The argumentative theory of reasoning

Back to projects.

Dan Sperber and I wrote a book that develops and extends the argumentative theory of reasoning, called The Enigma of Reason (USUK). It came out in April 2017.

Current philosophy and psychology are dominated by what can be called a classical, or ‘Cartesian’ view of reasoning. Even though this view goes back at least to some classical Greek philosophers, its most famous exposition is probably in Descartes. Put plainly, it’s the idea that the role of reasoning is to critically examine our beliefs so as to discard wrong-headed ones and thus create more reliable beliefs—knowledge. This knowledge is in turn supposed to help us make better decisions. This view is hard to reconcile with a wealth of evidence amassed by modern psychology. Tversky and Kahneman (and many others) have demonstrated the failures of reasoning in decision making. Johnson-Laird and Evans (and, again, many others) have shown how fallible reasoning can be. Others have shown that sometimes reasoning too much can make us worse off: it can unduly increase self-confidence, allow us to maintain erroneous beliefs, create distorted, polarized beliefs and enable us to violate our own moral intuitions by finding handy excuses. Sperber claimed that the full import of these results has not been properly gauged since most people still seem to agree, or at least fail to question, the classical, Cartesian assumptions.

The theory Dan Sperber suggested—the argumentative theory of reasoning—proposes that instead of having a purely individual function, reasoning has a social and, more specifically, argumentative function. The function of reasoning would be to find and evaluate reasons in dialogic contexts—more plainly, to argue with others. Here’s a very quick summary of the evolutionary rationale behind this theory. Communication is hugely important for humans, and there is good reason to believe that this has been the case throughout our evolution, as different types of collaborative—and therefore communicative—activities already played a big role in our ancestors’ lives (hunting, collecting, raising children, etc.). However, for communication to be possible, listeners have to have ways to discriminate reliable, trustworthy information from potentially dangerous information—otherwise speakers would be wont to abuse them through lies and deception. Listeners must have mechanisms of epistemic vigilance. One way listeners and speakers can improve the reliability of communication is through arguments. The speaker gives a reason to accept a given conclusion. The listener can then evaluate this reason to decide whether she should accept the conclusion. In both cases, they have used reasoning—to find and evaluate a reason respectively. If reasoning does its job properly, communication has been improved: a true conclusion is more likely to be supported by good arguments, and therefore accepted, thereby making both the speaker—who managed to convince the listener—and the listener—who acquired a potentially valuable piece of information—better off. 

Our evolutionary account is much more in touch with the prevailing view of the evolution of human cognition. According to this view—alternatively named the social brain hypothesis, or the Machiavellian intelligence hypothesis, among others—most of human cognition evolved to answer the demands of our social world. But, however evolutionarily plausible our theory might be, we still need to provide more solid evidence in its support. Here’s a short list of some of the predictions made by the theory that are well supported by empirical data:

Prediction #1. If reasoning evolved so that we can argue with others, then we should be reasonably good at arguing. Short answer: we are. When we have to make up or evaluate arguments in properly argumentative contexts—we truly have to convince someone, or someone truly has tried to convince us—we are good at it. This good performance stands in sharp contrast with the very poor performance observed in often much simpler reasoning tasks that are not set in argumentative contexts. 

Prediction #2. If reasoning evolved so we can argue with others, then reasoning should yield better results in groups than alone. Short answer: it does. When the performance of groups and lone individuals in reasoning tasks is compared, groups fare much better—sometimes dramatically so. Not only do groups have a better performance than the average individual, but they often perform as well, or even better, than the best group member (again, in reasoning tasks, this is not true across the board). 

Prediction #3. If reasoning evolved so we can argue with others, then we should be biased in our search for arguments. In a discussion, I have little use for arguments that support your point of view or that rebut mine. Accordingly, reasoning should display a confirmation bias: it should be more likely to find arguments that support our point of view or rebut those that we oppose. Short (but emphatic) answer: it does, and very much so. The confirmation bias is one of the most robust and prevalent biases in reasoning. This is a very puzzling trait of reasoning if reasoning had a classical, Cartesian function of bettering our beliefs—especially as the confirmation bias is responsible for all sorts of mischief (cf. prediction #4). Interestingly, the confirmation bias needs not be a drag on a group’s ability to argue. To the extent that it is mostly the production, and not the evaluation of arguments that is biased—and that seems to be the case—then a group of people arguing should still be able to settle on the best answer, despite the confirmation bias (which they do, cf. prediction #2). As a matter of fact, the confirmation bias can then even be considered a form of division of cognitive labor: instead of all group members having to laboriously go through the pros and cons of each option, if each member is biased towards one option, she will find the pros of that options, and the cons of the others—which is much easier—and the others will do their own bit.

Prediction #4. When people reason alone, there is often nothing to hold their confirmation bias in check. This might lead to distortions of their beliefs. As mentioned above, this is very much the case. When people reason alone, they are prone to all sorts of biases. For instance, because they only find arguments supporting what they already believe in, they will tend to become even more persuaded that they are right or will develop stronger, more polarized attitudes. 

Prediction #5. When reasoning is used to make decisions, it will do what it is supposed to do, namely, find arguments. As a result, instead of always pointing towards a better choice, reasoning will usually lead us to a decision that is easy to justify. Psychologists have shown that many a weird decision can be explained by this factor: people decide to do something because they can easily justify it rather than because it is right. For instance, people can be led to choose a big disgusting roach-shaped chocolate over a small heart-shaped one just because the former is bigger and more expensive, and so it’s hard to rationally defend the choice of the later—even though they would surely enjoy it much more!

This is the bulk of the theory, in a very short form (most of the evidence can be found there:  Mercier, H., & Sperber, D. (2011). Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences.) Here are some more recent developments (a short review of all the evidence can be found here).

All the results above are also true of children. Kids start to argue very young and are surprisingly good at it, compared to other reasoning tasks. Children—well, students in general, this is true from pre-school to MIT—understand things much better when they work in group. In the words of the leading expert: “research on cooperative learning is one of the greatest success stories in the history of educational research.” (Slavin, 1996, p.43) This fits very well with our contention that reasoning should work better in groups, and I think this is one of the very important fact that our theory helps make better sense of. Also, children’s confirmation bias is as strong as that of adults (not exactly a shocker). Interestingly, in some of the tasks for which reasoning trips adults up, children can do better because they reason less. (These results are described in this paper: Mercier, H. (2011). Reasoning serves argumentation in children, Cognitive Development).

It is also interesting to notice that argumentation is not a modern or a Western phenomenon. Some might have us believe that illiterate people cannot reason, or that Easterners never argue (this is too extreme, no one serious supports these views anymore, but run down versions are still common). This is not true. People argue all over the world, and all the available evidence indicates that they do it well. Also, people seem to reason better in groups everywhere—where data is available at least. And Cicero would have little to reproach to the Easterners’ tradition of argumentation and persuasion. (This is described there: Mercier, H. (2011). On the universality of argumentative reasoning. Journal of Cognition and Culture.)

Finally, the argumentative theory of reasoning also fits in well, and helps make sense of, the recent developments in deliberative democracy. Deliberative democracy—basically, the idea that more political decisions should be the outcome of deliberation, that citizens should deliberate more—is one of the most potent ideas in current political science. Our theory explains well why deliberation should be a very efficient tool. It also explains why groups can sometimes fail to deliberate properly. In particular if everybody agrees to start with, the confirmation bias of each participant will not be reined in by an opposite opinion and the group will polarize. In the same way as reasoning is not—we surmise—designed to be used by individuals on their own, it is not designed to be used by all-agreeing groups either—neither is a genuine argumentative context in which at least two people disagree about something. One of the most important consequences of our theory—I think—may be at the political / policy level. Based on the dominant, Cartesian view people have been trying for many years to reform reasoning: to teach critical thinking, to rid us of our biases, to make Kants of us all. This approach has not been very successful. According to our theory this is not surprising, as people have been trying to reform something that works perfectly well—as if they had decided that hands were made for walking and that everybody should be taught that. Instead, we claim that reasoning does well what it is supposed to do—arguing—and that it produces good results in appropriate—argumentative—contexts. So, instead of trying to change the way people reason, interventions based on the environment—institutional in particular—are much more likely to succeed. If we can increase people’s exposition to arguments, if we manage to make them argue more with people who disagree with them, then reasoning should produce very good results without having had to be reformed. (The consequences of our theory for deliberation are spelled out in this paper, Mercier, H., & Landemore, H. (2011). Reasoning is for arguing: Understanding the successes and failures of deliberation. Political Psychology)

I’m also working on applications of the theory to moral reasoning (our views are close to those of Jon Haidt, maybe with a more positive role for reasoning overall, see Mercier, H, (2011) What good is moral reasoning? Mind and Society), to expertise, to the history of ideas (if reasoning was designed to help us create better beliefs, why has our knowledge of the world progressed at such a glacial pace for most of our history?), and to science (which is a very social and indeed argumentative enterprise). 

Upshot. Reasoning is made for arguing. Because of this people have a strong confirmation bias that plagues lone reasoners. But when people argue, the biases of the arguers can balance each other out and lead reasoning to felicitous outcomes. Let’s reason together!

You can find a summary of the evidence in this paper.

Protagoras, one of the main sophists.


We have been lucky enough to have some very interesting coverage of our work.

Coverage of The Enigma of Reason

The Guardian (best book of 2017 list) 

Older coverage

The New York Times (and also here)(with a clarification here and a blog post here)
New Scientist (cover story)
Counterpoint (Australian radio)
Wired blog (and also here)
Chosun blog
Bloggingheads (also here and here)


Mercier, H. & Sperber, D. (2011) "Why do humans reason? Arguments for an argumentative theory". Behavioral and Brain Sciences. (Target article)

Mercier, H. & Sperber, D. (2011) "Argumentation: its adaptiveness and efficacy". Behavioral and Brain Sciences. (Response)

Mercier, H. & Sperber, D. (2009) “Intuitive and reflective inferences”. In Evans, J. St. B. T. and Frankish, K. (Eds.) In Two Minds: Dual Processes and Beyond. Oxford University Press.

Mercier, H. (2010) “The social origins of folk epistemology”.
Review of Philosophy and Psychology.

Sperber, D. & Mercier, H. (2012) “Reasoning as a social competence”. In Landemore, H. and Elster, J. (Eds.) Collective Wisdom

Mercier, H. (2013) "Using evolutionary thinking to cut across disciplines: the example of the argumentative theory of reasoning". In Zentall, T. & Crowley, P. (Eds.) Comparative Decision Making. Oxford University Press.  

Mercier, H. (2013) "The function of reasoning: Argumentative and pragmatic alternatives." Thinking and Reasoning

Mercier, H. (2012) "Some clarifications about the argumentative theory of reasoning. A reply to Santibáñez Yañez (2012)" Informal logic

Mercier, H. & Sperber, D. (2012) “Two Heads Are Better” Stands to Reason.Science (Letter)