233 | Hugo Mercier on Reasoning and Skepticism
What this covers
Hugo Mercier, a cognitive scientist at the Institut Jean Nicod in Paris, joins Sean Carroll to examine the origins and function of human reasoning. Rather than treating reason as a general-purpose tool for arriving at truth, Mercier argues that reason evolved as a social mechanism—specifically for producing justifications to convince others and for evaluating the justifications others offer in return. This reframing carries a sharp diagnosis: the human mind is not gullible but conservative, prone to rejecting evidence that conflicts with existing beliefs rather than accepting false claims wholesale. The conversation ranges across why propaganda rarely works through persuasion alone, how trust functions as a learnable skill, and why people often seem to dismiss good counterarguments in the moment yet privately shift their views later.
The episode explores several concrete domains where this theory reshapes common assumptions. Nazi-era propaganda, Mercier argues, succeeded only where antisemitism already thrived—it spread in receptive regions and backfired elsewhere, suggesting that mass media cannot implant arbitrary beliefs. On contemporary polarization, he contends that actual policy disagreement between Republicans and Democrats is smaller than perceived; what has grown is affective polarization, driven by greater ideological coherence rather than social media echo chambers, which he notes actually expose people to more opposing views than offline life does. He also examines conspiracy theories through the lens of institutional trust, arguing that government corruption—not skilled propagandists—creates the soil where conspiracies flourish. Throughout, Mercier positions open-mindedness as a corrective to the cultural valorization of skepticism, suggesting that the real problem is being too conservative about whom and what to trust, not too credulous.
Mercier argues that reason evolved primarily as a social tool for producing and evaluating justifications to others, and that as a result humans are not too gullible but too conservative—rejecting good evidence that conflicts with their priors rather than being duped by persuasion.
- Reason is a specific cognitive mechanism for producing and evaluating reasons, distinct from rationality and most of cognition
- Reason evolved for cooperation and communication, with biased production but objective evaluation of others' arguments
- Mass persuasion (advertising, propaganda) rarely changes minds; people mostly believe false things because they were predisposed to them and resist correction
Reason is a specific cognitive capacity for producing and evaluating communicated arguments, unique to humans and requiring complex communication systems.
- Reason is both the reasons we give each other when trying to convince someone to do or believe something, and the cognitive mechanisms that allow us to produce and evaluate those reasons; it is very specific, and most of the mind operates without using reason at all.
“reason is both the reasons we give each other. So when I try to convince you to do something or to believe something, I will give you reasons that you will evaluate. And reason is also the cognitive mechanisms that allow us to produce and to evaluate these reasons.”
- Humans are the only animals that reason at all, because producing and evaluating reasons requires a complex, flexible, open-ended communication system to express those reasons—something no other animal possesses.
“this ability we have of producing and evaluating reasons is something that as far as we can tell, no other animal has, in part, because it requires a complex system of communication to be able to express these thoughts, these reasons.”
- Although we are conscious of reasoning—which is why people equate it with System 2—much of reason is still unconscious: the reasons that convince someone, and the intuitions about whether a reason is good or bad, simply pop into the head without the person knowing how they got there.
“if I think of a reason to convince you of something, I don't know how I found it. It just kind of pops into my head. When I give you a reason... you have an intuition about whether it's a good or a bad reason, and you don't know why that's the case.”
- Reason could only have evolved against a backdrop of pre-existing complex communication; what is crucial is not language per se but the ability to express and decode meanings—even pointing can convey a reason if both parties share that interpretive capacity.
“for reason to have evolved, for it to have made sense, that reason would evolve, people, our ancestors would already have had a relatively complex communication in place... what's really crucial is not language per se, but more the ability to understand what people mean more generally.”
- Explaining one's own decisions requires a distinct cognitive mechanism analogous to a separate piece of code: a well-functioning system (animal brain or program) produces good decisions as a black box, but giving reasons for those decisions requires additional machinery that introspects on one's own mind and behavior—machinery that doesn't come pre-packaged with the rest of cognition.
“If you want your code to be able to explain to someone why it has come up with that decision, you need to write another piece of code... in order to do that, you need a different piece of software that will look into your own mind and look into your decisions”
- The dual process model dividing cognition into intuitive System 1 and deliberate System 2 is flawed because treating System 2 as the controlled rational part assumes a 'little person inside your head' doing the reasoning, which merely pushes the explanatory problem further back rather than solving it.
“you can't solve the riddle of how people act by assuming that there is a little person inside their head. That's just pushing the problem further. So I completely agree with that criticism.”
- Like any complex system—a cell, a piece of code—the human mind is made of countless specialized mechanisms working in relative isolation, each accomplishing a specific function; rationality is not one module but emerges from the coordination of many such mechanisms.
“the human mind exactly as much as any other organ or any, like a cell or even any complex artifact or complex piece of code is made of countless mechanisms that work in relative isolation from each other.”
- Neanderthals probably had language as sophisticated as ours, implying that our common ancestor and humans roughly 300,000 years ago had broadly the same cognitive capacities we have, while inferring the abilities of pre-sapiens species becomes much harder.
“we know that Neanderthals probably had language that was as sophisticated as ours. And so that means that our common ancestor and probably people about 300,000 years ago had about the same cognitive capacities broadly that we have.”
Reason evolved for social persuasion and cooperation, not for finding truth independently.
- The argumentative theory of reason explains the otherwise puzzling, well-attested confirmation (myside) bias: whenever people have a preconceived notion and start reasoning about it, they overwhelmingly find arguments for why they are correct, irrespective of whether they actually are—because the production side of reason evolved to persuade, not to find truth alone.
“one of the great strengths of our theory is that it explains this otherwise extremely puzzling feature of reason... the confirmation bias or the myside bias... Whenever people have any kind of preconceived notion... overwhelmingly they will find arguments for why they're correct.”
- Reason has two asymmetric sides: when producing arguments to convince others we are thoroughly biased toward our own view (confirmation/myside bias makes evolutionary sense for persuasion), but when evaluating others' arguments our incentives shift toward accuracy, so we judge arguments fairly—accepting good ones even when they challenge us and rejecting bad ones; the back-and-forth between biased production and objective evaluation is what makes group reasoning work.
“we use recent produce reasons and arguments and justifications, and when we do that we are quite thoroughly biased... But there's this other side of reason, which is how we evaluate arguments... the biases that cannot plague the production of reasons completely disappear. So we are quite objective.”
- Even though reason is social in origin and doesn't follow strict logical validity or Bayesianism, the exchange of reasons must on average lead to more accurate beliefs (or fitness-enhancing outcomes), otherwise it would not have been selected for; people's intuitions about good versus bad reasons therefore must track what is actually true.
“if on average the outcome of that process was bad, if on average we ended up with worse ideas... that would not be selected for... the intuitions people have about what is a good reason or what is a bad reason, must somehow track what is actually true.”
- Reason evolved because humans uniquely cooperate and communicate at scale: justifying oneself lets others accurately judge whether you are a competent, well-meaning cooperation partner, and exchanging reasons during disagreements lets communicating partners figure out who is most likely right—creating selection pressures favoring those who could reason.
“when you cooperate with people, it's really important for you to judge them... So if you can explain to me... then I can correct the way I perceive you... And so when people cooperate, it becomes super useful to be able to justify themselves”
- There is no evolutionary pressure to prove mathematical theorems, but the reasoning capacity that evolved for persuading fellow humans is, by lucky accident, the same crucial capacity we now repurpose to prove theorems and develop physical theories.
“reason is not the only ability that we use for science, but it is clearly a crucial one. It did not evolve for that at all. And it's our great luck that we can also use it to prove mathematical theorems”
Propaganda and misinformation exploit preexisting beliefs and coordination rather than implanting new convictions through persuasion.
- Studies (originating with Ian Kershaw and confirmed by later quantitative work) show Nazi propaganda only succeeded where people already held the belief: antisemitic propaganda spread in already-antisemitic areas and even backfired in low-antisemitism areas, while efforts to make workers abandon communism, accept euthanasia, or like the Nazis themselves largely failed.
“what Kershaw thought was that Nazi propaganda worked, so to speak, only when people already believed what the propaganda was saying... it seems to have made things worse marginally, in areas where people were already very highly anti-semitic. But it didn't have any effect or even backfired in areas with kinda low antisemitism.”
- Propaganda can have insidious effects without persuading anyone: it signals that the government backs certain acts (so perpetrators won't be prosecuted) and creates coordination points where like-minded people realize how many others share their views, emboldening collective action—and the same coordination mechanism can apply to good causes like climate concern.
“propaganda can have many insidious effects, even if it doesn't persuade anybody... It also creates these kind of coordination points in which people realize that actually, wow, I'm surrounded by a lot of people who don't like Jewish people either.”
- Vaccine hesitancy exists in every country because people have an intuitive resistance to vaccination, and this pre-existing resistance is then fed by wildly different and mutually incoherent stories across cultures (5G, autism, AIDS), which are custom-made to appeal to each culture's prevalent fears but do not cause the underlying resistance.
“it's really the other way around, it's that there is an audience for such views, because people have an intuitive resistance to vaccination... it is fed by many, many different stories... custom made to appeal to different cultures... but that does not cause the fear in the first place.”
- People believe false things not because a persuasive propagandist convinced them but because they were already predisposed to those beliefs; the evidence is that the same kinds of false beliefs—ancestors, ghosts, witches, certain gods—recur across virtually all human societies, which wouldn't happen if some manipulator could implant arbitrary beliefs at will.
“the type of false things that people believe are surprisingly consistent across the world... if you were someone who could just persuade people at will, why make them believe that?... It's just people are prone... to have some beliefs”
- Advertising largely doesn't shape people's deep preferences (the desire for an SUV pre-exists), but it can still profitably shift which brand of a near-identical product a consumer picks; these effects are small but enough to be a win for advertisers without implying people changed their minds, and much ad spending may be wasted, especially online.
“it is not the case that people's preferences are deeply shaped by advertising... advertising can still work. Not in making you buy an SUV instead of a mini, but in making you buy maybe one SUV compared to another model that is pretty much the same.”
Bias and motivated reasoning in argument production serve evolutionary persuasion functions, explaining why people find supporting arguments for preexisting beliefs.
- Reason is not the same as being logical or rational: most animals are rational in their behaviors without using reason, and humans often use reason in non-rational ways such as motivated reasoning to justify already-held conclusions or delude themselves.
“you can think that most animals are rational in most of the behaviors that they engage in. And according to us, they don't use reason to do that... people can use reasons to engage in moderated reasoning, to justify conclusions that they already believed”
- In the overwhelming majority of cases we form beliefs through other cognitive mechanisms (perception, testimony) and only generate reasons afterward when those beliefs are contested; the reverse—belief formed by accepting a good reason—does happen, as when people immediately revise answers to deceptive math problems once shown the correct reasoning, but it is comparatively rare.
“in the overwhelming majority of cases, we come to our beliefs through... many other clinician mechanisms, like perception... and then when these beliefs are contested... then we come up with reasons. But there are cases in which it's the other way around”
- It is important to distinguish irrational from false: the doomsday cult members who kept their belief after the prophecy failed were not necessarily irrational, because staying maintained their close-knit support group, so there are countless situations in which saying or believing false things can be perfectly rational given the social benefits.
“It's really hard to show that something is irrational and that it's really important to distinguish irrational from false... there are countless situations in which saying or believing false things may be perfectly rational.”
People form most beliefs through perception and testimony before reasoning; changing minds through argument alone is comparatively rare.
- Minds process communicated information very differently from perception: if you see your keys on the desk you immediately update, but if someone merely tells you where they are you stay skeptical—because perception evolved to give an accurate world model whereas other people evolved to serve their own often-divergent interests, so communicated claims require good reasons to be accepted.
“Our minds process communicated information very differently. And the reason is quite obvious, is that in an evolutionary context we can trust our perception... By contrast, other people evolved to serve their own interests, that often do differ from ours.”
People resist changing minds not from reason's failure but from social mechanisms fearing ridicule; good arguments often work privately even when publicly rejected.
- It is very rare to find someone exposed to better versions of their ideas who refuses them; people who believe in Bigfoot or conspiracy theories have almost certainly heard the good reasons against their belief, so their real problem is not being too credulous about the fringe claim but being too skeptical of the strong counter-evidence—a failure to accept correction.
“it's very, very rare to find someone who has not in our societies who hasn't been exposed to the correct versions or better versions of their ideas... Then the issue is that they fail to correct themselves or they fail to accept to be corrected by others.”
- People fail to change their minds in arguments not because reason fails but because other cognitive mechanisms—especially social ones telling us we'll look stupid, inconsistent, or unreliable if we reverse—interfere; reason often works behind the scenes so that someone refuses to concede in public yet privately comes to accept a good argument days later.
“there is kind of interference from other cognitive mechanisms... those that tell us that if you change your mind, you're gonna look stupid... you have a discussion and you don't budge, but then a few days later you're like, 'Oh, yeah, it kind of made sense.'”
- Skilled persuaders cannot easily make us believe false things, because our evaluation mechanisms err on the side of caution: if you don't understand an argument, you stay with your prior belief, so for topics you care about and could pay a real cost on, someone has to give you a genuinely good argument to change your mind.
“I think the way these mechanisms should work is they should err on the side of caution. If you don't really understand an argument then you should just stay put and stay with your prior belief.”
- Psychology theories are like toothbrushes: everybody has their own and nobody wants to use anybody else's, so the field has trouble agreeing even on basic terms.
“For psychologist theories, I like toothbrushes. Everybody has their own and no one wants to use any anybody else's.”
Trust and openness to others' ideas function like learnable skills; excessive skepticism keeps people in ideological bubbles missing valuable information.
- On 'non-attitudes'—issues where people have no prior intuition and don't expect personal impact—people are easily influenced by sources they already trust (e.g., a Bernie Sanders fan adopting his stance on an obscure topic); this is individually rational but can be collectively problematic because aggregated public opinion influences policy.
“political scientists call 'Non attitudes.'... for these kind of attitudes, people are quite easily influenced by people they already trust... it's completely rational at the individual level to do this... But it can still be problematic at the collective level”
- Following Toshio Yamagishi, trust is like other skills: the more you try trusting people the better you get at figuring out who to trust, because if you never trust anyone you never learn whether you should have; so as long as the stakes aren't massive, there is a net benefit to trusting people, whereas being too conservative keeps you stuck in your bubble missing opportunities.
“trust is something like other skills and the more you try trusting people, the better you get at figuring out who you should trust... if you never trust anybody, you never know whether you should have trusted them.”
- What most often stops us from trusting people in low-stakes situations is not fear of losing a little time or money but fear of looking silly if duped; it is worth overcoming this because even if you occasionally lose a bit and look foolish, the learning makes you the net winner.
“what stops us from trusting people is not so much that we think we may lose a bit of time or a little bit of money, it's that we're afraid we're gonna look silly if we get duped.”
- The core human shortcoming is not accepting things too easily but rejecting them too easily; rather than valorizing skepticism, individuals should ask how to open their minds more to information from people in different disciplines, parties, countries, and backgrounds whom they currently under-trust, because by default we are too conservative.
“it's very, very rare to find someone who has not in our societies who hasn't been exposed to the correct versions or better versions of their ideas... Then the issue is that they fail to correct themselves or they fail to accept to be corrected by others.”
Actual policy disagreement between parties is smaller than perceived; rising affective polarization driven by ideological coherence and distorted media exposure.
- In the US the average Republican and Democrat are much closer on actual policy positions than people believe; what is clearly growing is affective polarization—increasing mutual dislike—driven partly by greater ideological coherence, so that knowing someone's party now lets you infer disagreement on many issues and gives grounds to dislike them.
“the average Republican or Democrat tends to be much closer to each other than people think... the thing that is really clearly growing a lot in the US at the moment is affective polarization, where people from both parties increasingly dislike each other.”
- Contrary to the echo-chamber theory, social media exposes people to MORE crosscutting information from the other side than real life does; the problem is that the exposure is to the most extreme, outrage-generating versions of the other side, producing a distorted view that drives polarization—yet since affective polarization hasn't risen everywhere social media spread, social media alone is not sufficient to cause it.
“you are more exposed to crosscutting information from the other side of the spectrum on social media than in real life. But the type of information you tend to be exposed to on social media is some of the most extreme information... social media on its own is not sufficient”
- One should be skeptical of pessimistic stances about new media because across history every new medium, from print onward, prompted predictions of doom yet in every case made things better; new media likely speeds up social change and improves coordination rather than dramatically changing its direction, since people's underlying preferences remain similar.
“across history for every new type of media, whether it's print or whatever else, there has been a lot of people saying it's gonna be the end of the world, and in every case, it has made things better”
Government corruption and institutional distrust drive conspiracy belief; transparency and trustworthiness are the main levers on conspiracy prevalence.
- Chimpanzees usually fail to use human pointing to find hidden food because they live in a competitive world and don't assume others intend to help them; domesticated species like goats understand pointing because of their evolutionary history of beneficial interaction with humans, showing comprehension of pointing reflects relationship history, not general cognitive ability.
“if you take a chimp and there's two buckets... and you point to the bucket that has the food, they won't pick it... That's because they don't assume that you want to help them. They live in a very competitive world”
- At the population scale, the main lever on conspiracy theory prevalence is how trustworthy and uncorrupt a government is: countries with corrupt, dysfunctional governments (and actual conspiracies) put people in a mindset where conspiracies seem plausible, so improving institutional trustworthiness—and transparency from actors like pharmaceutical companies—is the only thing that meaningfully lowers aggregate conspiracy and vaccine-hesitancy rates, though never to zero.
“the only thing that really matters at that scale is how good of a government you have. So if your government is very trustworthy and not corrupt, there will be fewer conspiracy theories... the only thing that is going to meaningfully lower the rates of things like conspiracy theory is if people have... fewer grounds for believing in these things”
- A government tempted to paternalistically keep people from the truth would be better off being transparent, because over the long term transparency increases trust in government, which makes people more willing to believe it when belief matters most.
“transparency would clearly be important”
- When you drive you can't talk to other drivers, which Mercier proposes is a main factor in road rage: normally explaining yourself ('I cut you off because I'm in a hurry') smooths social friction, but removing the ability to justify yourself makes interactions go south quickly.
“when you're driving, you can't talk to the other drivers and I think that's one of the main factors in road rage... as soon as you remove that ability to justify yourself, things go south pretty quickly.”