abduction

横槊赋诗

In the philosophical literature, the term “abduction” is used in two related but different senses. In both senses, the term refers to some form of explanatory reasoning. However, in the historically first sense, it refers to the place of explanatory reasoning in generating hypotheses, while in the sense in which it is used most frequently in the modern literature it refers to the place of explanatory reasoning in justifying hypotheses. In the latter sense, abduction is also often called “Inference to the Best Explanation.”

This part is only about abduction in the modern way we think about it. There's a little extra information about the older meaning, which started with a thinker named Charles Sanders Peirce—check out the Supplement: Peirce on Abduction for that.

Also, you can look at the entry on scientific discovery, especially the part that talks about discovery as abduction.

Most philosophers think that abduction (which is the same as Inference to the Best Explanation) is a way of reasoning that people use often, both in daily life and in science. However, there's still some disagreement about exactly how abduction works and how important it is.

This section will:

• Compare abduction to other types of reasoning.
• Show some important examples of how it's used, both in philosophy and elsewhere.
• Look at different ways to explain what abduction is.
• Talk about how important it is to our understanding.
• Point out possible links between abduction and a method called Bayesian confirmation theory.
  

横槊赋诗

1. Abduction: The General Idea


You know that Tim and Harry had a big fight that ended their friendship. Then, someone tells you they just saw Tim and Harry jogging together. The best reason you can think of for this is that they have made up and are friends again. So, you decide that they are friends once more.

One morning, you walk into the kitchen and see a plate and cup on the table. There are breadcrumbs and a piece of butter on the plate, and there’s a jar of jam, a pack of sugar, and an empty milk carton around it. You think that one of your housemates got up at night to make a snack and was too sleepy to clean up afterwards. This seems like the best explanation for what you see.

Sure, it’s possible that someone broke in and decided to have a snack while stealing, or maybe a housemate just arranged everything on the table to trick you into thinking someone had a midnight snack. But those ideas seem much less likely than your first thought.

While walking on the beach, you notice what looks like a picture of Winston Churchill in the sand. You might think it’s just a mark left by an ant crawling around, like in a book by Hilary Putnam called Reason, Truth, and History. But a much easier and better explanation is that someone actually drew a picture of Churchill in the sand. So, that’s what you believe happened.

In these examples, your conclusions don’t follow directly from the facts you have. For example, just because Tim and Harry had a big fight and were seen jogging together, it doesn’t mean they are friends again. You also don’t have any useful statistics about friendships or fights that can help you decide if they are friends again.

What helps you believe they might be friends again is that this idea would be the best explanation for why they were seen jogging together. Many philosophers think that if Tim and Harry are friends again, it would explain their situation well.

This kind of reasoning is called abduction or, more commonly today, Inference to the Best Explanation.

横槊赋诗

Abduction is usually seen as one of three main types of reasoning, along with deduction and induction. The difference between deduction and the other two kinds (induction and abduction) is about whether the conclusions are definitely true or not.

In deductive reasoning, if the starting facts (premises) are true, then the conclusion must also be true. For example:

• All A's are B's.
• This specific thing (a) is an A.
• So, this specific thing (a) is a B.
  

But not all reasoning works like this. For instance, if you say "John is rich" based on "John lives in Chelsea" and "Most people living in Chelsea are rich," the first statement isn't guaranteed to be true just because the second and third ones are. It’s possible that John is one of the few people in Chelsea who isn't rich.

The same goes for your conclusion that Tim and Harry are friends again just because they were seen jogging together. They could be jogging together for another reason, like discussing old business matters, and still not want to be friends again. So, it’s possible they might not have made up even if they were seen together.

We usually divide types of reasoning that aren't definite into two groups: inductive and abductive. Inductive reasoning is a mixed group, but for now, we can think of it as reasoning that relies on statistics, like how often a certain thing happens in a certain group.

For example:

• 96% of Flemish college students speak both Dutch and French.
• Louise is a Flemish college student.
• So, Louise speaks both Dutch and French.
  

Sometimes, the statistical information can be less specific, like saying, "Most people living in Chelsea are rich." There’s a lot of talk about whether inductive conclusions should be stated in numbers (like saying there's a 96% chance Louise speaks both languages) or if they can sometimes just be described generally (like saying there's a high chance it’s true).

Also, it's worth noting that some thinkers, like Harman, believe that induction is a special kind of abduction. For more on these ideas, you can look at Kyburg 1990 (Chapter 4) and Weintraub 2013.

Just because an inference is based on statistics doesn’t automatically make it inductive. For example, if you see many gray elephants and no non-gray ones, you might conclude that all elephants are gray. This conclusion would be the best explanation for what you've seen, making it an example of abductive reasoning.

To tell the difference between induction and abduction, we can say this: both types of reasoning go beyond what is directly stated in the facts (that’s why they are called non-necessary inferences). However, in abduction, you look for explanations for why things happen, while in induction, you only rely on observed patterns or statistics.

It's important to note that although both can use statistics, abduction focuses more on finding explanations, while induction sticks to just looking at numbers and frequencies.

A notable thing about abduction, which it shares with induction but not with deduction, is that it can violate monotonicity. This means you might be able to draw certain conclusions from a smaller set of premises that you can’t draw from the whole set.

For example, if you know that Tim and Harry had a terrible fight and were just seen jogging together, you might conclude they are friends again. But if you add the information that they are former business partners who still have financial matters to discuss, this might change things. Now, you might no longer think they are friends again based on all the information combined, even though the first two premises alone seemed to support that conclusion.

The reason for this change is that what would best explain Tim and Harry jogging together might not hold true anymore once you consider their history as business partners.

横槊赋诗

1.2 The ubiquity of abduction

The type of reasoning seen in the earlier examples is something many people find familiar. Both philosophers and psychologists agree that we often use abduction in everyday thinking. Sometimes, we clearly rely on abductive reasoning, but other times it happens so automatically that we don’t even notice it.

One example is how we trust what others say. This trust is thought to be based on abductive reasoning. For instance, Jonathan Adler (1994) suggests that when someone claims something is true, the best explanation for this is usually that they believe it for good reasons and want us to believe it too. Because of this, we generally have good reason to trust what they say.

This idea can also apply to how we understand language. Some argue that figuring out what a speaker means involves inferring the best explanation for why they said something in a certain context. In particular, experts in pragmatics have suggested that listeners use Gricean maxims—guidelines for conversation—to help determine the best explanation for a speaker's words when those words are unclear, overly detailed, off-topic, or otherwise strange.

In both cases, whether trusting testimony or understanding speech, the necessary abductive reasoning often happens below our conscious awareness.

Abductive reasoning isn’t just used in everyday situations; it's also important in scientific methods. Many philosophers of science argue that abduction is a key part of how science works. For example, Timothy Williamson (2007) claims that “the abductive methodology is the best science provides,” and Ernan McMullin (1992) describes abduction as “the inference that makes science.”

To show how abduction is used in science, we can look at two examples.

At the start of the nineteenth century, scientists noticed that Uranus, one of the seven known planets at the time, was not following the orbit predicted by Isaac Newton's theory of universal gravitation. Scientists also assumed there were no more planets in the solar system. One possible explanation for this deviation was that Newton's theory might be wrong. However, since Newton's theory had been very successful for over 200 years, this didn’t seem like a strong explanation.

Instead, two astronomers, John Couch Adams and Urbain Leverrier, proposed independently (but nearly at the same time) that there might be an eighth planet that hadn’t been discovered yet. They believed this idea offered the best explanation for Uranus’ unusual orbit. Soon after, this unknown planet was discovered and named "Neptune."

The second example is about the discovery of the electron by English physicist Joseph John Thomson. Thomson did experiments on cathode rays to find out if they were streams of charged particles. He reasoned this way:

Since cathode rays carry a negative charge, are deflected by electric forces as if they are negatively charged, and respond to magnetic forces like a negatively charged object would, I can only conclude that they are negatively charged particles.

Thomson's conclusion that cathode rays are made up of negatively charged particles doesn’t logically follow from his experimental results, nor did he have any relevant statistical data to support it. However, he felt there was no other reasonable explanation for his results, which suggests that he believed this conclusion was the best—and possibly the only—plausible explanation he could come up with.

There are many other examples of how abduction is used in science, as discussed in various studies and literature. For instance, some researchers like Harré, Lipton, Campanero, Aizawa, Headley, and Dellsén have explored this topic.

Abduction is also considered the main way doctors reason during medical diagnoses. Physicians typically choose the hypothesis that best explains a patient's symptoms. Various sources highlight this, including Josephson and Josephson's work on medical reasoning, as well as studies by Dragulinescu and Kind that focus on abduction in medicine and psychiatry.

Finally, abduction plays a crucial role in important philosophical debates. For example, Shalkowski discusses its place in metaphysics, and other scholars like Bigelow, Biggs and Wilson, and Schurz also explore this topic. Krzyżanowska, Wenmackers, and Douven examine how abduction might relate to the meanings of conditionals, while Williamson and Baron look at its applications in logic.

However, abduction is perhaps most significant in epistemology (the study of knowledge) and the philosophy of science, where it is often used to counter underdetermination arguments. These arguments start with the idea that several hypotheses can be equally valid based on empirical evidence, meaning that no evidence can favor one hypothesis over another. This leads to the conclusion that we cannot be justified in believing any specific hypothesis.

A well-known example of such an argument is the Cartesian argument for global skepticism. This argument suggests that our belief in reality as we perceive it is just as plausible as various skeptical scenarios, such as being deceived by an evil demon or being brains in vats connected to supercomputers. Similar reasoning has been applied to support scientific antirealism, which claims that we can't justifiably choose between different competing theories about what lies beneath observable reality.

Responses to these arguments usually highlight that the idea of empirical equivalence overlooks important explanatory aspects. This notion is often defined just by whether hypotheses make the same predictions. Critics argue that even if some hypotheses predict the same outcomes, one might still offer a better explanation for those outcomes.

If explanatory considerations matter in deciding which conclusions we can justifiably draw—according to supporters of abduction—they argue that we could be justified in believing one of several hypotheses that all make the same predictions.

Many epistemologists, following Bertrand Russell, have used abduction to counter Cartesian skepticism. They claim that although skeptical hypotheses make the same predictions as the hypothesis that reality is as we typically perceive it, they are not equally good explanations. Specifically, the skeptical hypotheses are considered much less simple than the "ordinary world" hypothesis. Various scholars, including Harman, Goldman, Moser, and Vogel, have made similar points, with Carter discussing this further. Pargetter has also offered an abductive response to skepticism about other minds.

Philosophers of science have similarly argued that we are justified in believing in Special Relativity Theory rather than Lorentz's version of the æther theory. Although both theories make the same predictions, Special Relativity is seen as a better explanation because it is ontologically simpler—it does not require the existence of an æther. Janssen provides a thorough discussion on why many philosophers prefer Einstein's theory over Lorentz's.