THE FUNCTION OF REASON
The general question I've been living with is how do we go about getting a better scientific grip on everything social? The social sciences have developed away from the natural sciences, even with some bit of hostility toward natural sciences, and that, I believe, is a source of poverty. If we want to have a more ambitious understanding of how social life functions, of the mechanisms involved, the challenge is to achieve continuity with neighboring natural sciences. The obvious neighbors to begin with are cognitive neuroscience, ecology, biology, and others.
I started as a social scientist. I started as an anthropologist doing fieldwork in a small group of people in the south of Ethiopia, asking myself fairly standard anthropological questions.
I was in this tribe in the south of Ethiopia, studying rituals—sacrifices and divinations. They had a fairly rich ritual life with lots of symbols and so on, and I would keep asking them, “What is the meaning of the symbols you’re using? What are the reasons for why you do this ritual the way you do?” And I never got a satisfactory answer, or so I thought. When asked about the meaning they said, “We do it because that’s what our fathers did, and our forefathers.” That was always the answer: “We do it because that’s the way we’ve always done it.” I was very frustrated by this and went looking for possibly better informants—an older member of the society, a “wise man,” or whatever—who would know more, but I never found them.
One morning I woke up after having dreamt about all that fairly intensely, and in the dream I was telling myself, “You’re not paying attention. Listen to what they’re saying. Maybe what they’re saying is exactly right. Maybe these symbols don’t have meaning. Maybe their job is not to convey meaning. Maybe the reason why the people do these things is because of the force of precedent, because indeed they’ve done it all the time.”
I’d been there quite a few months, but I was so agitated by this that I flew back to Paris and started working on a book, which came out some years later, in 1975, called Rethinking Symbolism. In it, I argued that cultural symbolism is not in the job of conveying meanings. If you want to convey meanings, there are much better means to do that—as we're doing now, for instance, speaking. There's a high investment that’s involved in ritual symbols, the interpretation of which is always uncertain or vague. There are people in some societies who tell you “this means that,” but their answers themselves are mysterious and call for further interpretation. That can’t be the reason why they do it. That can’t be the function of these symbols and rituals.
My argument was that cultural symbolism has more of a cognitive function. What these cultural symbols do achieve is focus attention in certain directions. Rather than "mean" something, they evoke many things. They create a certain commonality of orientation, interests, and values among people without having any signification, properly speaking. That got me into cognitive science, which was very much at the beginning. The official beginning was in the late ‘50s, early ‘60s, but at that time there was little cognitive science, especially compared to what came out after, and practically none of it was about higher cognitive processes. Issues concerning the meaning of symbols were nowhere near the questions that people doing cognitive psychology at the time had in mind.
I got involved in cognitive science fairly intensely. I’ve been involved also in the study of human communication, particularly language. Part of the reason is that I developed my ideas in France, where Levi-Strauss was such a big influence. I was never a Levi-Straussian, but he was still the most interesting anthropologist around. He was always insisting that language, linguistics, provided the model for the study of culture, for the study of social science. It was the heyday of semiology or semiotics, the notion that a conceptual framework would unify all these human sciences and possibly even go beyond that.
I studied linguistics quite a bit, and I came across Chomsky when I was a student at Oxford in the mid-‘60s. For me, that broke down the Levi-Straussian view because, on the one hand, Chomsky’s work on language was much more impressive than Saussure or even Jakobson—the classical structuralist work. On the other hand, the kind of model that Chomsky was proposing, generative grammar or transformational grammar as it was called at the time, was quite specific to language. The notion that you find in Saussure that the model of language can be exported to talk about culture, about music, about art, about anything, didn’t make sense anymore in the case of language because the Chomskyan approach was clearly focused on the peculiarities of human language.
The whole idea that if you wanted to have a unified understanding of human communication, human culture, human language, all you needed was a framework provided by structural linguistics, that was out. Then either you could decide to become a student of syntax, or, say, do ethnography in South Ethiopia, or else, if you had the wide ambition that I had, then you had to go back to the drawing board and rethink very basic issues.
Also, I got involved in linguistics. The doubts I had about using a simple semiotic model of cultural symbolism—which, in my work, extended to a study of metaphors and symbolism in language—caused me to interact with people in linguistics and philosophy of language who were interested in forms of comprehension that went beyond semantics, beyond just getting the meaning of words as they may be described in a formal system, to understand how words are being used in a given context. This field developed under the name of pragmatics, and was very much influenced by the English philosopher Paul Grice.
I worked on this new approach to not just linguistic communication, but communication in general with an English friend and colleague Deirdre Wilson, who was a linguist who had studied at MIT with Chomsky and others. We developed a new approach to pragmatics, which was squarely grounded in cognitive science, in cognitive psychology.
A major challenge for human cognition is this: Humans have the ability to process a very wide range of information through their senses and through the conceptual framework they can bring to bear on monitoring their environment. Plus, they get all this information from communication with others. Plus, they have all this information in memory. Now what you have is a glut. You have too much information. This happens way before the Internet and too much soliciting of your attention. That’s true for prehistorical man in a traditional environment. We have a capacity to monitor many more things than we can process in an intensive manner, than we can attend to.
A crucial issue for cognitive efficiency for humans then is to decide which of all the information that is competing for your attentional resources, both from the environment and from your memory, should be prioritized. Which background information should be brought to bear on which new information in order to get the most efficient processing for information?
We developed what we called relevance theory, arguing that human cognition is geared towards the maximizing of relevance of the input that it processes. This, we argued, has a consequence for communication. When we communicate intentionally—I’m talking to you, requesting your attention—your attention spontaneously goes to what’s relevant to you, more relevant than the competition at this moment. When I try to get your attention through communication, I’m conveying that I assume that what I’m trying to communicate is worth your attention and is more relevant than anything else you could attend to at this very minute. And this, we argued, determines how you interpret in context what is meant by the words that are being used. So the right interpretation is the interpretation that will tend to confirm this expectation of relevance that every utterance raises about itself.
On this basis, we developed a view that it’s not that your linguistic utterances have a literal meaning, and that when you use them you use them to convey this literal meaning and then you can depart it for a rhetorical purpose with metaphor, irony, or implicit content from this literal meaning. No. Quite generally, whether you speak literally, or metaphorically, or ironically, or whatever, your words are not an encoding of your meaning; they are a piece of evidence from which your meaning has to be inferred. That meaning and these inferences are guided by this expectation of relevance, as I was mentioning before. That meaning can go from very specific—for instance, you ask me what the time is and I look at my watch tell you it’s 6:15—to very vague meanings. And this can also be expressed by behavior, by gesture, and indeed by cultural symbols, where you convey that relevance will be achieved by orienting in a certain direction, by looking at certain things rather than others, by approaching them with a certain kind of expectation. There’s a continuum of cases between precise meanings that you can paraphrase and much vaguer effects, which boil down to a mix of focalization and evocation.
Anthropological fieldwork is a nice job, a nice métier; I liked doing it. I like the company of fellow anthropologists; they are people who have an extraordinary curiosity and are willing to talk about lots of subjects. They are willing to spend hours listening to people who work in some small group in the Amazon or in Polynesia, who have studied some weird local practices. I like this kind of curiosity.
On the other hand, anthropologists each specialize in their own fieldwork. They’ve invested years in something that they cannot properly share. I might talk for two hours about my fieldwork; I spent years there. In a way, it’s quite solitary work. When you’re in the field you’re with people all the time, but it’s solitary in terms of sharing what you’ve learned.
I was working in a small group of farmers and weavers in the south of Ethiopia—typical anthropological fieldwork up in the mountains. Anthropologists nowadays work in all kinds of society, but the traditional fieldwork of anthropologists was in a small group with a very traditional culture, often with very simple technology, often without writing. These are all very interesting groups to study, and they all have a share of human experience that is rapidly vanishing. If only because of that it’s worth trying to document and to understand.
I was, however, more interested in theoretical issues. Most anthropologists have very little interest or even patience for theory. I was also more personally attracted to cooperative work—discussing with others, doing joint work. The work I have been talking about, on linguistic communication with Deirdre Wilson, involved working with somebody else. We discussed endlessly and that was great. You don’t get that when you do anthropological fieldwork. Then I got involved in experimental and cognitive psychology. Again, doing experimental work has a great quality in that you work with collaborators, you jointly do experiments, you get results as evidence, which may go in favor or against the kind of hypothesis you had.
A number of issues regarding what’s common to all systems of all communication had occupied many people in the early and mid-20th century, from the linguistics of Saussure to the cybernetics of Wiener. All this was before the cognitive revolution. If any psychology was involved at all it was shallow. We, on the other hand, came after the cognitive revolution. We could take advantage of this much richer understanding of human psychology and of the mechanisms involved.
Part of the origin of cognitive psychology, of course, was the same as programs for the development of intelligent machines, the discovery of the Turing-Church Thesis, and the idea that you could have precise mechanisms and machines that processed information. This led to a much richer way of asking questions that had been asked before. We could start thinking about communication, linguistic communication, or cultural communication, or rituals, (if rituals are in the business of communication, which they are only to a certain degree) within a richer framework and asking more precise questions about the mental cognitive mechanisms that are at work. That was at the scientific basis of our work.
I worked on and off in Ethiopia between ’69 and ’75. I did a regular stint of a field anthropologist. I spent a bit less than two years in the field, in the tribe altogether. But then I just got completely involved in more theoretical issues. I stayed with one foot in the anthropological community, but I was really interested in the more dynamic discussions about language, communication, the naturalistic approach, the evolutionary approach to culture and to cognition. All that was, to me, so much more intellectually exciting.
For a long time I did follow very closely work in linguistics. I was very influenced by Chomsky. I’ve been, however, doing so many other things that I stopped following central issues in linguistics a long time ago. So, if you ask me today whether I agree or disagree with Chomsky’s current view of syntax, I don’t even have the competence to answer that, but I’m very impressed by what he has done. Of course, he’s not an anthropologist and he didn’t do fieldwork. He did study more than one language, but that was not his point. He’s been incredibly important. He’s changed the field completely. Even people who are extremely hostile to him in the field have been at least indirectly massively influenced by his work.
Part of the intellectual excitement of Chomsky is that he was asking pretty fundamental questions. He related issues about certain constructions in the syntax of English to issues of what made human beings capable of acquiring language. He made the technical issues in linguistics relevant to general theoretical issues, and general theoretical issues relevant to the study of particular cases. It was intellectually extremely stimulating. For me, the most important intellectual encounter has been the one with Chomsky.
Initially, I started with this interest in society and culture, in collective things. The classical view of what culture is, very simply, that which is transmitted in a population by non-genetic means: by communication, imitation, and all forms of interaction. In the human case, imitation is an important factor which has been overplayed. Humans imitate better than any other animal, (except maybe parrots, but parrots have a narrow range of things that they imitate).
We humans are good imitators, but, more importantly, we’re great communicators. We transmit much more via communication than we do by imitation. Communication is the vector through which culture develops, is transmitted, builds, and evolves, more than anything else. The reason why I studied communication with Deirdre Wilson and did all this work on relevance theory is because I saw communication as a building block, as a crucial ingredient for understanding society and culture—which was also the idea of Levi-Strauss and others. But they thought they understood what communication was; communication was what Saussure’s structuralist model said it was. I thought that was wrong. We really had to rethink communication quite radically. But my goal in doing that was to understand society and culture, not to understand language per se, (though I'm interested in that, too).
How do you move, first, from individual cognition to the interaction between typically two individuals who might be involved in communication? And then how do you scale up to what happens at the scale of populations, of human groups? In those days the social sciences were completely divorced from the cognitive sciences (which were not even called cognitive sciences). It is still true to a large extent, but much less than it was then. I thought that a bridge could be built between the social sciences and the emerging cognitive sciences. This would give us greater insights and greater tools for understanding the social, and to establish this continuity between the natural and the social sciences, which I thought was essential to improving the social sciences themselves and better understanding the world in which we live. In this work, a better understanding of communication played a fairly central role.
How you move from communication in the ordinary sense to cultural transmission? This is a challenging question. And, on this, my mind was going at the time—we’re talking about the ‘70s—in the same direction in which Richard Dawkins was going when he started talking about memetics. One day an English friend of mine brought me an issue of the New Scientist where there was a long essay by Richard Dawkins, which was, in fact, the last chapter of The Selfish Gene. The book hadn’t been published yet, so he was selling memetics before selling The Selfish Gene. I found it illuminating in many ways. I’d been arguing, in a much more modest and vague manner, for similar ideas. Nevertheless, I had some reservations about Dawkins’s approach.
What I found really exciting there was an idea I was also arguing for at the time. Namely, to explain the success of bits of culture, of practices, of rituals, of techniques, of ideologies, and so on, the question was not how do they benefit the population in which they evolve; the question was how do they benefit their own propagation? Dawkins was saying that much better than I could have done at the time. This was exactly right, I thought. You don’t need to explain the success of social, cultural practices by assuming that they owe this success to the benefit that they bring to the population in which they evolve. It’s only marginally that cultural practices benefit themselves by benefitting the population in which they evolve. Helping their carriers is one way in which bits of culture can benefit themselves. But there are lots of other ways.
It was because I was involved in a fairly detailed study of how human communication works that I was struck by the fact that communication is not a replication system. When I communicate to you, you don’t get in your mind a copy of my meaning. You’ll transform it into something else. You extract from it what’s relevant to you. It involves both understanding and misunderstanding. But even if you’re understanding me perfectly, your goal will not be to have a copy of what was in my mind, it will be to extract from it some thoughts of yours which will have been usefully informed by mine, but which will be relevant to you.
In Dawkins’s memetics, replication was a crucial element. The idea was that you could generalize the Darwinian model of selection to all kinds of replicators. Memes were cultural replicators competing with one another for space in our minds and in our social interactions, and therefore, the object of process of selection. What seemed wrong to me was the idea that information in human transmission replicates. It doesn’t. You get this paradox of evolutionary approaches to culture, which takes its extreme form in Dawkins’s memetics. Dawkins has a kind of clarity of extremist views; I admire that.
Dawkins’s memetics is such a simple and clear idea, so what I think to be a problem with it is also more apparent. The same problem arises with most evolutionary approaches to culture. The problem, or the paradox, is that if you look at cultures, what you see is quite a bit of stability: The same words are being used more or less in the same sense for generations; the evolution of word meanings or word phonology is very slow; the same tales are being told to children one generation after the other; the same recipes are being cooked; the same laws are being followed, interpreted, and employed. So many aspects of culture seem to involve repetition again and again. How can things stay so stable? It has to be that they are being reproduced quite faithfully. You need high fidelity replication or reproduction to explain the stability of culture. Or so it seems.
Suppose that instead of looking at cultural phenomena generally you look at the micro mechanism of transmission: communication, imitation, and so on. What happens when you teach something to somebody? What you see is that, yes, humans are good at imitation. They’re good at communication. They’re better than any other animal species we know, but this however hardly ever results in replication.
When you communicate orally, people don’t copy in their mind what you have told them, they extract something from it. If you see a friend who has a great recipe for apple pie and you “imitate it,” you don’t really copy it. You look at it and you extract from it a way to do it your own way. There's a loss of information at every step, which is quite significant. But even to talk about a “loss of information” assumes that the goal was to replicate. Once you understand that the goal is to extract something that’s useful to the learner, to the imitator, to the addressee of communication, then it’s not loss of information; it’s a constructive use. You construct with what others provide you something that you want. And so, in fact, you rarely replicate.
So how can you have this macro stability of cultural things with this micro failure to replicate? There's got to be fidelity in copying, hasn't there? You look closely and, no! I only know two very clear cases of people who would copy faithfully. One is forging money, where the forger tries to copy the dollar bill exactly. And the other is a chorus line on 42nd Street. The rest of human interaction involves a lot of coordination, but very little copying, strictly speaking.
At that point there’s got to be something wrong with the idea, which is still very widespread, that what makes culture possible is high fidelity copying. And again, this is an idea that is at the center, in particular, of Dawkins’s memetics. I’ve been arguing for a long time not just what I think to be plain observation that, in fact, high fidelity is not common at all—and a lot of things are culturally transmitted without being copied in a faithful manner—but also to give a positive account of what’s happening.
Fidelity is not the only way to ensure stability. You can have stability in a population not just by faithfully copying, but also if the transformations that everybody produces at each step—again, each person is looking for what’s relevant to them—if these transformations converge, if you have what I can call a “cultural attractor” … let me give you an example. Think of the word “love.” Love is a very successful word in the English language used every day millions of times, billions of times. Each time the meaning is a bit different. The lover says, “I love you.” What does she mean exactly? Does she mean what you mean by it? Does she mean the same thing she meant yesterday? There are a whole variety of uses. You can copy the sound of the word “love,” but you cannot copy the meaning. Meaning, anyhow, is not something that you could observe and then copy. All you can do is infer—not observe—what the person means when she uses the word. If, however, our transformations converge towards attractors, towards ways of thinking that are of relevance to all of us, then you may get stability without fidelity. You get it because of a convergence of transformations rather than because of the absence of transformation (which is what fidelity would be). You can also model mathematically such converging transformations and their cultural effects.
I’m asking a big question, which is then divided into many sub-questions. The big question is, again, how do we get a naturalistic understanding of culture in society? For this we have to understand the micro mechanisms of communication. And for this we have to understand some basic aspects of cognition. And for this we have to understand something about the evolution of cognition in humans. The evolutionary psychology program has an important role in all that.
Then we have to put all these things together (including demographic and ecological factors) and see to what extent we can understand, and possibly even model, population-scale dynamics where these mechanisms interact and help explain how the cultural items transform, emerge, and vanish. That’s a large part of the program in which I’ve been involved.
I first mentioned that I like doing experiments because I like the cooperative work that’s involved in this work. It’s a fun activity. It’s intellectually very stimulating. And also it does sometimes provide important evidence.
I started doing experiments a long time ago. When I was still mostly an anthropologist, I’d been invited by Clifford Geertz at the Institute for Advanced Study in Princeton. He hoped he would correct my mistaken ideas and turn me into a proper anthropologist in his style, but I was too stubborn and too attracted by more naturalistic approaches. But I enjoyed the time spent there.
I met George Miller at Princeton University, and I found him a wonderful person. He asked me the same things you’re asking me—what was I working on, what questions I had in mind. That day I had been working with Deirdre Wilson on irony, and he asked me, “Are the ideas you have experimentally testable?” I started thinking, and I went back the next day and said, “We could do these experiments which, if our account is right, should produce these results, and if a more classical view of irony is right, should produce these other results.” We did the experiments and published them, and then this started a new cottage industry of experimental study of irony. This was my first experimental work in psychology. Starting with George Miller was not a bad start. It was fairly inspiring, so I continued. The experiments I did later were on reasoning, which I also found to be an exciting topic.
Few, if any, scientific ideas come bottom up. It’s never the case for me, and rarely the case for anybody, that you gather so much evidence and data that somehow an idea emerges. There’s a bottom up aspect, but more important is the top down, where you have hunches. You may call it intuition. I think it’s mostly luck, when you hit on a good idea. Other people, just as bright and smart as you, have the bad luck of hitting on a bad idea. They invest a lot on the bad idea and they don’t get anywhere. If you have been lucky enough to hit on a good idea then, indeed, you’ll find confirming evidence, good evidence that will start explaining lots of things. But initially I think we’re groping in the dark.
As I said, I started doing experimental work on relevance, on communication, irony, and so on. Then I became a mentor, for the Fyssen Foundation, of an Italian psychologist, Vittorio Girotto, who’d been working on reasoning and had been awarded a grant. I would argue with him that a lot of findings in psychology or reasoning had to do with relevance. People seemed to be making logical mistakes in reasoning, but what they were really doing was transforming the input they were presented with in a way that would make it more relevant to them. Vittorio and I did a lot of experiments going in that direction.
This is how I got involved in experimental work on reasoning. Besides the fun of doing this work, part of the reason why it mattered to me was because if you want to explain social interaction and cultural contents, reason and reasoning play a very important role. I was talking before about converging transformations in cultural transmission. One way we converge is by reasoning together or by exchanging reasons and coming to see things in the same way. It doesn’t always work, but it’s still a strong factor of convergence.
I'd been thinking that standard approaches to reason and reasoning were mistaken because they were seeing it as an individual adaptation, as a way to enhance your own individual cognition. That didn’t make much sense in these exchanges of reasons. Using your reason to produce an argument to convince others and establish a convergence of goals or ideas made more sense. True, sometimes you can convince others just because you have authority and they trust you. But when they don’t, or they don’t on the topics on which you would like to convince them, does it mean that there’s nothing you can do? No. You can still overcome the limits of trust by producing arguments that they can evaluate on their own merits, and if they find that these arguments are good enough, then they will possibly be convinced by what you’re saying.
Hugo Mercier came to work with me as a student. He decided he liked the idea of the social function of reasoning and proposed to do his PhD by developing this theme, which he did splendidly by going way beyond what I had envisaged, both in deepening the ideas and gathering so much good evidence for it. We were producing ideas and papers, which got picked up in the New York Times through the Edge meeting that you were alluding to, but also with a big misunderstanding.
The success of the “argumentative theory of reasoning” was in good part based on a misunderstanding. The misunderstanding being that we were taken to be saying, “Haha, you think reason is to get more clever, more intelligent, or discover the truth, but that’s not what it's for; it’s to persuade others who wouldn’t be persuaded otherwise. It’s a way to manipulate other people.” It was taken to be a cynical view of what reason and reasoning is about; it was taken to imply that people are naïve if they think reason is in the pursuit of truth. People who take this cynical view do not apply it to themselves and think that they reason objectively.
This cynical view doesn’t make any evolutionary sense. Why would something evolve to manipulate others and then nothing evolve in others so as not to be manipulated? It doesn’t work like that. If one can benefit by causing harm to others, then in others some counter measure is likely evolve, and you’ll have an arms race. So our argument was that reasoning evolved to produce arguments in order to convince others. This works, however, because reasoning also evolved to produce in each one of us a means to evaluate arguments so as to gain from the ideas of others when they're able to present good reasons for why we should accept them and to reject them otherwise. From an evolutionary point of view, reasoning has got to be beneficial on both sides otherwise people will stop listening to arguments and then producing them would be useless.
Our work, then, had a success partly based on misunderstanding. In general my work has had some impact, but it was very much within the scholarly community. This was the first time I had an article on my work in the general press, in the New York Times, the Guardian, the Corriere della Sera. The argumentative theory, which Jon Haidt liked and which had both been successful and misunderstood, was only part of our overall project. The argumentative theory develops an answer to the question, what is the function of reasoning? But it says nothing about the mechanisms. Hugo and I had made some allusions about mechanisms. Now we started working on this second aspect and developing it more thoroughly. What was initially a theory of reasoning and argumentation has now become a theory of reason, not just reasoning, and not just of its function, but also of its mechanism.
There’s something very weird about standard approaches to reason. On the one hand, from Aristotle onwards, reason is seen as what makes humans superior to all other animals, and this has been repeated ad nauseam. Before Darwin, humans were not only disposed to think that they were superior to all animals, but the more differences you could show between humans and other animals, the better. We have this capacity to reason that other animals don’t have, and that sets us completely apart from them? Very good!
After Darwin, the animality of humans became quite evident and apparently radical discontinuities became very puzzling. Do humans really have this kind of superpower, which doesn’t seem to fit anywhere in our natural traits, not even among our other cognitive capacities—from perception to unconscious inference, motor control—the kind of things we share with so many other animals. It’s a bit like we’re Superman or Spiderman with this fantastic capacity that only we have. This cannot be right.
Dawkins once had a nice article about why animals don't have wheels. You might think wheels would be a nice adaptation, so why didn’t it evolve? Well, it probably wouldn’t be such a good adaptation because it would be useful on only very specific terrains. On most terrains it wouldn’t help you. But even so, there might be animals living on a terrain where wheels would be very handy, so why didn’t they evolve? It’s not that it’s inconceivable; it’s that the design problems are very specific, and there are no in-between steps in the evolution of wheels that would each be adaptive. So how do you go from a non-wheeled animal to a wheeled animal? It can only be through a series of improbable steps, so improbable that it never happened on Earth.
In the case of reason being seen as this improbable superpower that exists only in one species—humans—you have moreover an extra problem, because reason, which is described as a way to enhance cognition in all domains, might be useful not just to humans, but to many other species. Investing massively in cognition the way humans have could be advantageous to other species. That’s one enigma proposed by this view of reason.
The second enigma is well known to psychologists. Kahneman and Tversky, Peter Wason, and others have described reason as being flawed, as making egregious mistakes all the time. On the one hand it's a superpower, but it’s a superpower that doesn’t work properly; this makes even less sense. You have this double paradox of having a superpower that doesn’t fit in an evolutionary perspective in any clear way and that, moreover, doesn’t even deliver what it’s supposed to deliver.
Hugo and I set out to resolve the double enigma; first, by showing that human reason fits perfectly well among other cognitive capacities. Reasoning is only one form of inference among others. Inference, the capacity to use some input information to derive further consequences that are not given but that you can draw on the basis of, is something that all animals do. They guide their action on that basis. Cognition, in general, is inferential. Perception is inferential. The way we use the activation of our retina to infer properties of the objects that have caused this activation by reflecting light is inferential. The way we guide our body movements is inferential. We draw inference all the time. Insects, slugs, birds, any organism that locomotes, that moves around, couldn’t do that without doing inferences. Plants stay put, so they don’t need cognition. They stay in the same place. They don't take the risk of moving. Moving provides new opportunities, but also huge risks. To benefit from the opportunity and avoid the risks, you need cognition and you need to infer precisely what’s beyond your skin’s surface. Inference is ubiquitous in animal life.
In which way is reason different from other forms of inference? In the literature you get some people who don’t even see the difference and assume that animals reason too, and the capacity to do logical inference must be present in lowly animals as it is in us. Then you get others who assume animal inference and human reason are completely separate capacities. We disagree with both approaches. We also disagree with an approach that is dominant today, the dual system approach, defended in particular by Daniel Kahneman and others like Stanovich or Evans, and according to which the enigma of reason can be solved by assuming that two kinds of processes are involved in inference. One kind, we share with other animals, and the other is more specifically human.
According to the dual system approach, first, there is a more basic System 1, which in humans is just what we might call intuition. It's an automatic system that operates spontaneously. It uses heuristics that generally work but are not fully warranted from a logical or epistemological point of view. Other animals use similar heuristics. We rely on them most of the time because doing so takes less time, less energy, less investment. It's a good way of approaching everyday tasks of inference, which go from not banging into furniture when you walk around to knowing how to talk to one another. When this doesn’t fully work or when we meet a problem that we cannot solve in this way, then we resort to reasoning, or System 2, which involves applying rules, proceeding in a more self-conscious manner, in a way that is more linked to a proper justification of an epistemological or logical nature.
If you assume that there is such a partition between two systems of inference—intuition and reasoning—then you can explain the apparent flaws of reason, the fact that in so many experiments people make these egregious mistakes, by assuming that they are guided by System 1, by the more intuitive kind of inference. Intuition is not geared toward handling atypical problems. It can be tricked in so many different ways. And so all the failures that you find in the literature of reasoning, all the cognitive illusions that you find are due to the fact that what you get is the output of System 1, of intuition.
The other system—reasoning proper—is a costlier and more painstakingly acquired system. We’re capable of deploying it when the investment is worth it, and that gives us this relatively superior power, which may not be as superior as classical philosophers like Aristotle assumed, but the possession of which is very much linked to the possession of language and to the possibility of entertaining higher order thoughts. Dual system approaches tend to be rather sketchy. There are many versions and it’s being readjusted all the time. For instance, more and more it’s being recognized that reasoning, the higher system, makes mistakes, too. So it’s not that good. Intuition often is even better than reasoning. The notion that you can explain the mistakes or successes of human inference through dual system theory isn’t that convincing. The actual workings of reason proper are still vastly mysterious.
We took an alternative route. We don’t think that there are two systems. Intuition, anyhow, is not based on one system. We don’t have a faculty of intuition; it is instead based on a great variety of cognitive mechanisms, some of which have a strong innate basis while others have to do more with the acquisition of competences in the course of cognitive development. There are many autonomous systems involved.
Some of these mechanisms of intuition are not just intuitions about the facts in the world, about space, about time, about solid objects, about living creatures; they are also about representations or even meta-representations. So we have intuition about, for instance, the meaning of words, the truth or falsity of ideas, about what other people may think. We have intuition about not just objects in the world, but how they are represented in the minds of others, in our own minds, in abstract ideas, and so on.
This is still the domain of intuitions. Like other systems, it’s a system that’s highly specialized. It works on a very special kind of object, namely representations. Most things in the world are not representations. Representations occur inside and in the vicinity of animals that have a cognitive system. Most of the world is representation free, so to speak. We have a specialized system to have a variety of mechanisms to develop intuitions about representations. Among the intuition we may have about representations are reasons, intuition about reasons for beliefs or decisions.
Why are reasons of any relevance to us? In our own individual thinking, reasons don’t matter very much. We trust ourselves. In any case, you have to rely on your own cognition. You don’t need to look for a reason for what you intuitively believe. If you intuitively believe something, most of the time, that’s it. But if we want to communicate to others what we believe and they don’t have the same intuitions, we may still share intuitions about reasons for our belief, and then we may end up converging.
We also have other uses of reason that don’t have to do with argumentation or convincing others. We have reasons that are of more retrospective character. We use reason to justify ourselves. When we interact with one another, we depend on our good reputation, on the willingness of others to interact and cooperate with us in a variety of ways, and for that they have to think that the way we think and behave makes us reliable partners. The evidence they have is from what we do, which can be interpreted in a variety of ways. What we can do is provide reasons for our actions and our thoughts, not to convince others to adopt the same thoughts or behave in the same way, but to show that we had good reasons and can be trusted to have similarly good reasons in the future.
Reasons have two functions. One is to justify ourselves and the other is to convince others. Reason is one intuitive mechanism among others; it produces intuitions about reasons. Reason serves two main functions. One function is the argumentative function (which we had discussed in our earlier work about the argumentative theory of reasoning), and the other is the justificatory function.
What we are arguing, then, is there is no division between intuition and reasoning. Reasoning is just a certain use of intuitions about reasons. Reason is just as intuitive as all the rest. It doesn’t stand in contrast with another kind of system. It’s one particular kind of intuition, which plays a very important role. It’s just as if you took another kind of intuition, say, about emotions or aesthetic emotion, and said, “Oh, that’s completely different from other intuitions in all the world.” Yes, it has a certain particular role, as do any specialized intuitions, but it’s not a second system. It’s one mechanism of intuition among many others, which, in the case of intuitions about reason, plays an important role in human interaction.
The enigma of reason, we argue, gets resolved in the following manner. To begin with, reason is no superpower. Human beings, like other animals, have lots of mechanisms of intuitive inference. We have, in particular, the ability to represent the representation, to think about them, to have intuitions about them, but it’s still an intuitive capacity. It’s not a new type of capacity, but a new kind of object that we’re capable of having intuitions about. Having objects of thought that are specific to one species to think about or to use in cognition is not only for humans. Animals that have echolocation can exploit ultrasound to perceive their environment in detail and to navigate it, and we cannot. We, on the other hand, exploit reasons in our cognitive work. This is not a second system; it’s just an ordinary cognitive capacity among others, which has important implications for interaction because that’s what drove its very evolution. It’s an ability to understand others, to justify ourselves in the eyes of others, to convince them of our ideas, to accept and to evaluate the justifications and arguments that others give and be convinced by them or not.
Contrary to the standard view of reason as a capacity that enhances the individual in his or her cognitive capacities—the standard image is of Rodin’s "Thinker," thinking on his own and discovering new ideas—what we say now is that the basic functions of reason are social. They have to do with the fact that we interact with each other’s bodies and with each other’s minds. And to interact with other’s minds is to be able to represent a representation that others have, and to have them represent our representations, and also to act on the representation of others and, in some cases, to let others act on our own representations.
We arrive at an integrated view of reason that doesn’t assign it a fantastic goal of unique access to knowledge at the individual level. We think reason evolved in humans and not in other species because there is a specific ecological niche that humans inhabit, which is the sociality that they themselves created. It’s a niche that’s created by a social relationship and culture. In that niche, reason is adaptive and that’s why it evolved
The kind of achievements that are often cited as the proof that reason is so superior, like scientific achievements, are not achievements of individual minds, not achievements of individual reason, they are collective achievements—typically a product of social interaction over generations. They are social, cultural products, where many minds had to interact in complex ways and progressively explore a lot of directions on which they hit, not because some were more reasonable than others, but because some were luckier than others in what they hit. And then they used their reason to defend what they hit by luck. Reason is a remarkable cognitive capacity, as are so many cognitive capacities in human and animals, but it’s not a superpower. It’s well integrated in the minds of one animal and it’s well adapted to a special niche in which this particular animal, humans, live.
The dual system approach is trying to salvage something on the ruins of areas of psychology of reasoning as it had developed in the past fifty years. It has hit on a number of difficulties. It showed that the number of directions that had seemed obvious were, in fact, blind alleys, dead-ends. It ended up having problems and no solution. Sorting the evidence with the idea of two systems—System 1 and System 2—at least seemed to be a step in the right direction. But while it seemed to provide the way to explain why such a capacity as reason might malfunction, as experimental psychology has shown it does, it didn’t solve the other aspect of the enigma of where this unique superpower come from. Instead of having reason in a wider sense as a superpower, now we have just System 2, and that's still highly mysterious.
There are a number of nice gestures, hand-waving in a plausible direction, but what we’re suggesting is at least more precise. Still hand-waving maybe, but more precise hand-waving, leading to unexpected predictions that are experimentally testable and which make more sense, both of the psychological evidence and of the everyday and historical evidence regarding the role of reason in human affairs, in interaction, in the development of science, in the development of negotiation, in politics, and so on. Rather than seeing as a paradox the fact that people can use reason to defend absurd ideas, as we see happen all the time, this is exactly part of what we assume is going to happen.
There’s nothing particularly mysterious about reason as we describe it. The devil is in the details, of course, which we are not going to explore now. Right or wrong, ours is a novel approach to human reason. (This, actually, should make me guess that we must be wrong, because, if you have a deeply novel approach it’s probably a wrong idea.) Our approach really is at odds both with classical views of reason and reasoning and, indeed, with more recent developments like dual system theory.
The overall view I would defend is that we each have a great many mental devices that contribute to our cognition. There are many subsystems. Not two, but dozens or hundreds or thousands of little mechanisms that are highly specialized and interact in our brain. Nobody doubts that something like this is the case with visual perception. I want to argue that it’s also the case for the so-called central systems, for reasoning, for inference in general.