Hugo Mercier and Dan Sperber


THE ENIGMA OF REASON

A New Theory of Human Understanding

ALLEN LANE

UK | USA | Canada | Ireland | Australia

India | New Zealand | South Africa

Allen Lane Books is part of the Penguin Random House group of companies whose addresses can be found at global.penguinrandomhouse.com.

Penguin Random House UK

First published in the United States of America by Harvard University Press 2017

First published in Great Britain by Allen Lane 2017

Copyright © Hugo Mercier and Dan Sperber, 2017

The moral rights of the authors have been asserted

Cover image from the title page of Discourse on the method, Dioptrics, Meteors, and Geometry, by René Descartes (1596–1650), edition published in Paris, 1668.

Bridgeman Images.

Design: Jim Stoddart

ISBN: 978-1-846-14558-2

Contents

Introduction: A Double Enigma

I  SHAKING DOGMA

 1 Reason on Trial

 2 Psychologists’ Travails

II   UNDERSTANDING INFERENCE

 3 From Unconscious Inferences to Intuitions

 4 Modularity

 5 Cognitive Opportunism

 6 Metarepresentations

III  RETHINKING REASON

 7 How We Use Reasons

 8 Could Reason Be a Module?

 9 Reasoning: Intuition and Reflection

10 Reason: What Is It For?

IV WHAT REASON CAN AND CANNOT DO

11 Why Is Reasoning Biased?

12 Quality Control: How We Evaluate Arguments

13 The Dark Side of Reason

14 A Reason for Everything

15 The Bright Side of Reason

V REASON IN THE WILD

16 Is Human Reason Universal?

17 Reasoning about Moral and Political Topics

18 Solitary Geniuses?

Conclusion: In Praise of Reason after All

Notes

References

Illustration Credits

Acknowledgments

Follow Penguin

Introduction: A Double Enigma

They drink and piss, eat and shit. They sleep and snore. They sweat and shiver. They lust. They mate. Their births and deaths are messy affairs. Animals, humans are animals! Ah, but humans, and humans alone, are endowed with reason. Reason sets them apart, high above other creatures—or so Western philosophers have claimed.

The shame, the scandal of human animality, could at least be contained by invoking reason, the faculty that makes humans knowledgeable and wise. Reason rather than language—other animals seemed to have some form of language too. Reason rather than the soul—too mysterious. Endowed with reason, humans were still animals, but not beasts.

Reason: A Flawed Superpower?

With Darwin came the realization that whatever traits humans share as a species are not gifts of the gods but outcomes of biological evolution. Reason, being such a trait, must have evolved. And why not? Hasn’t natural selection produced many wondrous mechanisms?

Take vision, for instance. Most animal species benefit from this amazing biological adaptation. Vision links dedicated external organs, the eyes, to specialized parts of the brain and manages to extract from patterns of retinal stimulation exquisitely precise information about the properties, location, and movement of distant objects. This is a hugely complex task—much more complex, by any account, than that of reason. Researchers in artificial intelligence have worked hard on modeling and implementing both vision and reasoning. Machine vision is still rudimentary; it comes nowhere near matching the performances of human vision. Many computer models of reasoning, on the other hand, have been claimed (somewhat optimistically) to perform even better than human reason. If vision could evolve, then why not reason?

We are told that reason, even more than vision, is a general-purpose faculty. Reason elevates cognition to new heights. Without reason, animal cognition is bound by instinct; knowledge and action are drastically limited. Enhanced with reason, cognition can secure better knowledge in all domains and adjust action to novel and ambitious goals, or so the standard story goes. But wait: If reason is such a superpower, why should it, unlike vision, have evolved in only a single species?

True, some outstanding adaptations are quite rare. Only a few species, such as bats, have well-developed echolocation systems. A bat emits ultrasounds that are echoed by surfaces in its environment. It uses these echoes to instantaneously identify and locate things such as obstacles or moving prey. Most other animals don’t do anything of the sort.

Vision and echolocation have many features in common. One narrow range of radiation—light in the case of vision, ultrasounds in the case of echolocation—provides information relevant to a wide variety of cognitive and practical goals. Why, then, is vision so common and echolocation so rare? Because, in most environments, vision is much more effective. Echolocation is adaptive only in an ecological niche where vision is impossible or badly impaired—for instance, when dwelling in caves and hunting at night, as bats do.

Is reason rare—arguably unique to a single species—because it is adaptive in a very special kind of ecological niche that only humans inhabit? This intriguing possibility is well worth exploring. It is incompatible, however, with the standard approach to reason, which claims that reason enhances cognition whatever the environment it operates in and whatever the task it pursues. Understanding why only a few species have echolocation is easy. Understanding why only humans have reason is much more challenging.

Think of wheels. Animals don’t have wheels. Why not?1 After all, wheeled vehicles are much easier to construct than ones with legs or wings (just as models of reasoning seem much easier to develop than models of vision). However, artificial wheels are made separately and then added onto a vehicle, whereas biological wheels would have to grow in situ. How could a freely rotating body part either be linked to the rest of the body through nerves and blood vessels or else function without being so linked? Viable biological solutions are not easy to conceive, and that is only part of the problem.

For a complex biological adaptation to have evolved, there must have been a series of evolutionary steps, from rudimentary precursors to fully developed mechanisms, where every modification in the series has been favored (or at least not eliminated) by natural selection. The complex visual systems of insects, mollusks, or mammals, for instance, have all evolved from mere light-sensitive cells through long series of modifications, each of which was adaptive or neutral. Presumably, a similar series of adaptive steps from nonwheeled to wheeled animals was, if not impossible, at least so improbable that it never occurred.

Perhaps, then, reason is to animal cognition what wheels are to animal locomotion: an extremely improbable evolutionary outcome. Perhaps reason is so rare because it had to evolve through a series of highly improbable steps and it did so only once, only very recently in evolutionary time, and for the benefit of just one lucky species—us.

The series of steps through which reason would gradually have evolved remains a mystery. Reason seems to be hardly better integrated among the more ordinary cognitive capacities of humans than are the superpowers of Superman or Spider-Man among their otherwise ordinary human features. Of course, it could be argued that reason is a graft, an add-on, a cultural contraption—invented, some have suggested, in ancient Greece—rather than a biological adaptation. But how could a species without the superpower of reason have invented reason itself? While reason has obviously benefited from various cultural enhancements, the very ability of a species to produce, evaluate, and use reasons cries out for an evolutionary explanation. Alas, what we get by way of explanation is little more than hand waving.

The problem is even worse: the hand waving itself seems to point in a wrong direction. Imagine, by way of comparison, that, against the odds, biological wheels had evolved in one animal species. We would have no idea how this evolution had taken place. Still, if these wheels allowed the animals to move with remarkable efficiency in their natural environment, we would have a good idea why they had evolved; in other terms, we would understand their function. We might expect animal wheels, like all biological organs, to have weaknesses and to occasionally malfunction. What we would not expect, though, is to find some systematic flaw in this locomotion system that compromised the very performance of its function—for instance, a regular difference in size between wheels on opposite sides, making it hard for the animals to stay on course. A biological mechanism described as an ill-adapted adaptation is more likely to be a misdescribed mechanism. Reason as standardly described is such a case.

Psychologists claim to have shown that human reason is flawed. The idea that reason does its job quite poorly has become commonplace. Experiment after experiment has convinced psychologists and philosophers that people make egregious mistakes in reasoning. And it is not just that people reason poorly, it is that they are systematically biased. The wheels of reason are off balance.

Beyond this commonplace, polemics have flared. Reason is flawed, but how badly? How should success or failure in reasoning be assessed? What are the mechanisms responsible? In spite of their often bitter disagreements, parties to these polemics have failed to question a basic dogma. All have taken for granted that the job of reasoning is to help individuals achieve greater knowledge and make better decisions.

If you accept the dogma, then, yes, it is quite puzzling that reason should fall short of being impartial, objective, and logical. It is paradoxical that, quite commonly, reasoning should fail to bring people to agree and, even worse, that it should often exacerbate their differences. But why accept the dogma in the first place? Well, there is the weight of tradition … And, you might ask, what else could possibly be the function of reasoning?

Reason as standardly understood is doubly enigmatic. It is not an ordinary mental mechanism but a cognitive superpower that evolution—it used to be the gods—has bestowed only on us humans. As if this were not enigmatic enough, the superpower turns out to be flawed. It keeps leading people astray. Reason, a flawed superpower? Really?

Our goal is to resolve this double enigma. We will show how reason fits in individual minds, in social interactions, and in human evolution. To do so, we challenge the tradition, reject the dogma, and rethink both the mechanisms of reason and its function.

Where We Are Going

There have been more than two thousand years of philosophical work on reason, and more than fifty years of intense experimental work on reasoning. Some of the greatest thinkers of all time have contributed to this work. It would be beyond presumptuous to claim that most of this thinking has been on the wrong track, if it were not for the fact that both the philosophical and the psychological tradition have been vigorously contested from within.

How good is reason at guiding humans toward true knowledge and good decisions? How good are humans at using reason? We won’t attempt to tell the convoluted story of these old debates that in recent times, with psychologists joining the fray, have intensified to the point of being called “rationality wars.” What we will do instead in Part I of this book, “Shaking Dogma,” is single out clashes that reveal how serious are the problems posed by standard approaches to reason, and how wanting the solutions. We will suggest that parties to these heated debates have managed to weaken one another to the point that the best course may well be to collect from the battlefield whatever may still be of use and to seek new adventures on more promising ground.

We are less interested anyhow in debunking shaky ideas than in developing a new scientific understanding of reason, one that solves the double enigma. Reason, we will show, far from being a strange cognitive add-on, a superpower gifted to humans by some improbable evolutionary quirk, fits quite naturally among other human cognitive capacities and, despite apparent evidence to the contrary, is well adapted to its true function.

To understand how reason could have evolved and how it works, one should pay attention not only to what makes it special but also to how it fits among other psychological capacities and how much it has in common with them. There are many mechanisms involved in drawing inferences. Reason is only one of them. In Part II, “Understanding Inference,” we situate reason in relation to other inferential mechanisms, the overall picture being schematized in Figure 1.

Animals make inferences all the time: they use what they already know to draw conclusions about what they don’t know—for instance, to anticipate what may happen next, and to act accordingly. Do they do this by means of some general inferential ability? Definitely not. Rather, animals use many different inferential mechanisms, each dealing with a distinct type of problem: What to eat? Whom to mate with? When to attack? When to flee? And so on.

image
Figure 1. How reason is embedded in several categories of inference.

Humans are like other animals: instead of one general inferential ability, they use a wide variety of specialized mechanisms. In humans, however, many of these mechanisms are not “instincts” but are acquired through interaction with other people during the child’s development. Still, most of these acquired mechanisms have an instinctual basis: speaking Wolof, or English, or Tagalog, for instance, is not instinctive, but paying special attention to the sounds of speech and going through the steps necessary to acquire the language of one’s community has an instinctual basis.

As far as one can tell, other animals perform all their inferences without being conscious of doing so. Humans also perform a great variety of inferences automatically and unconsciously; for instance, in acquiring their mother tongue. However, there are many inferences of which humans are partly conscious. We are talking here about intuitions. When you have an intuition—for example, the intuition that your friend Molly is upset even though she didn’t say so and might even deny it—this intuition pops up fully formed in your consciousness; at the same time, however, you recognize it as something that came from within, as a conclusion somehow drawn inside your mind. Intuitions are like mental icebergs: we may only see the tip but we know that, below the surface, there is much more to them, which we don’t see.

Much recent thinking about thinking (for instance Daniel Kahneman’s famous Thinking, Fast and Slow)2 revolves around a contrast between intuition and reasoning as if these were two quite different forms of inference. We will maintain, on the contrary, that reasoning is itself a kind of intuitive inference.

Actually, between intuition in general and reasoning in particular, there is an intermediate category. We humans are capable of representing not only things and events in our environment but also our very representations of these things and events. We have intuitions about what other people think and about abstract ideas. These intuitions about representations play a major role in our ability to understand one another, to communicate, and to share opinions and values. Reason, we will argue, is a mechanism for intuitive inferences about one kind of representations, namely, reasons.

In Part III, “Rethinking Reason,” we depart in important ways from dominant approaches; we reject the standard way of contrasting reason with intuition. We treat the study of reason (in the sense of a mental faculty) and that of reasons (in the sense of justifications) as one and the same thing whereas, in both philosophy and psychology, they have been approached as two quite distinct topics.

Whereas reason is commonly viewed as a superior means to think better on one’s own, we argue that it is mainly used in our interactions with others. We produce reasons in order to justify our thoughts and actions to others and to produce arguments to convince others to think and act as we suggest. We also use reason to evaluate not so much our own thought as the reasons others produce to justify themselves or to convince us.

Whereas reason is commonly viewed as the use of logic, or at least some system of rules to expand and improve our knowledge and our decisions, we argue that reason is much more opportunistic and eclectic and is not bound to formal norms. The main role of logic in reasoning, we suggest, may well be a rhetorical one: logic helps simplify and schematize intuitive arguments, highlighting and often exaggerating their force.

So, why did reason evolve? What does it provide, over and above what is provided by more ordinary forms of inference, that could have been of special value to humans and to humans alone? To answer, we adopt a much broader perspective.

Reason, we argue, has two main functions: that of producing reasons for justifying oneself, and that of producing arguments to convince others. These two functions rely on the same kinds of reasons and are closely related.

Why bother to explain and justify oneself? Humans differ from other animals not only in their hyperdeveloped cognitive capacities but also, and crucially, in how and how much they cooperate. They cooperate not only with kin but also with strangers; not only in here-and-now ventures but also in the pursuit of long-term goals; not only in a small repertoire of species-typical forms of joint action but also in jointly setting up new forms of cooperation. Such cooperation poses unique problems of coordination and trust.

A first function of reason is to provide tools for the kind of rich and versatile coordination that human cooperation requires. By giving reasons in order to explain and justify themselves, people indicate what motivates and, in their eyes, justifies their ideas and their actions. In so doing, they let others know what to expect of them and implicitly indicate what they expect of others. Evaluating the reasons of others is uniquely relevant in deciding whom to trust and how to achieve coordination.

Humans also differ from other animals in the wealth and breadth of information they share with one another and in the degree to which they rely on this communication. To become competent adults, we each had to learn a lot from others. Our skills and our general knowledge owe less to individual experience than to social transmission. In most of our daily undertakings, in family life, in work, in love, or in leisure, we rely extensively on what we have learned from others. These huge, indispensable benefits we get from communication go together with a commensurate vulnerability to misinformation. When we listen to others, what we want is honest information. When we speak to others, it is often in our interest to mislead them, not necessarily through straightforward lies but by at least distorting, omitting, or exaggerating information so as to better influence them in their opinions and in their actions.

When we listen to others, then, we should trust wisely and sometimes distrust. When we talk to others, we often have to overcome their understandable lack of trust. If we distrusted others only when they don’t deserve our trust, things would be for the best. Often, however, we withhold our trust out of prudence, not because we know that others are untrustworthy but because we are not sure that we can trust them. This reticence may be wise—better safe than sorry—but still, we miss valuable information. Communication, which could be beneficial to speakers and listeners alike, often falters for lack of confidence.

The second function of reason—a function carried out through reasoning and argumentation—is, we claim, to make communication effective even when the communicators lack sufficient credibility in the eyes of their audience to be believed on trust. Reason produces reasons that communicators use as arguments to persuade a reticent audience. Reason, by the same token, helps a cautious audience evaluate these reasons, accept good arguments, and reject bad ones.

Much of our earlier joint work focused on this argumentative function of reason and developed an “argumentative theory of reasoning.”3 In this book, we broaden our perspective, consider both the argumentative and the justificatory functions of reason, and develop an interactionist approach to the mechanisms and the two functions of reason.

Part IV, “What Reason Can and Cannot Do,” offers a tour of what reason does. Throughout this tour, we show how our interactionist perspective is in a good position to explain why reason behaves the way it does. We revisit some well-established but ill-explained apparent weaknesses of reason such as the confirmation bias. We also draw attention to some of its neglected strengths.

The tour starts with a pair of observations: human reason is both biased and lazy. Biased because it overwhelmingly finds justifications and arguments that support the reasoner’s point of view, lazy because reason makes little effort to assess the quality of the justifications and arguments it produces. Imagine, for instance, a reasoner who happens to be partial to holidays at the beach. When reasoning about where to spend her next vacation, she will spontaneously accumulate reasons to choose a sunny place by the sea, including reasons that are manifestly poor (say, that there’s a discount on the flight to the very place where she would like to go, when in fact the same discount applies to many other destinations as well).

The solitary use of reason has two typical outcomes. When the reasoner starts with a strong opinion, the reasons that come to her mind tend all to support this opinion. She is unlikely, then, to change her mind; she might even become overconfident and develop stronger opinions. But sometimes a reasoner starts with no strong opinion, or with conflicting views. In this case, reason will drive her toward whatever choice happens to be easier to justify, and this sometimes won’t be the best choice. Imagine she has a choice between visiting her horrible in-laws and then vacationing at the beach, or starting with the beach and then going to see the in-laws, the latter option being somewhat cheaper. Reason will drive her toward what seems to be the rational decision: taking the cheaper option. It is likely, however, that she would come back more satisfied if she started with the in-laws instead of letting the prospect of this visit spoil her time at the beach: a better choice overall, but involving a hard-to-justify extra expense.

Psychologists generally recognize that reason is biased and lazy, that it often fails to correct mistaken intuitions, and that it sometimes makes things worse. Yet most of them also maintain that the main function of reason is to enhance individual cognition—a task it performs abysmally. The interactionist perspective, on the other hand, offers for the first time an evolutionarily plausible account of the often decried biases and shortcomings of reason.

It makes sense, we will show, for a cognitive mechanism aimed at justifying oneself and convincing others to be biased and lazy. The failures of the solitary reasoner follow from the use of reason in an “abnormal” context. Underwater, you wouldn’t expect a pen—which wasn’t designed to work there—or human lungs—which didn’t evolve to work there either—to function properly. Similarly, take reason out of the interactive context in which it evolved, and nothing guarantees that it will yield adaptive results.

What, then, happens when reason is put back in its “normal” environment, when it gets to work in the midst of a discussion, as people exchange arguments and justifications with each other? In such a context it properly fulfills the functions for which it evolved. In particular, when people who disagree but have a common interest in finding the truth or the solution to a problem exchange arguments with each other, the best idea tends to win; whoever had it from the start or came to it in the course of the discussion is likely to convince the others. This conclusion might sound unduly optimistic, but it is supported by a wide range of evidence, from students discussing logical problems, to juries deliberating, and to forecasters trying to predict where the next war will erupt.

In the last three chapters (Part V, “Reason in the Wild”) we demonstrate how robust are the features and effects of reason reviewed earlier. We find that solitary reasoning is biased and lazy, whereas argumentation is efficient not only in our overly argumentative Western societies but in all types of cultures, not only in educated adults but also in young children. Few will be surprised to hear that reason is typically biased and lazy when it is applied to moral and political issues. More surprising may be evidence that shows how, even in the moral and political realms, argumentation may work quite efficiently, allowing participants to form more accurate moral judgments and citizens to form more enlightened opinions. Such findings, however, are what one should expect in an interactionist perspective.

The last chapter (Chapter 18, “Solitary Geniuses?”) is about science, generally considered the pinnacle of human reason. Science is exceptional in many ways, but is the way scientists reason itself exceptional? Scientific progress is often attributed to solitary geniuses, from Newton to Darwin or Einstein. Their superior reason, we are told, doesn’t suffer from the shortcomings that plague the rest of us. Not only can these geniuses dispense with discussions with others in order to come up with new theories, such discussion might even hinder them when their revolutionary insights would be misunderstood and scorned by their not-quite-peers. Better wait for a less prejudiced new generation to see the light. Fortunately (for our theory and for scientists), science doesn’t work this way. Scientists make do with the same reason that all humans use, with its biases and limitations. But they also benefit from its strengths and in particular from the fact that reason is more efficient in evaluating good arguments than in producing them: when the arguments are there, the scientific community is able to elevate the status of a new theory from fringe to textbook material in a few years.

In these five parts and eighteen chapters, what we will put to you, then, is an interactionist approach to reason that contrasts with standard intellectualist approaches: reason, we maintain, is first and foremost a social competence. We do not deny that reason can bring huge intellectual benefits, as the case of science well illustrates; on the contrary, we explain how it does this: through interaction with others.

You are unlikely to accept what we say just because we say it, so we will present you with arguments that you will be able to assess on their own merits. We will show you how considering reason as a mechanism that draws intuitive inferences about reasons solves the first half of the enigma: reason is not a superpower implausibly grafted onto an animal mind; it is, rather, a well-integrated component of the extraordinarily developed mind that characterizes the human animal.

To resolve the second half of the enigma, we will demonstrate how apparent biases that have been described as deplorable flaws of reason are actually features well adapted to its argumentative function. A number of sometimes surprising predictions about human reason follow from our approach. The evidence we will present confirms these predictions. It is by force of argument that we hope to persuade you that the interactionist approach is right or, at least, on the right track. This, of course, makes the book itself an illustration of the perspective it defends.

I


SHAKING DOGMA

Reason, the faculty that gives humans superior knowledge and wisdom? This dominant view in the Western tradition has been radically undermined by fifty years of experimental research on reasoning. In Chapters 1 and 2, we show how old dogmas were shaken, but not nearly enough. The now dominant view of reasoning (“dual process” or “fast and slow thinking”), however appealing, is but a makeshift construction amid the ruins of old ideas.