Skepticism and human reason

by Henry Farrell on January 3, 2023

[attention conservation notice: I am neither a philosopher nor a cognitive scientist]

A quick friendly-critical response to this piece by Liam Kofi Bright, which also plugs some of my own collaborative work with Hugo Mercier and Melissa Schwartzberg.

The short version – many arguments against the human capacity for reason rest on shaky empirics, as Liam argues. But Liam’s counter-claim – that human beings are individually good at reasoning – isn’t necessary to make the case that I think he wants to make.

Even if human beings are bad at (some forms) of individual reasoning, they may be able to reason quite well collectively. That provides a different set of grounds for optimism about human reasoning that is maybe less congenial for analytic philosophy (I’ve no idea how you would begin to model it formally – perhaps others do) but that is robust against possible empirical criticisms that the usual analytic philosophy arguments are not.

Liam’s argument – as I understand it, omitting citations and summarizing it a bit crudely – is as follows. There is a lot of work out there that demonstrates, or purports to demonstrate, the ubiquity of individual cognitive bias. This “sceptical zeitgeist” of “general worries about our reasoning capacities” adds up to a general sense that:

we are ignorant little wretches, prone to error and delusion. We are mere pawns of powerful political agents and indifferent technologies that are far beyond our control. Our science is corrupt, and at most our intelligentsia advocate a managed surrender.

Liam sharply disagrees with the claim that this is the inevitable result of flawed individual reasoning. He argues that

the studies purporting to show that each of us individually is a hotbed of bias and illogic have not held up so well to inspection. Cognitive scientists have not been able to agree on how to reliably test for the presence and degree of implicit biases or what is even at
stake in such debates … Some philosophers and cognitive scientists think the whole notion is simply unsalvageable … and others have been quick to point out that this would undermine many sceptical conclusions people have been inclined to draw …

Furthermore, there is some evidence at least that individuals are pretty rational:

research has been done on how people decide at what level of abstraction they will reason about causal claims …. The results of that research suggested that people are well modelled by a normative account of such reasoning that has us as rational decision makers sensibly handling choices given the value of the information we seek to gain …. Likewise, far from irrational prejudice or mere confirmation bias, experimental and theoretical work has been done which suggests that political polarisation could flow from rational responses to ambiguous evidence ….

Finally, “there is an atmosphere of optimism about the deliverances of scientific reasoning.” All this leads Liam to say that a “stand of thought in analytic philosophy (and allied bits of cognitive psychology) thus runs sharply counter to the pessimistic picture.” He speculates that “the analytic optimists are right about our individual level capacities, although:

at the social level the pessimists’ case seems to me unanswered. While we live in a society wherein the rich and powerful have the means and incentive to subvert scientific inquiry and spread lies that better secure their own position I think there is only so much we can know.

It may be that Liam’s optimistic take on individual human reason is right. I’m considerably more skeptical than he is – I don’t think that many of the experimental results are going to go away. I could be wrong!: as Liam says, we “shall see how these turn out.”

But what I want to argue is that you can arrive at a qualified optimism that is broadly similar to Liam’s, from a very different set of assumptions. Specifically – we don’t need strong assumptions about individual rationality to think that social reasoning processes can, under the right circumstances, work quite well.

Here, I am leaning on a short piece that Hugo Mercier, Melissa Schwartzberg and I have forthcoming in the American Political Science Review. It’s Creative Commons, so I’ll probably stick it up here sometime after it has been formally published (it seems to me to be a bit rude to do so before). So anything intelligent that I have to say below should be attributed to Hugo and Melissa; the idiocies and incomprehensions remain mine and mine alone.

Like Liam, we wrote this piece in reaction against the current “skeptical zeitgeist” about ordinary people’s ability to figure stuff out for themselves. Specifically, we object to the arguments made by skeptics of democracy (and we will have more to say on this soon elsewhere).

But our starting point is not an assumption that people are generally individually rational. Instead, we build our counter-claims on foundations that are loosely shared with some of the skeptics of individual reasoning whom Liam describes (e.g. Haidt). Specifically, we start from the arguments that Hugo and Dan Sperber advance in their book, The Enigma of Reason (short precis here).

What Mercier and Sperber argue is that reasoning emerged not as a way of figuring out the world, but as a social competence. Specifically:

Reasons are essentially social. We produce reasons to justify our actions or beliefs, and to convince others.

They argue that the study of reason is:

divorced from that of logic (or of any other system of formal rules). Instead of being the foundation of reason, logic is a rhetorical tool that helps us express arguments more clearly by highlighting and often exaggerating the relation between premises and conclusion.

On this account, we advance reasons, not because we are logically trying to puzzle out the world, but because we want to justify ourselves. That means, among other things, that our reasons are often illogical and at odds with the evidence, in ways that we are ourselves incapable of seeing and understanding. Our reasons are biased.

So far, so skeptical. But where it gets interesting is when we start looking not just at how humans give reasons, but how they evaluate reasons that have been given by others. And here, Mercier and Sperber argue that we are pretty good. We may have ‘myside bias’ – i.e. we are not sharp-sighted about the flaws in our own arguments – but the experimental evidence suggests that we are readily able to see the flaws in the arguments of those whom we disagree with.

That in turn means that even if individual reasoning is basically flawed, collective reasoning, under the right circumstances, can work very well. With the right kinds of group structures, and a bare minimum of goodwill or mutual endurance, we can correct each other’s errors, each, through frank and forthright criticism, obliging each to recognize the weaknesses in their own arguments, and improve them. This is not a Habermasian paradise – there may be no convergence on a shared consensus. But there may be better reasons given on both sides of a continuing disagreement. And bias may have benefits – my curmudgeonly unwillingness to concede may oblige others to improve their claims.

What this implies (as we argue in the forthcoming piece) is a kind of qualified optimism about democracy. If you can construct group institutions that oblige this kind of debate, while making it harder for people to cluster around their shared misconceptions, you can construct relatively unbiased collective reasoning on the foundation of highly biased individual reasoners. Of course, figuring out how to build such institutions is a challenging research agenda in itself. Similarly, although we don’t do this, you could explain the “deliverances of scientific reasoning” without resting your explanation on strong claims about the rationality of individual scientists, or alternatively relying on funerals as the primary engine of scientific advance.

Finally, this framework could accommodate many of Liam’s implied arguments about the problems of currently-existing social arrangements for knowledge, while also providing a helpful frame for folk explanations of the demonstrable idiocy of many very rich and powerful people. If reasoning is a social process of the kind that Mercier and Sperber describe, then we must usually rely on others’ ruthless criticism of our own stupidest ideas, rather than our own capacities to realize when we are, in fact, being fools. Those who are very rich and very famous are notoriously likely to see their crochets indulged rather than excoriated. The Elon Musk text message corpus – and its still-unfolding consequences – demonstrate how very badly this can go.

None of this is to say that our account is necessarily better than Liam’s. Instead, it seems to me to have different tradeoffs. It may be more empirically robust than Liam’s – at least in the sense that it is loosely compatible with a broader chunk of the existing research out there, and less reliant on new research uncovering further evidence for strong individual rationality. Equally, it is based on a less analytically tractable set of initial assumptions than Liam’s proposed alternative. If people are mostly individually rational, then their interactions can more readily be modeled using standard formal techniques, making analytic social epistemologists and game theorists happy, and providing a crisper account of how social knowledge works, without skimping too much on plausibility.

So you can reasonably pick and choose. But if you want to be more constructive, you can also see this as reason (irony intended) to believe that Liam’s broad conclusions (that we can be optimistic about reasoning, but only under circumstances that are different and more equal to those we have) are plausible, even if one starts from quite different premises.

Equally (and again we talk more about this in forthcoming stuff), our account provides evidence against some of the alternatives that people have advanced to democratic reasoning. There is no reason to think that the reasoning of experts or CEOs or whoever is any less liable to bias than that of ordinary people, absent appropriate structures of social accountability, and some good reasons to suspect that it will on average be more biased in problematic ways.



both sides do it 01.04.23 at 4:07 am

you can construct relatively unbiased collective reasoning on the foundation of highly biased individual reasoners

This is a perfect description of a group of machine learning algorithms called “ensemble learners” (which include workhorses of ml: random forest, xgboost, etc). Right down to the term “bias”.

If some future audience for this argument is at all familiar with data science, chuck that metaphor in there, putty in your hands


William S. Berry 01.04.23 at 4:16 am

@Henry Farrell:

You haven’t posted here in some time; I’m pleased to see you back with such an interesting post. To my mind, at least. you’ve always been one of the best writers and thinkers at CT.

I intend to reread your piece again (in the light of day!), but I want to express some random thoughts in advance.

“Long story short”, as the kids say, do we actually understand what we’re talking about here?

I don’t think so.

I think that our minds deal in (relatively) discrete, or intuitive concepts. There is nothing “I think, therefore I am”, there is only the affirmation: “I think: I Am”.

The entire concept of logic is literally meaningless.

Everything we know and understand is what we experience.


Jens 01.04.23 at 5:46 am

What role does the state play in these processes? Is it an arena for the interplay of individual discourses? Or is it a lever for certain individuals and groups to impose their limited view? Doesn’t the word “democratic” conceal a structurally very biased factor here? Is this collective reflection a centralized or a decentralized process? Can institutions that promote collective reasoning really be planned top-down? Or does a particular collective outcome, taking into account individual perspectives, not result from the interaction of individuals ( or not) ?


Tim Worstall 01.04.23 at 10:15 am

How much of this is a retread of Galton’s Ox? Or is informed by that and the Wisdom of the Crowds thing. Because the issues and even conclusions look very similar. Which is good, if we’ve two independent lines of reasoning which lead to the same conclusion we’ve more support for that conclusion. Or are they not independent lines, is all that stuff from economic reasoning (where I tend to put Galton’s Bovine etc) already in the bases of this discussion?

There is one difference:

“It may be that Liam’s optimistic take on individual human reason is right.”

As the other strand takes it as possible (at least) that individuals can be wildly deluded as to fact but that the social interaction or averaging zeroes in on truth.

“If you can construct group institutions that oblige this kind of debate, while making it harder for people to cluster around their shared misconceptions, you can construct relatively unbiased collective reasoning on the foundation of highly biased individual reasoners.”

That other strand of reasoning describes this as the necessity of avoiding groupthink.

What strikes here is that the discussion does seem to be about very much the same thing. But how much are the two discussions informing each other and how much is being done in the absence of reference to that other discussion?


MisterMr 01.04.23 at 10:27 am

Question: how do you define “reason” and “rational”?

For example, I vote for party A and believe everyone who votes for party B is a hack and a liar. Therefore I don’t believe in what people from B say, and even if it sounds true I always suspect a trick.

Do you count this as rational or irrational?

IMHO in a world with limited information (which is to say in most real life situations) it is perfectly rational, however it seems to me that if you contrast rationality with “bias” this would be irrational to you.


Liam 01.04.23 at 11:36 am

Thank you for this lovely response piece! I read it with great interest, and then the short coauthored essay you linked too.

I think we agree more than you think we agree! This for two reasons. First, I think you somewhat misunderstand my basis for individual level optimism. Much of what I find impressive in the anti-sceptical counter-story is not so much the empirical fightback refutation of prior results as the greater sophistication in analysis of what exactly a normatively rational response would consist in. I think much of the trick in the arguments for pessimism about our epistemic categories relies on never quite stating what good behaviour would have been, and just presenting what is actually found in a negative light. Even in the Mercer & Sperber story after all what you find is people very intelligently using their rational capacities to improve their group standing – this is rational activity just directed towards a different end than one might have guessed from a surface level analysis.

For a more detailed example, you mention myside bias, which I take to be a sort of confirmation bias crossed with group identification. This is precisely the sort of thing which I think has been overly hastily ascribed as irrational. But it seems to me that people who think about what rationality would ask of us in a sophisticated way at both the individual ( and group ( level quite often find that this behaviour is not so disreputable at all. If you just leave what good rational behaviour at an intuitive level it seems bad, but when you seriously think through the consequences of constrained agents doing their best in scenarios we face it seems not so unreasonable after all.

So I don’t see the sort of things you point to as reason for pessimism – and I say that despite agreeing on the substance of a) the empirical results, and b) the overall conclusion that what ultimately grounds reason for optimism is the hope that we could properly arrange our society.

Second, oddly enough from the other direction I do agree that individual reasoning behaviour is not what we would want from a social perspective. This is because I am by trade a social epistemologist, someone who thinks about how we should want group inquirers to behave, and far from finding it convenient that people there are epistemically ideal in fact this is the opposite of our usual tenor! It is a foundational result that the two do not align ( and are the classics usually cited) and much of our activity is spent working out how what to ask of individuals that they may facilitate group success (e.g. or So I am far from hostile to the idea that good group reasoners may be composed of “bad” individual reasoners; this is my stock in trade.

So yes, I am happy to note the split between what we want from groups and what we want from individuals, and also in agreement with the sort of empirical results you mention without finding that shaking my humanistic optimism about us as people. What is best though is, as you noted, where we come to in the end. Ultimately the place we all arrive is a faith in the capacity of democracy – and I might add ( a far greater one than our formal democracy under capitalism, leaving much of our social life and productive capacity subject to the whims of a tiny few – to guide us aright. I think this agreement is far more important than any other points of disagreement (and for the sorts of reasons we are discussing here in any case disagreement has its good place – just a shout out to this great book I wanna plug so I am happy for the engagement by people i view to be true comrades.

I hope in future to play my small part in the project of what you call analytic democracy. I am grateful to you for making me aware of it, and thinking my work worthy of response. Thank you!


ShImran 01.04.23 at 12:29 pm

Skepticism is important because it helps us avoid blindly accepting claims or ideas without proper justification. It encourages us to seek out evidence and to consider alternative explanations, which can help us arrive at more accurate and reliable conclusions. Skepticism can also help to prevent us from falling victim to misinformation, deception, or manipulation.


Henry Farrell 01.04.23 at 12:46 pm

More in response to Liam soon-ish (this is a great discussion and one I’m delighted to do a small bit to get going). But quickly, “both sides can do it” might be interested in taking a look at this:

Hong and Page’s work provides a particularly clear, if stylized, model of how diverse individual perspectives or heuristics can combine for better problem solving. This observation is highly familiar in machine learning, where the large and rapidly-growing class of “ensemble methods” work, explicitly, by combining multiple imperfect models, which helps only because the models are different (Domingos 1999; Schapire and Freund 2012). In some cases it helps exactly to the extent that the models are different (Krogh and Vedelsby 1995). Different ensemble techniques correspond to different assumptions about the capacities of individual learners and about how to combine or communicate their predictions. The latter are typically extremely simplistic, and understanding the possibilities of non-trivial organizations for learning seems like a crucial question for machine learning, for social science, and for the cultivation of effective citizenship.

As is likely clear to anyone who knows us, this point is all down to Cosma rather than to me (but then pretty well all my thoughts on these questions are in large part down to co-authors). Scott Page has been talking for a few years about applying bagging and boosting to social decision making processes, but I don’t think that this has turned into published work yet. So yes – you are right – and plausibly right in the sense that this is more than just a neat metaphor.


Henry Farrell 01.04.23 at 12:56 pm

And Tim – this is a set of debates that is certainly quite closely related to the wisdom of crowds, but that invokes a very different set of mechanisms than e.g. Galton. The weight of the ox is determined by an average of independent guesses – people don’t talk with each other or influence each other to get their answers. Some other mechanisms of aggregation – e.g. Condorcet’s voting argument – basically require that individual judgments be completely independent of each other. This paper by Cosma is likely helpful on the differences; as he pungently puts it:

It might be thought that the theoretical explanation is rather simple, and goes (currently) under the name of “the wisdom of crowds” (Surowiecki 2004): individuals make noisy guesses, which on average are unbiased and uncorrelated, so simple averaging leads to convergence on the appropriate answer. Taken seriously, this explanation implies that our economy, our sciences and our polities manage to work despite their social organization, that science (for example) would progress much faster if scientists did not collaborate, did not read each others’ papers, etc. While every scientist feels this way occasionally, it is hard to take seriously. Clearly, there has to be an explanation for the success of social information processing other than averaging uncorrelated guesses, something which can handle, and perhaps even exploit, statistical dependence between decision makers.


Trader Joe 01.04.23 at 1:01 pm

So if I’m following this right, there are basically two simple premises

No one really knows what they are talking about


Everyone’s a critic

Accordingly – by iteration and dialogue, eventually the collective “we” get it right.

It sorta passes the smell test of my own (biased) experience. Though maybe I’ve vastly oversimplified.


MisterMr 01.04.23 at 3:02 pm

I skimmed the article, and it didn’t answer my question, so I’ll repropose it in a more specific way:

I believe we can divide “thinking” in a part construens, where we came up with ipotheses, intuitions etc.; and a part destruens, where we check our intuitions empirically, or for consistence, or against other people’s opinions etc.

I take it as obvious that “rationality” can only apply to the part destruens and not to the part construens, for example formal logic can show if an argument in itself is flawed but cannot initiate a new argument.

Even this pars destruens can only be described as an algorithm were we check the consistence of our intuitions, we can’t check “rationalty” against the truth because this implies that we already know the truth, but the one who is thinking doesn’t know the truth from before or he would have no need to think.

So rationalty can only consist in checking intuitions VS other forms of knowledge, for example I might trust this guy’s opinion a lot and double check my opinions if they differ.

I can’t really find a better definition of “rationality” and I’m quite surprised when I read stuff that rationality somehow is the same of formal logic, and that other forms of thinking are therefore treated as “euristics”.

It gives me the impression that people are used to know the solution to a problem, and therefore treat as “rational” the processes that lead to the correct solution, but evidently rationality if it has to have any meaning cannot imply the previous knowledge of the solution.

For example, suppose I offer you a bet, you pay 1$ and if I roll two dices and get a 7 I’ll pay you 7$. Is it rational for you to accept the bet?
If you know probabilities two dices will roll a 7 with a probability of 1/6, so it is a good bet for you.
In some sense it would be “rational” for you to accept the bet, but then most people do not know probabilities, should we call all of them irrational?
Some might think, for example, that if I offer this bet I expect to win, and refuse the bet. Is this irrational or an euristic? Or is this rational in absence of knowing probabilities?

In most of our daily lives we have to make choices in situations where we cannot have all the relevant data, and our brains evolved to solve this kind of problem, but it would be too restrictive to call all this irrational or an euristic IMHO.

Therefore I would like a definition of rationalty that doesn’t imply knowledge of all the relevant data.


J, not that one 01.04.23 at 5:45 pm

I was wondering what “democratic decision making” is intended to be, as it seems like it could be anything from consumerism to the general assembly to the mores of the traditional farming community, which would appear to be quite different from one another, and all of them from something like formal structures for aggregating opinions efficiently.


Henry Farrell 01.04.23 at 7:35 pm

An initial response to Liam. First, thank you! It’s just great to have this conversation up and running. And thanks for the many cites, some of which I’d read (Zollman; Gabriel and O’Connor) and some of which I’d never even heard of (Solomon). I’m an enthusiastic but very erratic amateur reader of social epistemology, mostly because I think that the questions it is asking are exactly the ones we need to know more about. I also agree that this conversation is mostly about pretty specific differences in the context of a high degree of overall friendly agreement. For many of the important questions, perhaps most of ’em, you could frame the claims in terms either of individual rationality or of Paris school cog psych without a huge amount of violence.

So I think that maybe the most useful framing of what disagreements there are there, is in their cash value. What kinds of practical changes does a Mercier and Sperber account suggest we ought have, and where do they differ from an individual rationalist one?
First, the strong areas of concordance. Both suggest that the current pervasive skepticism about the capacity of ordinary people to reason is unwarranted, and that we need to think constructively about improving the group circumstances under which people collectively can figure out what they want.

I think though – and this is less even a couple of intuitions than handwaving in the vague general direction of same, that there are two possible areas of divergence. First – if the Mercier/Sperber account is right, then we ought to design institutions so that they capitalize on some mental modules rather than others. Specifically, we ought to build institutions that build on our capacities for criticism. Second, and even more vaguely, I think it’s easier to figure out how institutions like political parties ought work on a Paris school basis than on more purely rationalist foundations. There’s some interesting discussion in Boyer’s Minds Make Societies of how we seem to have hardwired mental mechanisms for coalitional politics, which perhaps complicate the epistemic benefits of parties in some important ways.

But those are vague intuitions – and the more important things, I think, are the broad agreement, and the possibility for a real debate about democracy and related questions across different perspectives. I feel that there’s a nascent conversation beginning to come into focus between various bits of political science, political theory, philosophy, network science/sociology (e.g. James Evans, machine learning a la Melanie Mitchell, Scott Page etc, and other places. It would be lovely to figure out more ways to push this along collectively.


Chris Stephens 01.04.23 at 8:02 pm

Henry: you might find Elizabeth Anderson’s essay “The Epistemology of Democracy” of interest.


LFC 01.04.23 at 9:50 pm

from the OP:

“That…means that even if individual reasoning is basically flawed, collective reasoning, under the right circumstances, can work very well.”

I think the phrase “under the right circumstances” is crucial and is therefore doing a lot of work here. That was also the case in Knight and Johnson’s The Priority of Democracy, where they argued that effective collective reasoning (i.e., deliberation) requires not only “equal opportunity for political influence” but also rough equality of individuals’ capacities to make coherent and potentially persuasive arguments.

To get those conditions, you need at a minimum a society with substantially less income and wealth inequality, and much more educational equality, than currently exists in the U.S. And until and unless those conditions are achieved, all these arguments about how well collective reasoning can work will remain theoretical — in the sense that there will not be a way to test if the arguments are correct, because the “right circumstances” and conditions for effective collective reasoning will not exist.


J-D 01.04.23 at 10:12 pm

For example, I vote for party A and believe everyone who votes for party B is a hack and a liar. Therefore I don’t believe in what people from B say, and even if it sounds true I always suspect a trick.

Do you count this as rational or irrational?

That must depend at least in part on why you vote as you do for party A and why you believe what you do about party B. It’s not rational to reach a conclusive evaluation without some information about that.

For example, suppose I offer you a bet, you pay 1$ and if I roll two dices and get a 7 I’ll pay you 7$. Is it rational for you to accept the bet?

It’s not rational to make my choice, without additional information, on the basis that your dice** are fair. Will you let me choose my own dice and roll them myself?

** ‘Dice’ is plural; the singular is ‘die’.


Alex SL 01.04.23 at 10:55 pm

Many agreeable individual observations here. Yes, we are often individually biased but can spot flaws in others’ arguments. Yes, we often rationalise a position we already hold instead of reasoning ourselves into the right position. Yes, reasoning collectively can work better than a single person, because we can critique each other. But the conclusions seem a bit vague and over-optimistic.

There are two important areas of collective reasoning that are interesting to contrast: politics and science (or academia more broadly, but I am a scientist, so I will stick to that).

Science as currently practiced has flaws, and it is run by humans who have flaws, but for all that, it is deliberately and purposely designed to maximise people poking holes in each other’s reasoning. We peer review each other’s manuscripts, each other’s grant proposals, each other’s promotion and tenure cases. We are trained to formulate ideas and then test them, i.e., try to disprove our own ideas. We take each other’s results apart on social media or via rebuttals and responses in the literature. All knowledge is treated as tentative by default, only valid until somebody comes up with something better. And, crucially, the whole system isn’t a democracy: we don’t vote about what is correct, and we don’t expect the current state of knowledge to provide any benefits to an interest group or to be more popular than a different possible conclusion.

Democracy as currently practiced ensures that voters can, if they want to, replace officeholders who are incompetent, corrupt, or advancing harmful policies with new officeholders who are competent, honest, or advancing beneficial policies. That is much better than, say, feudalism. However, even if we imagine a media ecosystem not owned by far-right billionaires but instead providing voters with accurate information, such a system is still by structural necessity set up to reward what is popular and in people’s short-term self-interest over what is correct or long-term sensible. Also, through institutional incentives democracy unavoidably gravitates towards tribalism, because banding together in parties increases your ability of achieve the outcomes you want compared to not doing so. Having a parliament full of independents is extremely inefficient, and the first group of them to form a party will wipe the floor with everybody else, so that others will have to follow suit.

As for having a direct democracy of plebiscites, and assuming for a moment you could organise that for a society significantly larger than Switzerland, you may quickly get incoherent results that make rational governance impossible simply because people have competing values interests, and things are complex. A concrete example is Brexit, which was decided via plebiscite, but where the majority was clearly composed of at least two or three groups that had entirely incompatible ideas of how Brexit would be delivered or what it was for (fortress Britain with no immigration; globalist and deregulated “Singapore-on-Thames”; and only out of the political EU but stay in single market; not to mention protest voters etc).

Much of the Brexit debate was characterised by people wanting a benefit (e.g., “sovereignty”) without accepting that there will be an unavoidable trade-off (e.g., economic damage). But a quick thought experiment illustrates an even worse problem with direct democracy. Assume a society in which a third of the voters want well-funded public services paid through higher taxes but not public debt; a third want well-funded public services paid through increased public debt but not through higher taxes; and a third want public services slashed so that both taxes and public debt can be minimised. Importantly, in contrast to unicornist Brexiters, all three groups have internally consistent wishes: neither of them wants to have a cake AND eat it; they are reasoning well from their values.

But you can see where this is going at the societal level. Assume three plebiscites: one to increase spending in public services; one to make it illegal for the government to raise taxes; and one to make it illegal for the government to increase public debt. Logically, they would all win by two-thirds of the vote, if all three groups of voters turn up to in equal proportions, and nobody has done anything illogical either at the individual level or at the level of collective debate and decision making.

In other words, I agree that collective reasoning can work well and overcome many of the problems of individual faults and biases, if we “construct group institutions that oblige this kind of debate”, as we have done in contemporary science. I just don’t see how that kind of institutional arrangement is compatible with democracy, which is by its nature a mixture of (1) popularity contest and (2) struggle of competing values and (3) struggle of competing interests instead of a reasoning process. Really, reasoning well only properly applies to “logically trying to puzzle out the world”, so it is unfortunate that this domain was effectively put aside for the argumentation in the blog post.


MisterMr 01.05.23 at 12:30 am

@J-D 16

But in the real world, outside of tought experiments, you can never be certain that you have all the relevant information.
I will not let you throw your dice, because you might have unfair dices.
So you’ll have to choose based on how much you trust me that my dice are fair. This clearly is not what we usually mean by “rationality” but it is an extremely common situation, and our brain evolved for it.
Hence my question, what exactly do we mean by “rationality”.


J-D 01.05.23 at 5:14 am

But in the real world, outside of tought experiments, you can never be certain that you have all the relevant information.

The problem is not one of not having all the relevant information, it’s a problem of not having sufficient information to justify a conclusion. In my experience, this problem is at least as great inside thought-experiments.

I will not let you throw your dice, because you might have unfair dices.

It’s exactly that consideration which is a rational basis for not taking the bet, or at least for hesitating about doing so. If somebody offers you odds which seem too good to be true, one common reason is that they are too good to be true. If you can’t figure out what can possibly in it for them from the proposition, it’s rational to consider that there may be some information which is being withheld from you. This is a serious practical point: it is, for example, the reason why you should never play Three-Card Monte. If you don’t include this kind of thinking under the heading of ‘rationality’, why not?


TM 01.05.23 at 9:37 am

Is this ox story of Galton’s credible at all or did he just make up a good anecdote?

Alex 17: “assuming for a moment you could organise that for a society significantly larger than Switzerland” It isn’t clear to me why size should matter for this. Switzerland has 26 different states and four official language communities. Why should running plebescites be any easier in Switzerland than say in Germany or even the US (the US btw does have very frequent plebiscites, just not on the federal level)?

You are correct that incoherent results can occur (I’m aware of one relevant case in the last 50 years or so), although this can happen in other systems as well (take the debt ceiling in the US which can lead to almost the exact scenario you have outlined). In Switzerland, as in the UK btw, it’s the Parliament’s responsibility to resolve such situations. The problem with the Brexit vote, in my opinion, was that Britain’s political culture simply isn’t suited for plebiscitary democracy. The Swiss political culture and institutional setup is very different. It’s not just the referendums.


Liam 01.05.23 at 10:15 am

Thanks for the reply Henry! I’ll just note that I agree that this is a good conversation to get going, and I will be interested to see how it develops! I doubt I will have much to contribute, but always happy to read along :)


MisterMr 01.05.23 at 1:11 pm

@J-D 19

I think people use the world “rationality” to mean very different things, and depending on the use of the term this “trust” calculus might be in or not.
If we oppose “rationality” to “bias”, as the OP and the linked article do, then choosing a course of action depending on how much I trust this or that other person or group would be an example of bias.
In my previous example, I trust party A and not party B, and from my point of view this is rational, but from someone else’s point of view this would be bias.
In the bet example, you don’t trust my intentions, this might count as rational or not depending on the definition of rationality.
Generally trust is based on emotions so I would not see it as a part of “rationality”, but as you say it could be seen as such, however my understanding of the OP and of the linked article is that they have a more abstract and philosophical understanding of “rationality” as a sort of pure reason; such pure reason is defined too poorly IMHO and this is my problem with both the OP and the linked article.

I’ll muse on this a bit: many years ago I read Jung’s book “psychological types”, and recently I reread it out of a curiosity for psychanalysis (together with other old books by Freud, Adler, and others). Anyway Jung postulates 4 brain functions, Tought, Feeling, Intuition and Senses, without defining them very well. Reading the “character types” linked to the functions I realized that Jung meant very different things by those four names than what I mean, for example by “feeling” he for some reason means mostly pro-social feelings and the habit of going along with others. By Intuition he probably implies something about his theory of archetypes, though he doesn’t tell this explicitly; in other schools I’ve seen the term “intuition” applied to the ability of understanding other people’s emotions, that in my opinion is not intuition at all, etc.
All these terms like intuition, coscience, self coscience, reason, emotions etc. do not have fixed meaning and often the meaning is defined in a specific way inside a school so that it is impossible to say if Freud, Jung, Lacan or a philosopher mean the same thing when they speak of “consciousness”, and the same goes with “reason”.
So if we have to distiguish “reason” from “euristics” I would need fist a clear definition of what is meant by “reason”.


TM 01.05.23 at 2:49 pm

The OP uses the terms reason or reasoning 39 times and I have to agree with MisterMr that I still don’t know the author’s intended meaning of the word.


steven t johnson 01.05.23 at 4:40 pm

“Science as currently practiced has flaws, and it is run by humans who have flaws, but for all that, it is deliberately and purposely designed to maximise people poking holes in each other’s reasoning. We peer review each other’s manuscripts, each other’s grant proposals, each other’s promotion and tenure cases. We are trained to formulate ideas and then test them, i.e., try to disprove our own ideas. We take each other’s results apart on social media or via rebuttals and responses in the literature. All knowledge is treated as tentative by default, only valid until somebody comes up with something better. And, crucially, the whole system isn’t a democracy: we don’t vote about what is correct, and we don’t expect the current state of knowledge to provide any benefits to an interest group or to be more popular than a different possible conclusion.”

The adversarial model of science, like the adversarial model of the US courtroom, as the singular path to truth works perhaps as well for science as it does for the justice system? It seems to me that J.D. Bernal lived in vain. But he convinced me that a great part of science progresses by technical progress in instruments and simple description of the world, not by trials. The creation of new problems by mere {“mere?”) technological advances is essential. Maybe most of all, the generalization of scientific principles (once called “laws of nature”) is not the whole of science…but like new discoveries with new instruments and whole new fields, it’s not clear how this part of science fits in with the vaguely Popperian falsificationist paradigm.

Taking another angle, one of the key tasks in effective scientific research is asking the right questions. One part of physics has been dominated by research into what’s called string theory, all of which has perfectly met the procedural criteria put forth above. But has it been successful? The example of parapsychology should provoke concerns. It seems to me to be misleading to say all knowledge is tentative, whether by default or whatever. It seems to me that what counts as scientific knowledge should better be called “corrigible,” with the clear understanding that no mere skepticism counts as a correction. Only certain kinds of evidence (usually labeled the “testing” part of science,) count in defining something as knowledge, even corrigible knowledge, but most of all only scientific evidence counts as critique. The issue of course is what counts as scientific, which in practice tends to mean, measured or taking into account experience.

It is “democracy” that is truly skeptical, admitting religious, racial, all manner of “explanations” for the way things are. Reasoning from such flawed premises may only accidentally find the correct answer, I think. GIGO.

In relation to the OP, Mercier and Sperber should be accompanied by the very closely related Hanson and Simler. Evidence that the mind is really a modular as presumed by the quartet should be presented. And, actual traces of differential reproduction that selects for traits (either positive or negative) should be addressed as well. As I recall the bits I’ve read, I don’t think any of these conditions are well-met.

As to the interesting questions as to, what is rationality? In general, actions that are likely to attain ends. Reason is instrumental or it is nothing in my opinion. A lot of old writing on the irrationality of individuals focused on the interests and the passions, both. But a lot of it focused on the irrationality of the masses, disdained as the mob or whatever. The old objection that a lot of what is held to be true is deliberately taught seems to me to be valid. Thus the conclusion the rabble will kill us all seems to me more a premise for an agenda than a conclusion. Or, is the problem with Fox News the audience or the people buying the audience?


Alex SL 01.06.23 at 2:54 am


Not saying that incoherent results cannot occur in representative democracy – of course they could even in an autocracy if the dictator was just addled enough. The point of my thought experiment was that whereas in representative systems they may or may not happen depending on the rationality of the politicians, they will happen in direct democracy even if everybody is individually perfectly rational.

I am not a political scientist, but it appears to me that despite its linguistic diversity, it is still easier to build the political culture that can handle plebiscites in small Switzerland than in a large, much more diverse nation like India or the USA.

Must really disagree about the problem with Brexit, though. It isn’t just the political culture of UK versus Switzerland, it is so much more. First, answering a complex question like that with a simple yes or no in the first place. This isn’t a binary like “should this disadvantaged minority get this additional right” yes or no, where it is very clear what happens in either case. The ‘leave’ answer conflated everything from remaining in both SM and CU all the way to what was later called No Deal. Second, massive ignorance and denial, ranging from the belief that the EU would disintegrate the day after a successful leave vote to the belief that the UK export industry could simply replace all its trade with the EU by trading more with New Zealand or, I dunno, Nepal or something. Third, and relatedly, making a decision about something where you don’t even have freedom to implement the decision in the way you promised to voters. It is one thing to first negotiate terms with an external party and then put them to the people for either acceptance or rejection of that agreement, it is quite another to promise to the people what beneficial terms you will get from the EU upon leaving, and then the EU laughs in your face when you try to implement the mandate from the plebiscite.

None of that is different in Switzerland, regardless of its political culture, as was amply demonstrated at a smaller scale with the freedom of movement issue. Correct me if I am wrong (as I have not lived in that country for over a decade now), but my understanding is that the Swiss decided via plebiscite to opt out of freedom of movement with the EU, as if they could just do that in isolation without consequences, and then the EU made clear what benefits Switzerland would lose as a logical consequence if they went through with it? I mean, I can use my glorious human Freedom Of Will, like a nation would use its democratic sovereignty, to decide that a company or university should pay me a large amount of money, but if I am not willing to trade them anything in return I may soon run into a snag with my free and unfettered decision.

But ultimately, these considerations are IMO all slightly off-topic anyway, because my main point was that applying “reasoning well” to democratic decision making is a bit of a category error. Democracy is mostly about interests and values. That being said, I see in responses in this thread already hints how one can redefine ‘rational behaviour’ to mean ‘blindly support one’s own group even when it is completely unhinged, because it promotes group cohesion and thus strengthens one’s own group over the opposing group’. Okay then.


J-D 01.06.23 at 8:50 am

think people use the world “rationality” to mean very different things …

I am confident people use it with different meanings (that’s utterly unremarkable), but unsure how much difference there is.

If we oppose “rationality” to “bias”, as the OP and the linked article do …

I am also confident that the word ‘bias’ is used in different ways, and therefore …

… then choosing a course of action depending on how much I trust this or that other person or group would be an example of bias.

… I am not sure whether the kind of bias you are referring to here is an example of the kind of bias referred to by Liam Kofi Bright or by Henry Farrell.

All of that aside, I think that the kind or degree of trust (or of distrust) extended to people should be affected by evidence. If there is some way of using the words ‘rationality’ and ‘bias’ which means that the only way to be rational and unbiassed is to extend an equal degree of trust (or an equal degree of distrust) to each and every individual, regardless of the evidence, then what I would say about that is that people should not seek to conform to that standard of rationality or to avoid that kind of bias.


LFC 01.06.23 at 3:40 pm

My impression from reading H. Farrell at CT over the years (I can’t claim to have read, for the most part, his academic publications) is that he believes that electorates or “mass publics” are capable of reaching “reasoned” decisions under the right conditions— i.e., among other things, where reasonably accurate information is widely disseminated and where a range of different viewpoints have roughly equal chances to be heard. (One of the attractions of the Internet from this standpoint is that it has the potential to facilitate this, at least in theory.) “Reasoned” decisions in this context means decisions that successfully identify the main societal problems that need to be addressed and result in the election of politicians who take “evidence-based” steps to address them and who are open to revision of their approaches based on how effective their attempted solutions actually prove to be. In this sense a well-functioning democracy is supposed to be self-correcting, as Knight and Johnson argue.

A view of “reason” as concerned in its origins with justification-giving rather than “puzzling out the world” fits into this general perspective because it emphasizes that “reason” was always tied to collective deliberation or decision-making. This view of reason as originating as a “social competence” supports the notion that collective decision-making can yield “reasoned” outcomes in the sense defined above. Or at least, it might support that.

A criticism of this perspective, as some of the comments in this thread suggest, is that democracies in practice have not usually worked this way. Maybe the New England town meetings, which Tocqueville admired so much, did, but it’s at least an open question whether once you get beyond that setting they work this way. Possibly some very good local governments still work this way. Politics at the national level, at least in the U.S., does not seem to. Congress agreed to spend billions on infrastructure primarily because members of Congress love having millions of dollars poured into their districts, and without that consideration I’m not sure a process of reasoned collective decision-making would have reached the same result.


William S. Berry 01.08.23 at 4:53 am

I apologize for my embarrassingly stupid comment @2. I barely remember posting it (I was more than half drunk, I think); it was a bit of a shock to read it in this thread!

But, I stand by its “substance”, for the most part. The “logic is meaningless” bit is the stupid part, and requires a bit of interpolation:

Obviously, logic is/ can be meaningful in our use of it in mathematics and analysis (importantly, operations that are projected from, but make a record that is outside of, our minds). I only want to/ meant to suggest that there is no evidence whatsoever that our minds use logic in our everyday, unself-conscious thinking.

Logic involves the use of operators and copulas, evidence of which is completely lacking with respect to our understanding of the work of the mind (what we call thinking).

We have no evidence to suggest that our minds operate in any way other than as a mirror of the world of experience. A more up-to-date version of the theory of “the association of ideas”/ of intuitive recognition of the relations of objects and events (and their associated impressions and ideas), might go some way toward explaining human thought. Beyond genetically instinctual and “racial” memory, there is the epigenetic development of the mind through its interplay with the world of experience. There is a kind of “logic” involved, but it is the apparent logic of cause-and-effect*, contingency, etc., that appears to govern action in the world, and imprints itself on our consciousness.

We’re animals. Without culture, history, memory, we’d be incredibly stupid as individuals. Our existence as social animals with the capacity for learning and yearning is all we have. Human culture must be preserved in order for humans to thrive as a race.

Anyway, it’s a broad and deep subject. I’m still collating. Carry on.

*There is a loud noise, or sudden movement, and flocks of birds ascend. It always happens that way.


TM 01.09.23 at 11:58 am

Alex 25: “I am not a political scientist, but it appears to me that despite its linguistic diversity, it is still easier to build the political culture that can handle plebiscites in small Switzerland than in a large, much more diverse nation like India or the USA.”

It appears to you, but it’s unclear why. Are you aware that plebiscites are very common in the US, just not on the federal level? What do you think about how this works? I find it strange that the strong plebiscitary component in US democracy is so little discussed (see also “a republic not a democracy”). Whether we like it or think it’s a good idea is beside the point, but why isn’t there more attention to this aspect of the US political system?


TM 01.09.23 at 1:29 pm

Alex: “It isn’t just the political culture of UK versus Switzerland, it is so much more. … None of that is different in Switzerland”

The Brexit vote was a political stunt pulled off by the UK government to appease its right flank. That this was even allowed to happen is a problem if the UK political culture and institutional setup. Part of the problem is that the UK doesn’t even have any kind of rules about the employment of referendums, it’s just something that the government arbitrarily decides. The Swiss government, in contrast, could not force a referendum just because they think it’s in their political interest. They have to follow the constitution, after all… You are simply wrong to think that “none of that is different in Switzerland”.

You are also overlooking that the actual chaos ensuing after the Brexit vote wasn’t the result of the vote as such, which was totally non-binding, but of the ineptitude of the UK government and parliament. Again this is a problem of UK political culture and institutional setup.

“my understanding is that the Swiss decided via plebiscite to opt out of freedom of movement with the EU”

I apologize if this discussion is getting way off topic but perhaps this is of common interest. In the 2000s, the Swiss have concluded a series of agreements with the EU, all backed by popular referendums with fairly large majorities, that include the agreement on freedom of movement with the EU (overview at ( paragraph 7)

In 2014, the Right indeed forced and narrowly won a vote on limiting immigration that contradicted the EU agreements but without saying so explicitly. In fact the promoters claimed that the intention was to keep the EU agreements (everybody knows that the Swiss economy would tank big time if they didn’t). It is the case of incoherent outcomes that I mentioned at 20. Given that the democratically decided agreements and the 2014 vote were in contradiction, and that there was no explicit mandate to abrogate the agreements, the government and parliament stated that pacta sunt servanda and the 2014 vote was unworkable. You could see it as the Swiss version of Brexit and I think the way this was handled in both countries is well worth studying (for those interested in such issues). The Right was forced to show its colours by initiating a new proposition this time explicitly calling for abrogating the EU agreements. This came to a vote in 2020 and was defeated by more than 20 points.

Switzerland is not, contrary to common misperception, a direct democracy. It’s a parliamentary democracy with a very unique system of government (a leaderless government of seven equals) that incorporates plebiscitary instruments and has evolved a distinct political culture.


Alex SL 01.11.23 at 9:55 pm


I think the fact that US plebiscites are not at the federal level is the salient point here. That means it works, for the USA, at a smaller scale, just as it seems to work for other nations if they are relatively small. Maybe I am wrong, and it could work just as easily for hundreds of millions of people, but yes, I’d like to see some concrete proof of that. So far, one can only say, odd that they wouldn’t do it at the federal level too. Maybe there is a reason for that.

For “none of that is different in Switzerland”, you are picking on a single point that I did not make and that is irrelevant anyway: how a plebiscite is initiated. What you are leaving out are all the points I made: that plebiscites can’t work well for questions that are ill-defined or leave unclear what exactly is to be done if one of the two options wins, that the ideologically heterogeneous public are more likely than a more ideologically homogeneous and at least potentially (if they listen) expertly advised governing coalition to make incoherent decisions even when every voter is individually rational, that people aren’t all individually rational and expect to get benefits without having to accept the associated trade-offs, and that voters do not have unfettered sovereignty over outcomes that their nation will first have to negotiate with external actors who have greater leverage.

These are just facts about human nature and about how the world works, and they do not change just because you change a nation’s political culture.

As for Brexit itself, I think it is a bit naive to assume that the issue would just have gone away if Cameron hadn’t promised the referendum. The “Eurosceptics” had been a growing concern in the Tory party since all the way back to Major. The media had been spewing lies and hate about how the EU is outlawing church bells or apple trees for decades and weren’t showing any sign of stopping.

If you want to get an idea of that subculture even in 1995, I recommend reading the novel The Aachen Memorandum. Europe is equated with feminism, grade inflation, Hitler, socialism, speed limits, Polish taxi drivers, and everything else the author doesn’t like. Nation states are dissolved into a single European state, but France can still buy the Channel Islands off the UK, although neither exist anymore. There are continental armies in Britain waiting to squash any potential unrest, but (spoiler upcoming) in the end the entire EU collapses within days because a British royal gives a speech, prefiguring how some leading Brexiters would later believe the rest of the EU would dissolve if Brexit went ahead, because, hey, the UK is the main character of every story. The book is stunningly unhinged, hateful, and intellectually incoherent, but the author is a respected academic, a successful author, and a member of the House of Lords. Brexit was not just Cameron making an oopsie.

Yes, Cameron blew his government up unnecessarily, but if something falls over this easily, the problem wasn’t the specific incident that caused it to fall but its structural weakness. Why did Brexit win? Because there was a major constituency of powerful and well-connected anti-Europeans, including most of the media, that would have kept pushing, and if they hadn’t got their referendum in 2016, they may have got one in 2021 or 2025. Really the main question is if the result could have been remain if the referendum had not been held in the middle of austerity; then again, if the result would have been 52% remain, Brexiters had already promised that they would try again as soon as possible.

Thanks for the added detail re Switzerland. I see that reporting outside of Switzerland on the meaning of the first referendum has been simplistic.

Comments on this entry are closed.