The Enigma of Reason: A New Theory of Human Understanding

by Dan Sperber

  This example of numbers and numerals illustrates three important points:

  Our intuitions about things (here, numbers) are not the same as our intuitions about their representations (here, numerals).

  Our intuitions about representations exploit properties of the representations that need not match properties of the things represented (such as roundedness).

  Our intuitions about representations of things may nevertheless be a source of insight about the things represented themselves. (For instance, that 900 is three times 300 is a fact about the numbers themselves; this fact is intuitively grasped because of the intuitive relationship between the numerals used in the decimal system to represent these two numbers.)

  Numbers are a very special kind of thing, and numerals are a very special kind of representation. All the same, the three general points we have just made about their relationship readily extend to other kinds of metarepresentational abilities.

  Take explanations. Explanations are a certain kind of representation. Children, well before being themselves capable of providing anything resembling a genuine explanation, start asking all kinds of why-questions. That is, they start requesting explanations for a wide variety of things. Soon enough, they start themselves providing explanations. More generally, asking for and providing explanations is a common aspect of conversation across culture.

  We have clear intuitions about the “goodness” of various explanations. As the psychologist Frank Keil and his colleagues have shown, these intuitions may not be very reliable when they concern our own ability to provide explanations.16 We often greatly overestimate, for instance, our ability to explain how domestic appliances we use every day actually work. We are, however, better at evaluating the explanations given by others. Even children are typically quite adept at recognizing the expertise of others in specific domains of explanation, and at taking advantage of it. Keil describes a division of cognitive labor between seekers and providers of explanation in different domains that is quite similar to the division of cognitive labor that is often at work in the exchange of arguments (as we will show in Chapter 15).

  Quite parallel to the case of numbers and numerals,

  Our intuitions about good and bad explanations are not the same as our intuitions about the things explained.

  Our intuitions about explanations exploit properties such as cogency, generality, or coherence that are properties of the explanations themselves and not of the things explained.

  Our intuitions about explanations (which make us prefer good explanations) is nevertheless a major source of insight about the things explained.

  More generally, representations are a very special kind of thing in the universe. They are found only inside and in the vicinity of beings with minds. By any sensible criterion, metarepresentational modules and the inferences they perform about representations are very specialized, domain- or task-specific devices. At the same time, our inferences about representations are quite relevant to our understanding of the things represented.

  If you intuitively grasp that 900 is three times 300, your intuition is driven by properties of the numerals, but the relevant information you gain is about the numbers these numerals represent. If you recognize the cogency of a good explanation of, say, how a dual-technology motion detector works, then you have learned something not just about the explanation but also, and more importantly, about motion detectors.

  Mindreading, arguably the most important of our metarepresentational modules, is no exception. It informs you, through your intuitions about what others believe, about the subject matter of their beliefs. When the woman in the waiting room looks at her watch and sighs, you guess not only that she believes that the doctor is keeping her waiting but also that indeed the doctor is keeping her waiting. From this, you may draw further inferences of your own, such as about how well the doctor keeps her appointments. We learn a lot about the world by discovering what other people think about it. Mindreading provides a window on the world at large.

  Metarepresentational modules provide information not only about the representations metarepresented but also, indirectly, about the things these representations represent. So while, as we insisted, these modules have very specific domains, namely, specific aspects of specific kinds of representations, they may nevertheless have a different and much wider virtual domain corresponding to the things represented.

  A standard objection to modularist views of the human mind is that reasoning couldn’t possibly be modular since it is not specialized, not restricted to a single domain. Indeed, humans can and do reason about everything and anything. They bring together in their reasoning evidence pertaining to quite different domains.

  To compute the circumference of the earth, the Alexandrian scholar Eratosthenes, for instance, used a mixture of astronomy, geography, geometry, and local observation of shadows. Or, to take a more recent debate, consider the ancient piece of cloth known as the Shroud of Turin. Is it, as some claim and others deny, the very burial shroud of Jesus and hence a sacred relic? Arguments on the issue have drawn on history, theology, various branches of chemistry, and physics (radiocarbon dating methods in particular). Aren’t these two examples—Eratosthenes’s discovery and the debate on the Shroud of Turin—perfect illustration of the fact that human reasoning isn’t domain specific, let alone modular?

  Actually, while such illustrations raise interesting issues, they do not provide strong objections to the modularist approach. They ignore the fact that metarepresentational modules may have virtual domains that extend way beyond their real domain.

  Reasoning, we will argue, is based on a metarepresentational module that provides intuitions not about the world in general but about reasons. Reasons are a kind of representation. The real domain of the reason module—reasons—is rather narrow. Still, reasons themselves can be about anything or any combinations of things in the world, bringing together, for instance, the pace of camels in the desert and the position of the sun in the sky of Alexandria at noon on the summer solstice, or the story of the crucifixion of Jesus and the rate of radioactive decay of radiocarbon. Reasoning, therefore, can be both quite specialized in its operations and quite general in its import. In fact, its universal import is best explained by the specific properties of its specialized domain. Inferences about reasons that are themselves about anything result in a kind of virtual domain-generality.

  III

  * * *

  RETHINKING REASON

  What is reason? How does it work? What is it for? How could it evolve? In Chapters 7 through 10, we develop a novel interactionist approach that answers these questions. Reason, we argue, is a mechanism of intuitive inferences about reasons in which logic plays at best a marginal role. Humans use reasons to justify themselves and to convince others, two activities that play an essential role in their cooperation and communication. Just as echolocation evolved as an adaptation to the ecological niche inhabited by bats, reason evolved as an adaptation to a very special ecological niche, a niche that humans built and maintain for themselves with their intense social relationships, powerful languages, and rich culture.

  7

  How We Use Reasons

  Humans appeal to reasons not just in reasoning but also in explaining and justifying themselves.1 Nevertheless, explanation and justification on the one hand and reasoning on the other have been studied independently of one another, as if two quite different psychological mechanisms were involved. We believe that the difference is less one of mechanism than of function, and we want to articulate, in this third part of the book, an integrated approach to the psychology of reasons.

  Why do you think this? Why did you do that? We answer such questions by giving reasons, as if it went without saying that reasons guide our thoughts and actions and hence explain them. These reasons are open to evaluation: they may be good or bad. Good reasons justify the thoughts or actions that they explain. This picture of the role of reasons in explanation and in justification may see
m self-evident. It is based, however, on a convenient fiction: most reasons are after-the-fact rationalizations. Still, this fictional use of reasons plays a central role in human interactions, from the most trivial to the most dramatic.

  The Commonsense Picture

  Start with the dramatic: in the middle of the night of November 2, 2013, Theodore Wafer, a middle-aged man living in Dearborn Heights, Michigan, a white suburb of Detroit, was awakened by loud banging on his front door. Earlier that night, Renisha McBride, a young African American woman, had crashed her car, walked out, and wandered for hours in a state of confusion. She ended up at Wafer’s door, probably looking for help. Wafer thought his house was being attacked, took his shotgun, opened the door, and fired, killing McBride. At the trial, he explained his action by saying that he had been afraid for his life and had shot in what he believed was self-defense against several attackers. An unreasonable fear, the prosecution argued. In any case, even if the fear had been reasonable, the reasonable response would have been not to open the door but to lock himself in and call the police. Wafer was found guilty of second-degree murder.

  Why did Wafer act the way he did? No doubt the reasons he gave at the trial were the best he could muster. We will never know for sure, and it may well be that he himself didn’t know what exactly had gone on in his mind at the time. But say we accept that he acted for the reasons he himself invoked: he felt under attack and wanted to defend himself. Even so, most of us would agree with the prosecution and the jury that he didn’t have good enough reasons to believe he was in great danger or to act the way he did.

  As the case of Wafer illustrates, some reasons may seem good enough to explain but not good enough to justify. We can accept an explanation and, at the same time, be critical of the reasons it invokes. For the purpose of explanation, it is enough that these reasons should have seemed adequate to the person we are trying to understand. On the other hand, to judge that what the person thinks or does is justified, the same reasons must seem adequate to us.

  Much less dramatically, we all invoke minor or even minute reasons at every turn in daily social interactions. Rob, for instance, asks Ji-Eun, “Do you want a milkshake?” and she answers, “Thank you, but, you know, most of us Koreans are lactose intolerant.” Why doesn’t Ji-Eun just say no? Why does she bother to give a reason for her refusal? By mentioning a reason for declining Rob’s offer of a milkshake, Ji-Eun suggests that she is appreciative of the offer and might have accepted it otherwise. Rob is unlikely to question the reason invoked by Ji-Eun. What should matter to him is that by giving a reason rather than just saying, “No!” Ji-Eun has been considerate toward him. In such ordinary interactions, our giving reasons manifests the kind of consideration others can expect of us and we might expect of them.

  How must people understand reasons to be able to use them as they do in their thinking and interactions? To answer this question, it would be of limited use to look at the way people understand the word “reason” itself (in the sense of a reason for something). This understanding varies across people, across circumstances, and, obviously, across cultures. The English word “reason” in that sense doesn’t have straightforward translations in all languages. This fuzziness and this variability in the way people think and talk about reasons in general are worth studying in their own right. What we are investigating here, though, are not folk notions or folk theories of reasons but the way people make use of particular reasons and do so whether or not they categorize them as “reasons.”

  To make better sense of the way reasons are produced and used, it will be helpful to make a distinction between objective and psychological reasons. Such a distinction (with subtle differences and various terminological choices) is common in the rich philosophical literature on reasons. For our present purpose we will, when useful, draw inspiration from this literature without getting involved in its controversies.2

  An objective reason, as we will use the phrase, is a fact that objectively supports some conclusion. This conclusion may be descriptive (about what is the case) or practical (about what is desirable). For instance, the fact that today is Friday is an objective reason to conclude that tomorrow will be a Saturday; the fact that the plums are ripe is an objective reason to pick them from the tree without waiting.

  Facts, as we are using the term, are true propositions, abstract objects without causal powers. It is not, strictly speaking, the fact that some plums are ripe that causes them to fall from the tree; it is their ripeness itself or, to be more precise, it is the chemical and physical changes that define ripeness. Objective reasons, being facts, are themselves without causal powers. From our empirical point of view, it doesn’t matter whether objective reasons exist in the world independently of human interests and what well-defined criterion of objectivity they meet, if any. If we talk about objective reasons at all, it is because they are represented both in what people think and in what they say and, unlike facts, representations of facts do have causal powers.

  A psychological reason is a mental representation of an objective reason. Like all mental representations, it may be a misrepresentation; that is, it may represent as an objective reason for a conclusion a false proposition rather than a fact, or it may represent a genuine fact as supporting a conclusion it actually does not support. Still, in our cognitive and evolutionary perspective, we assume that in general, people tend to represent as objective reasons for a given conclusion facts that do indeed support it: cognition is imperfect but not random. Psychological reasons are mental representations in the brain and, as such, play a causal role in people’s lives. (When we use “reasons” without qualifying the term, we are talking about psychological reasons.)

  It is generally thought that the main role psychological reasons play is to motivate and guide people’s actions and beliefs (guidance being little more than a fine-grained form of motivation and motivation a coarse-grained form of guidance). We disagree. The main role of reasons is not to motivate or guide us in reaching conclusions but to explain and justify after the fact the conclusions we have reached.

  The uses of reasons to explain and to justify are not just related; they are intertwined. To explain people’s beliefs or decisions, psychological reasons must at least point in the direction of a conceivable justification, that is, of a good, objective reason. Wafer, for instance, could, perhaps not with sufficient good sense but not absurdly either, see his reasons as objectively justifying his shooting the person at the door. Had he, on the other hand, given as his reason for having fired a shot that Elvis Presley was dead and that therefore life was meaningless, we would see this not as a genuine reason-based explanation, not even a defective one, but as an admission (or a claim) of temporary insanity.

  Similarly, when we invoke reasons to justify other people’s thoughts or actions, the normal implication is that they had these reasons in mind and were guided by them. How could people be justified by reasons they knew nothing about? This seems to be a commonsensical constraint on justification, but is it really? Actually, there are interesting exceptions described in the philosophical literature as cases of “moral” or “epistemic luck.”3

  Imagine, for instance, what would have happened if the person Wafer shot and killed had turned out to be not a young woman looking for help but a dangerous criminal on the “Most Wanted” list. In that case, Wafer’s action might well have been commended even though he would have acted without knowing that there happened to be good, objective reasons to do what he did—a typical case of moral luck. Reasons must be in people’s minds to explain their behavior, but in some cases at least, they need not be in their minds to make their behavior either blameworthy or praiseworthy.

  Just as there are cases of moral luck—an action undertaken with inadequate reasons in mind but that turns out to have had, unbeknownst to the agent, an excellent objective justification—there are cases of epistemic luck—a belief held for inadequate reasons but that turns out to be justified all the same.

  T
he Alexandrian scholar Eratosthenes, as we saw in Chapter 1, was the first person ever to measure the circumference of the earth. The result of his calculation, 252,000 stades, was just 1 percent shy of the modern measurement of 24,859 miles, or 40,008 kilometers. What makes this precision even more astonishing is that Eratosthenes had made two mistakes in his assumptions. It was essential to his calculation that the town of Syene should be right on the Tropic of Cancer and due south of Alexandria, and he was convinced that such was indeed the case. Actually, Syene is 1° north of the Tropic of Cancer and 3° east of Alexandria. So how could he come so close? By pure chance, the effects of these two errors practically canceled one another out. This bit of epistemic luck has not been used to downplay Eratosthenes’s accomplishment. It is as if his overall method was so brilliant and his result so impressive that the inadequacy of some of his reasons was irrelevant.

  Moral and epistemic luck are puzzling phenomena: they suggest that we may find people justified by reasons that actually didn’t motivate or guide them and that therefore cannot explain their thoughts or actions. Could the commonsense picture of reasons not be as clear and coherent as it seems at first sight?

  The Commonsense Picture Challenged

  Contrary to the commonsense picture, much experimental evidence suggests that people quite often arrive at their beliefs and decisions with little or no attention to reasons. Reasons are used primarily not to guide oneself but to justify oneself in the eyes of others, and to evaluate the justifications of others (often critically). When we do produce reasons for guidance, most of the time it is to guide others rather than ourselves. While we would like others to be guided by the reasons we give them, we tend to think that we ourselves are best guided by our own intuitions (which are based, we are sure, on good reasons, even if we cannot spell them out).