In science as well as everyday life, we often feel the pull of simpler, more elegant, or more beautiful explanations.
What makes an explanation a good explanation? Isn’t the simplest explanation always the best? Why do people often swallow crazy explanations? Those are just some of the questions we’re asking in this week’s episode about the nature of explanation.
Let’s start with an innocent seeming question. Suppose we have two competing explanations of the same thing. How do we decide between them? One perfectly reasonable thought is that we should always start with the simpler one. But what justifies this thought? You might suspect that the simpler explanation is more likely to be true. There is, I think, something to this thought. But it is also important to say that sometimes a simple explanation might be too simple. If we are going to take simplicity as a guide to explanation, we need what might be called a relative rather than an absolute notion of simplicity. Our explanations had better be simple relative to the evidence we seek to explain. We should not offer explanations that are so simple that they don’t explain the evidence, but we also shouldn’t make our explanations more complicated than the evidence demands.
But now we face another question. How exactly do we strike a balance between explanations that are too simple given the evidence, and explanations that are too complex given the evidence? Here, I’m afraid that there are no hard and fast rules, and certainly no algorithm that we can apply. It’s more a case-by-case, you know it when you see it, kind of thing.
There is, however, a name for the kind of reasoning we employ in trying to arrive at explanations in the sort of situations I have in mind. Philosophers call it abductive reasoning. I bet most people are familiar with inductive reasoning and deductive reasoning, but abductive reasoning is probably a less familiar term. Abductive reasoning or inference is also sometimes called inference to the best explanation, but that is probably not any more transparent. Let me give an example to illustrate the concept.
Suppose you’re a detective investigating the murder of Jimmy Socrates. You’ve got two main suspects, Freddie Nietzsche and Tommy Hobbes. Here’s the case against Nietzsche. He and Socrates were spotted having a heated argument shortly before Socrates’s death. And the police found the murder weapon in the bushes in Nietzsche’s backyard.
That sounds pretty damning. But now let’s consider Hobbes’s. Hobbes hated Socrates too. That night, an eyewitness saw a man matching Hobbes’s description running from the scene of the crime. Before you conclude that Hobbes is our man, consider this additional fact. Only an hour after the murder, Hobbes was spotted in a restaurant a hundred miles away, having a quiet, intimate dinner with his partner.
One response you might have here is to think that since we got inconclusive evidence that supports two different hypotheses—on the one hand, we’ve got the heated argument and the gun in the bushes pulling us toward one hypothesis, while on the other, we’ve got eyewitness testimony pulling us toward the other—our only choice is to dig for more evidence in the hopes of finding some crucial bit of evidence that will conclusively confirm one hypothesis, while disconfirming the other.
But an abductive reasoner might say we don’t need to do that at all. We’ve got enough evidence already. We just need to bring with a little abductive reasoning to bear to help us settle the question. Surely, the adductive reasoner might say, it was Nietzsche who did it. If not, how did the gun get into his yard? And how did Hobbes get so far away from the scene of the crime so fast? That’s the beauty of abduction, even if our evidence is in some sense incomplete and inconclusive, it can help us settle on the best explanation, without a lot of aimless mucking about in search of more evidence.
But is abductive reasoning really a reliable means of finding the truth? It may reliably lead us to hit upon the simplest explanation, but why think simplest is always best, especially if by best we mean "true?"
Consider, for example, the following alternative explanation of Socrates’ demise. Perhaps Hobbes is framing Nietzsche. Knowing they would come to blows, Hobbes arranged for Nietzsche and Socrates to meet where he knew they would be seen. After they part company, Hobbes murders Socrates and plants the gun in the yard. Of course, so far, we’ve left out the inconvenient fact of the intimate dinner a hundred miles away. How do we account for that fact? Well, perhaps Hobbes had a get-away-vehicle, capable of whisking him a hundred miles away in a relatively short time, waiting at the ready. His vehicle of choice was a helicopter! It was parked at a nearby heliport. He kills Socrates, gets into the chopper, drops the gun as he flies over the yard, lands the chopper a hundred miles away, where a driver is waiting to take him to his dinner date.
This may sound … a little bonkers, at least given the evidence so far. Certainly its complicated, at least given the evidence so far. And it's just such apparently evidentially unmotivated complications, as we might call them, that abductive inference tends to shy away from. So unless and until we decide to check out heliports in the area for suspicious takeoffs and landings, we seem have almost no reason to consider it as a hypothesis. Certainly, abductive reasoning alone will not force of even tempt us to consider it.
But that fact raises the following issue. How can we be confident that we know the best explanation, if we won’t even allow ourselves to consider every logically possible explanation?
On the one hand, you might think that that’s the beauty of abductive inference. It doesn’t require us to drive ourselves crazy, searching for evidence to rule out every merely conceivable alternative. If an alternative is ad hoc, or too complicated, or inconsistent with what we already know, we can just ignore it.
Point taken, but none of that guarantees that the alternatives that we rule out because they lack certain explanatory virtues—like simplicity, to name just one—will necessarily be false. Unless we have some more reliably truth-tracking method for determining which logically possible alternatives are worth taking seriously and which aren’t, we will have no guarantee that the explanation which strikes us as best will necessarily be true. And in the absence of any such guarantee, abductive inference can come to seem at least a little like hocus-pocus.
It may be mysterious, but I am not prepared to dimisss abduction altogether. People engage in abductive reasoning all the time, after all. And it’s not just scientists. It’s ordinary people too. And even babies seem to do it. And for the most part, we are remarkably good at it, despite its pitfalls.
This means that our task as philosophers and psychologists and cognitive scientists is to take the frequent success of abductive reasoning as a given and then to try figure out how and why it works so well, but also when it goes wrong. And that’s precisley what we try to do this week. We’d love to have your help.