math. When you get new evidence, you should adjust how strongly you believe things according to a specific formula.

Bayesianism makes no mention of explanation at all. It just talks about probabilities. So some philosophers have asked: is abduction compatible with Bayesianism, or are they rivals?

Some say they’re compatible—in fact, that abduction can help Bayesianism. The problem with Bayesianism is that it doesn’t tell you where your initial probabilities come from. You have to start somewhere, and explanatory considerations might help you decide what’s plausible. If one theory is much more elegant and simple than another, maybe it deserves a higher starting probability.

Other philosophers think this is backward. They think Bayesianism and abduction are just doing different jobs. Bayesianism is about how confident you should be (degrees of belief), while abduction is about what you should actually believe (categorical belief). They might both be useful, but for different purposes.

So What Should You Do?

Nobody really knows whether abduction is ultimately justified. What philosophers mostly agree on is that humans do use it constantly, and we’d be paralyzed without it. If you refused to ever infer the best explanation, you couldn’t trust anything anyone told you, you couldn’t diagnose why your bike won’t start, and you couldn’t even make sense of most conversations.

The question is whether this is just a useful habit that sometimes works, or whether it’s a genuinely rational method that tells us something true about the world. Philosophers are still arguing about this, and it’s not clear the debate will ever be completely resolved.

In the meantime, you’ll keep using abduction—and you should. Just maybe with a little more awareness that every time you say “the best explanation is…”, you’re betting that the explanation you thought of is actually the right one, and that there isn’t some better explanation you never considered.

Key Terms

Term	What it does in this debate
Abduction	A type of reasoning where you infer that something is true because it would be the best explanation for what you observe
Deduction	Reasoning where the conclusion must be true if the premises are true (used as a contrast to show what abduction is not)
Induction	Reasoning based on statistics or patterns (another contrast—abduction uses explanatory value, not just frequencies)
Inference to the Best Explanation (IBE)	Another name for abduction; captures the idea that we’re picking the best candidate from a set of possible explanations
Bad lot objection	The worry that the best explanation among the ones you thought of might not actually be the true one; the real explanation might not have occurred to you
Bayesianism	A theory about how to rationally update beliefs using probability; doesn’t mention explanation, which creates a puzzle about how it relates to abduction
Rule-circularity	When an argument uses the very rule it’s trying to justify (like proving abduction works by using abduction); some think this is okay, others don’t

Key People

Charles Sanders Peirce (1839–1914) – An American philosopher and scientist who first named and described abduction as a distinct type of reasoning, different from deduction and induction.
Bas van Fraassen (born 1941) – A philosopher who has been one of the strongest critics of abduction, arguing that we have no good reason to think the best explanation is likely to be true.
Peter Lipton (1954–2007) – A philosopher who defended abduction and argued that even Bayesians should make room for explanatory reasoning in their theories.
Stathis Psillos (born 1965) – A philosopher who defended the reliability of abduction by arguing that rule-circular arguments aren’t necessarily bad.

Things to Think About

Think of a time you were wrong about something because the “best explanation” turned out to be false. What was the explanation you believed? What was the actual explanation? Was there any way you could have known you were missing something?
The bad lot objection says we can’t be sure the real explanation is among the ones we thought of. But practically speaking, how would you even try to think of all possible explanations for something? Is that even possible? Does it need to be possible for abduction to be useful?
If you had to choose between two rules for your life—“always believe the best explanation you can think of” or “never believe anything that isn’t proven by logic or statistics”—which would you pick? What would go wrong with the other choice?
Some philosophers say abduction is how we know other people have minds. The best explanation for why your friend acts like they have thoughts and feelings is that they do. But you can’t prove it. Is that a good enough reason to believe it?

Where This Shows Up

Medical diagnosis: Doctors constantly use abduction when they match symptoms to the most likely illness. It’s why they ask about things that seem unrelated—they’re testing different explanations.
Detective work and crime shows: Every episode of a mystery show is basically a contest of competing abductions. The detective picks the explanation that makes the most sense of all the clues.
Science classes: When you learn about how scientists discovered atoms, germs, or tectonic plates, you’re seeing abduction in action. None of those things were directly observable when first proposed—they were the best explanations for what was observable.
Everyday conversation: When you interpret someone’s tone of voice or figure out whether they’re being sarcastic, you’re using abduction. You’re picking the meaning that best explains why they said what they said the way they said it.
AI and machine learning: Some computer scientists are trying to build AI systems that use abduction to make inferences, because logic-only systems are too rigid for messy real-world problems.