1. Introduction
Why do some groups hold beliefs that mostly align with scientific evidence and expert opinion, while others don’t? Why do whole communities buy into young earth creationism, or absurd conspiracies, while living alongside others who reject such patent falsehoods?
One type of answer here adverts to differences among the people in different groups. It says young-earthers are stupid or dogmatic or crazy. This answer favours an individualistic approach to social epistemic policy. It favours policies that try to make people more rationally responsive to evidence – and less stupid, dogmatic, etc. But this approach downplays how similar people’s belief-forming processes are across different communities. As much as some of us try hard to think for ourselves, still, a lot of us, a lot of the time, form our beliefs in a more socially scaffolded way. We align our beliefs with what’s accepted by respected sources in our own communities. Granted, if the respected sources in your community form their beliefs in a way that’s well-tuned to the evidence, it seems better to defer to them, than if you’re in a community whose respected sources form their beliefs in a more erratic fashion. The point, though, is that most of us form our beliefs via social deference, in a way that’s inattentive to these differences. If we are assessing how rational individuals are, in how they go about forming their beliefs, most scientifically literate people are going to receive a similar assessment to most young-earthers, because most of them are, like the young-earthers, uncritically accepting the views of the people that their in-group recognises as authorities.Footnote 1
An alternative approach to social epistemic policy focuses on building a truth-conducive informational environment, e.g. by ensuring that school syllabi, media, and other key nodes in our epistemic networks are disseminating material that’s credible and well-evidenced. The basic idea here is that because there’s little individual variation in epistemic abilities between people in different groups – and little to be gained by trying to improve people’s epistemic abilities – there is a greater potential for veritistic gains by focusing on environmental variables.Footnote 2
This moderate environmental approach to social epistemic policy has a lot going for it. The approach is environmental in that it tells us to prioritise setting up a truth-conducive environment. But it’s moderate, in that it doesn’t totally rule out the possibility of beneficial interventions targeted at individuals. This approach concentrates on environmental interventions, simply because, in a veritistically bad environment, promoting critical thinking, etc. tends to be less efficient in realising our epistemic goals. Environmental ‘clean-up’ has more impact. At any rate, there are very good reasons to think so, given what we know about how powerful people’s cognitive biases are, and how difficult it is for people to evaluate the credibility of different sources that they’re exposed to.
In this paper, we criticise a more radical environmental approach to social epistemic policy, espoused by Neil Levy in Bad Beliefs (2022). Levy’s approach is radical in that it is partly based on a radical view of the rationality of uncritical deference. In Levy’s view, people who defer to in-group ‘authorities’ are in general believing rationally. And this is still the case, he argues, if this uncritical deference leads people to hold beliefs that flout scientific evidence. Contrary to what we ordinarily suppose, it isn’t irrational to uncritically assent to what your community’s favourite preacher or podcaster says. Indeed, for Levy, attempting to critically assess their views is actually an irrational response.
This claim about rational deference can be used to help defend a particularly radical environmentalist approach to social epistemic policy. If young-earthers and their ilk are in fact being rational, then there is nothing to be gained, vis-à-vis the promotion of our epistemic goals, by urging them towards more rational doxastic practices. They are already believing rationally. If we want people to eschew beliefs that flout the evidence, then the only kind of social epistemic policy that it makes sense to pursue is one that seeks to depollute the information environment.
The reason we want to criticise Levy’s account isn’t that we reject an environmental approach to social epistemic policy, but because we endorse it. Our worry is that Levy’s radical ideas about deference end up discrediting an environmental approach. How? First, by saddling it to independently unattractive claims about the nature of rationality. Second, by subverting some of the judgments that are involved in setting up the sort of truth-conducive environment that Levy himself endorses.
So our aim here isn’t simply to critique uncritical deference. We’re challenging the idea that epistemic environmentalism and radical deference norms come as a package deal. The environmental approach has considerable appeal. It recognises that individual epistemic conduct is shaped by structural factors, and prioritises interventions that target those factors. Nothing about this approach requires us to endorse Levy’s radical claims about the rationality of uncritical deference. There are no entailment relations between the claim that “it’s rational to uncritically defer to in-group authorities” and the claim that “providing true information is more effective than promoting critical thinking skills for achieving collective epistemic aims.” Nor is there any other compelling reason why these commitments have to stand or fall together. If our goal is promoting true beliefs, we should favour both information policies and belief-formation norms that conduce to that aim. Levy’s error lies in thinking that his uncritical deference norm better serves this aim than alternative, more nuanced – perhaps more boring – approaches to navigating between deference and discrimination. Disentangling these elements of Levy’s view helps clarify the key attractions of epistemic environmentalism.
The plan for the paper is as follows.
-
In §2 we continue the stage-setting. We explain Levy’s notion of Bad Beliefs, we say a bit more to characterise and motivate the environmental approach, and we explain Levy’s argument about why Bad Beliefs formed via social deference are rational.
-
In §3 we present our first objection – that Levy’s account saddles the environmental approach with a view of rational deference that’s independently unattractive. We start by explaining just how easy Levy makes it to qualify as rational when you defer to a putative in-group authority. We then consider how Levy distinguishes rational and irrational bases of deference, and argue that his way of drawing this distinction doesn’t rectify the initial worries about ‘easy rationality’ for in-group deference.
-
In §4 we present our second objection – that Levy’s radical account of rational deference discredits some of the critical thinking necessary to create the truth-conductive environment that the environmental approach calls for. To create that environment we need a significant number of people to adopt critical, questioning, un-deferential attitudes towards authority – the very attitudes whose rationality Levy is denying.
2. Stage-setting
2.1. Bad beliefs and the structural perspective
Bad Beliefs, as Levy defines them (2022: xi), are beliefs that (i) conflict with what domain-relevant authorities believe, and (ii) flout the publicly-available evidence. Footnote 3 Paradigmatic examples include the belief that climate change isn’t a real phenomenon, the belief that Barack Obama isn’t a US citizen, or the belief that the earth is only a few thousand years old. There’s plenty of evidence for the falsity of these views, and relevant experts agree they are false. Yet many people hold these beliefs nevertheless. Those are Bad Beliefs.
It seems natural to view Bad Beliefs as straightforwardly irrational. If there’s a weight of credible evidence that’s been made available via the media, via online sources like Wikipedia, or in academia, then we presumably have some rational duty to align our beliefs with what that evidence shows, especially when relevant experts digest and explain that evidence for us, offering clear advice about what to believe.Footnote 4 Believing Obama was born in Africa, or that climate change is a conspiratorial hoax, in the face of all that, seems like a paradigm case of irrationality.
Levy challenges this way of thinking about Bad Beliefs by challenging the individualistic framing that it presupposes. For Levy, Bad Beliefs are a structural problem – they can’t be effectively understood or addressed by focusing on individuals’ doxastic practices.
To analogise: the reason why more people are obese nowadays, in many countries, isn’t that individuals have become less ‘dietarily virtuous’. It’s that economic, technological, and political factors are producing an obesogenic environment. If we want to reduce obesity, we have to rebuild that environment, e.g. by limiting the easy availability of cheap, addictive, unhealthy foods, or increasing the availability (and/or affordability, and/or appeal) of healthier foods. It’s not illuminating to fixate on individual choices downstream from more potent structural factors.Footnote 5
Similarly, to reduce the frequency of beliefs that flout evidence and expert opinion, we need to rebuild our information environments, so that people aren’t inundated with misinformation and have reliable default information sources. Just as it’s largely futile to urge people in an obesogenic environment to be more dietarily virtuous, it’s also largely futile, in a veritistically bad environment, to urge people to be more epistemically virtuous.
The crux of the environmental approach isn’t to deny the value of basic education. It’s to resist the idea that non-truth-conducive environments can be effectively counteracted by promoting individual reasoning skills like critical thinking or digital literacy. Levy associates such individualistic approaches with virtue epistemology, and he treats Quassim Cassam (Reference Cassam2018) as his foil and exemplar of this approach. But the approach Levy is criticising isn’t just an epistemologist’s contrivance. It shows up regularly in political theory and public policy work on content moderation and online governance. It’s commonplace, in that literature, to see people talking about the new norms of online discourse that people must adopt, in order to navigate around echo chambers, misinformation, and related epistemic hazards that seem to become heightened online (e.g. Chambers Reference Chambers2021; Cohen and Fung Reference Cohen, Fung, Bernholz, Landemore and Reich2021). The environmental approach regards such interventions as mere band-aids that treat surface-level symptoms while ignoring root causes.
2.2. The rationality of bad beliefs
Bad Beliefs may be a structural issue, calling for a structural solution. But isn’t uncritically believing a podcaster or preacher, against expert opinion and credible evidence, still irrational? Even if obesity is produced by obesogenic environments, eating lots of sugary, ultra-processed food is still unhealthy. And even if Bad Beliefs are produced by polluted environments, accepting misinformation still seems irrational. After all, isn’t it just paradigmatically irrational to believe things that the relevant evidence disconfirms?
Yes and no. Bad Beliefs come from trusting putative authorities who misjudge the evidence due to corruption, bias, or delusion. But even so, Levy argues, deferring to them is part of a practice that, when iterated at scale, conduces to collective epistemic success. We most effectively acquire epistemic goods via simple deference, or “outsourcing”.Footnote 6
Levy’s argument for this draws on evidence of how humanity’s epistemic success is a consequence of cumulative culture. In contrast to the cultures of other intelligent animals that can solve complex problems and engage in sophisticated communication,
Only [human beings’] cumulative culture builds on these innovations, enabling cognitive achievements that go beyond what any individual or any generation can achieve… Cumulative culture opens up horizons for knowledge that are closed to individuals, no matter how individually gifted they are. [40]
The most striking empirical findings, in this connection, and those most congenial to Levy’s account, find that humans are more disposed than other primates to engage in apparently pointless imitative behaviour. Footnote 7 This supports our intellectual success by making it easier for us to intergenerationally transmit knowledge, even in cases where the recipients can’t yet grasp why the transmitters believe what they’re transmitting. If humans are so smart, Levy asks, then how can other primates “outperform us in identifying efficiencies and more successful routes to a goal?” [44]. The answer is: because our culture is extra-imitative.
Imitation is an adaptation for culture. It allows us to acquire knowledge and practices developed by… individuals dispersed across space and time. It allows us to acquire, and then to build on, deeply social knowledge: adaptive behavior that could not have been developed by any individual de novo. [44]Footnote 8
Solo inquiry yields less than when we instead defer to the prevalent (often inherited) judgements of others around us. In Levy’s view, this kind of outsourcing is entirely rational: “being open to cues for what others believe is being open to reasons” [77].Footnote 9
So why isn’t believing against evidence paradigmatically irrational? Generally, relatively uncritical deference conduces to collective epistemic successes by enabling cumulative culture. We might prefer to call this rational irrationality [153], or ecological rationality [142–43] – locally irrational deference that yields success. But Levy provocatively suggests that uncritical deference is rational all the way down, denying any sharp distinction between ecological rationality (what conduces to collective success) and individual rationality (what’s locally responsive to evidence) [73]. A climate change denier, who uncritically defers to a podcaster or politician who is viewed as an authority within that person’s (the denier’s) community, is rational. Their belief is bad, in flouting evidence, but rational, in being duly responsive to the social evidence, enabling cumulative culture’s long-run epistemic success.Footnote 10
3. Against uncritical deference
Levy argues – forcefully, creatively – for an environmental approach to social epistemic policy. By our lights, though, the environmental approach is undermined by being saddled to Levy’s account of rational deference. The prima facie worrying issue with his account is that it makes it too easy to count as rational in deferring to others. The deeper issue is that Levy’s revisionist conception of rationality – which collapses the conceptual gap between what’s rational for the individual to believe, and what conduces to a group’s collective epistemic success – doesn’t vindicate his permissive view of rational deference. It doesn’t support the controversial claims that are meant to differentiate Levy’s social epistemic policy prescriptions, from the prescriptions that are favoured by more individualistically-minded epistemologists.
Levy thinks it’s rational for Christians to believe in a young earth, for instance, because such belief results from relatively uncritical deference to local authorities. He sees this as rational because it conduces to our species’ collective epistemic success. But in fact uncritical deference ought to be adjudged irrational by Levy’s own standards. Why? Because indiscriminate deference tends to set back a group’s collective epistemic success, and there’s no way to differentiate relatively uncritical deference from indiscriminate deference – not without appealing to the kind of individualistic norms of critical thinking whose rationality Levy wants to deny.
3.1. Uncritical deference
So, the basic worry: Levy’s account makes it too easy to count as rational in deferring to in-group authorities. He thinks the young earther isn’t holding this belief in spite of evidence against it, but that she’s holding it on the basis of supporting evidence – the evidence constituted by the fact that it’s affirmed by her in-group’s local authorities.
As a starting point, notice how this places a lot of justificatory weight on one source of evidence, and very little weight on other sources. To paraphrase Alex Worsnip (Reference Worsnip2022),
To get the result that young-earthers are believing rationally, Levy needs to explain why young-earthers are rational in giving so much weight to the testimony of other young-earthers. Young-earthers are aware that lots of people aren’t young-earthers – including many people who are part of their “community”, in some sense. So Levy’s claim has to be not just that we’re rational in deferring to the community, but that we’re rational in deferring to our co-partisans, even when we’re well-aware that many people outside this immediate community disagree.
The worry here should be familiar. Groupthink seems clearly epistemically bad, and Levy’s account of rational deference seems to license it. Even if Levy is correct that outsourcing is beneficial (when iterated at scale), it doesn’t follow that chauvinistic or parochial outsourcing is beneficial. Indeed, some of his own examples, whose point is to show how outsourcing is rational, also serve as illustrations of why chauvinistic outsourcing isn’t rational. Consider his example of the 19th-century colonial explorers who dismissed life-saving dietary knowledge offered by indigenous people [37]. The lesson Levy draws is that we do better, epistemically, by outsourcing, in the way these explorers didn’t. But the explorers were outsourcers, in general – they wouldn’t have been able to embark on their endeavour in the first place if they hadn’t relied on all manner of testimonial knowledge about, e.g. navigation. They missed out on vital dietary knowledge not because they failed to outsource generally, but because they refused to outsource to out-group members – hence their plight does not seem very different from that of young earthers.
Outsourcing is rational, for Levy, because it enables the accumulation of knowledge via cumulative culture. Constantly trying to critically audit the reports or opinions of apparently authoritative people in your in-group isn’t a fruitful approach to belief formation, when it’s iterated at scale. We should endorse the rationality of relatively uncritical deference, Levy argues, insofar as this belief-forming heuristic conduces to the epistemic benefits of cumulative culture.
But the devil is in the detail of the italicised caveats – relatively, and insofar as. Outsourcing as practiced by the chauvinistic colonial explorers cut them off from knowledge. In order for Levy’s account of rational deference to work, it needs a criterion that can be used to identify overly uncritical, chauvinistic, or parochial, forms of deference – the forms that are tantamount to pernicious groupthink, and which inhibit collective epistemic success. The criterion has to differentiate these counterproductive forms of deference from the preferable, only relatively uncritical forms of deference – the ones that conduce to collective epistemic success. This criterion can’t just recapitulate the ideals of critical thinking and epistemic virtue in opposition to which Levy is pitching his account.
3.2. Modesty and discrimination
Levy has a story to tell here, but we don’t think this story can do the work that’s needed. (We will focus on exposition in this sub-section, and then return to criticism in §3.3.)
Solo inquiry isn’t an effective practice – long-term, at a community level – for realising epistemic goods. So we should outsource. But there’s still good and bad outsourcing, by Levy’s own lights. Rational deference requires us to engage in processes of discrimination: seeking to differentiate the deference-worthy authorities from non-authorities. Thus Levy’s account has to answer the question of how novices are to assess expertise – building on recent epistemological work on this issue, whether of a bent which is more optimistic (e.g. Goldman Reference Goldman2001; Anderson Reference Anderson2011), or more pessimistic (e.g. Millgram Reference Millgram2015).
Levy’s proposal here calls upon a familiar, attractive norm: we should respect our intellectual limitations. We aren’t being rational if we’re trying to carry out intellectual tasks which outstrip those limitations. Let’s call this norm, as loosely defined, modesty.Footnote 11
So, which grounds for adjudicating expertise are and aren’t rational? The modesty norm says we have to respect our intellectual limits. Some discriminations are easy to get right; others are far harder. Schematically, modesty tells us it’s rational to make the easy discriminations, but, generally, irrational to attempt the harder ones. The easy discriminations are the ones where we’re just looking for markers of expertise. The idea isn’t that it’s easy to judge who the real experts are in a given domain, i.e. who has domain-relevant knowledge and complementary intellectual abilities. Indeed, those are the immodest discriminations that we’re meant to avoid, which Levy’s account sees as irrational. The idea, rather, very simply, is that it’s usually easy to judge who the putative experts are, in a group or institution that you’re a part of. Consider university students. They don’t find it hard to work out which people are the professors, and which people are administrators or security staff. The university’s institutional format makes these discriminations easy. And so, apart from rare mistakes – which are easily correctible – students reliably identify whom the experts are that they’re meant to be learning from.
The hard discriminations, by contrast, are those where one is trying to judge manifestations of expertise, as per the conundrum Goldman (Reference Goldman2001) sets out in his Novice/2-Expert problem. Suppose Lia takes a course in evolutionary biology, where she’s told by Professor Dawkins that the central tenets of evolutionary biology are basically incorrigible and that this is why they’re accepted by every expert who understands the relevant data. The evidence vindicates those tenets. Suppose Lia then takes a course in philosophy of science, where Professor Feyerabend tells her that biologists accept evolutionary theory not (or not only) because the evidence supports it but as a result of social processes that are strikingly similar to processes of cultural habituation or indoctrination.
How should Lia discriminate? It’s easy to identify markers of expertise, but both Professors have those markers. What she needs to do is to assess their manifestations of expertise. And that seems nearly impossible. Unless Lia has a way to review their track records of predictive success, or identify biasing factors that undermine the reliability of one of them – unless she can do these things armed only with her novice’s understanding of the data and methods that define their respective fields – she seems to have no good way to discriminate. Footnote 12 Maybe if a lot goes well, over many years, Lia can eventually acquire enough expertise to rationally judge whether Dawkins or Feyerabend is doing better at understanding and interpreting the evidence that’s pertinent to their dispute. But for most of us, most of the time, the attempt to make such discriminations will, in terms of its rational pedigree, be tantamount to making a random stab in the dark.
This part of Levy’s thinking is summed up in his commentary on Cassam’s views about how epistemic virtues can help people resist misinformation and conspiracies. Levy agrees with Cassam that we may “deploy the virtues to choose between competing experts” [103]. But Levy puts a qualifying asterisk next to this acknowledgement. Insofar as Cassam’s prescription calls for individuals to “directly and virtuously adjudicat[e] the second order evidence”, i.e. the evidence about which expert is actually more reliable, “it’s still too demanding and too individualistic.” Thus, Levy thinks, “we face the same risks of losing knowledge by engaging at this level,” and hence, “dogmatism remains a better strategy” [103].
But surely this is incredibly risky, isn’t it? To say that dogmatism is the better strategy – better in the sense of mitigating our risk of losing knowledge? Levy holds firm. Such outsourcing “remains a better strategy”, where the strategy at issue, specifically, involves “relatively unquestioning deference to authoritative sources because they’re authoritative, and not because we’ve assessed their degree of expertise ourselves” [103, our emphasis].
3.3. The unquestioning stance
The novice can judge markers of expertise, but they don’t have the expertise to directly judge manifestations of expertise. This is why relatively unquestioning deference to authoritative sources is the rational approach to outsourcing, on Levy’s account. To attempt to defer only in a critical, discriminating fashion, is to infringe against the norm of modesty.
The problem we were pressing in §3.1 remains, though. If the putative authorities disagree, and if we defer to our preferred authorities uncritically, we are undermining the epistemic benefits of collective culture, just as surely as if we immodestly attempted to judge the experts’ manifestations of expertise. Under both approaches, we are deciding whom to defer to in a way that’s tantamount to reckless guessing. If what Levy’s approach to discrimination involves, in the end, is that you can ascribe authority to whichever view or idea happens to chime for you, then discriminating deference ends up being no better – with respect to our collective epistemic goals – than hubristically trying to figure everything out yourself. Footnote 13 And Levy’s account seemingly isn’t giving us the resources to do better than this.
To illustrate the worry, consider a follower of a doomsday cult, who has been told by their leader that the world will end in some sort of spiritual-celestial cataclysm, on a specific date. Suppose the leader is clever, knowledgeable, and charismatic – they have a mixture of social traits that mark them out as an authority within their community. And suppose that lots of smart people (teachers, lawyers, financial advisors) are deferring to the leader.
Now suppose the foretold world-ending date goes by, and the cult leader has concocted an after-the-fact rationalization that explains why the doomsday event didn’t happen on that date, and why it’s actually going to occur in twelve months’ time. In order to give the example a ‘two rival experts’ format, let’s suppose that our follower – who is now vaguely worried that they’re being deceived – goes to ask another apparently credible person whether they think the world is going to end on this newly appointed date, in a year’s time. And suppose this other putative expert says precisely what we would expect: “no, the world isn’t going to end then, you’ve fallen into some sort of cult.”
Recall that Levy’s account of what it’s rational to do, in such situations, is to evince relatively unquestioning deference to authorities because they’re authoritative – not because you’ve assessed their intellectual credibility directly. But of course, there are two putative authorities in the frame of reference here, so the fact that one is an authority cannot settle the question of which one should be (relatively unquestioningly) deferred to.
Intuitively, it seems clear what this person ought rationally to do. They should see that someone who rearranges their worldview impromptu, and produces post hoc rationalizations for why they’re doing so, is probably a liar or bullshitter – at any rate, not the reliable expert they’re making themselves out to be – and accordingly, stop deferring to that person.
If the follower reasons like this they will be adjudicating expertise in a manner that doesn’t neatly fit into the markers vs. manifestations schema. The cult leader may present a bunch of complicated math or astronomy, to try to explain why their prediction of the apocalyptic date shifted. The follower may have no ability to assess that putative evidence. The follower’s growing mistrust of the leader isn’t because they have spotted granular faults in the leader’s reasoning. But equally, the leader may have all of the social markers of expertise that they had previously. So the follower’s growing mistrust isn’t stemming from a shift in the leader’s social markers of expertise. The follower is just perceiving that things don’t add up. The best explanation of the pattern of reasoning and testimony that they’re observing from the leader isn’t that the leader is a real expert who’s telling the truth. The best explanation is that they’re a smart person who has, for whatever reason, become a zealot on behalf of a belief system that doesn’t line up with reality.Footnote 14
People make these kinds of discriminations fairly often. They obviously aren’t foolproof, and it’s possible that some more individualistically-minded virtue epistemologists overestimate the median person’s ability to step into this kind of gestalt critical vantage point and to make sound judgements upon doing so. The modesty norm might be a sensible corrective to that overestimation. Nevertheless, people seem to make these discriminations, successfully, with some regularity. More to the point, for Levy’s purposes, the execution of these kinds of discriminating judgements seems to be conducive to the realization of cumulative culture’s epistemic benefits. True, we reap some collective epistemic rewards from a tendency towards unquestioning deference. But those rewards would be jeopardised if this imitative tendency were not partly counterbalanced by some preparedness to question and demur.
To be fair, Levy never endorsed a stance of purely unquestioning deference to authority. He endorsed relatively unquestioning deference. So then the issue is: what does that caveat permit? How close does it bring Levy’s account to the individualistically-minded epistemologists that he’s criticising – the ones who think people should use critical thinking to adjudicate between experts? At a more concessive moment in his argument, Levy says
If virtue epistemology can help… it’s not by substituting for apt deference to others and socially distributed cognition; instead, it’s by playing a (small) role in helping us to do these things better. Virtue epistemologists… appear to aim to bring us each to inculcate the virtues in ourselves and then… to tackle hard problems largely on our own. [91]
Insofar as virtue epistemologists are telling us to eschew collaborative learning, and to figure out everything ourselves, they are liable to criticism. But as Levy’s discussion shows, the real locus of disagreement between him and his opponents isn’t whether to outsource, but when and how to outsource. The clear point of difference between them, besides subtleties of emphasis (Levy believes the role of individual virtue is small; his opponents presumably see its role as larger), is that Levy wants to say – and his opponents generally want to deny – that it’s rational to defer to putative authorities because they are putative authorities.
Levy’s stance on this looks bad in the cult case. The fact that the leader is a putative authority isn’t a good reason to defer to him – not when there are other experts who demur, when the evidence doesn’t support his claims, and when there are grounds for questioning his judgement. This is what individualistically-minded epistemologists are urging us to do: to question authority when there are grounds for doing so, rather than taking the fact that someone is an authority among our in-group as a sufficient reason to defer to them. The person who believes climate change is a hoax just because their favourite podcaster said so is being too credulous. If that extreme level of unquestioning deference counts as rational, then rational deference counts for little. It loses the theoretical link to cumulative culture and epistemic success that was supposed to be theoretically underpinning Levy’s whole proposal.
4. Structural solutions
Our second criticism is that Levy’s view about the rationality of uncritical deference undermines his own environmentalist approach to social epistemic policy. The environment we want is one where it’s easy for people to believe true, well-evidenced things, largely by going along with what they are told. Concretely, Levy argues for improving our norms of scientific publishing and media coverage to discourage disseminating biased or spurious findings [125–31]. But the policies he’s endorsing to this end require us to discriminate between credible and non-credible expertise.Footnote 15 This discrimination, when done successfully, seems like a paradigm example of rational judgment. It involves carefully evaluating the track records, methodologies, and incentives of putative experts to determine their reliability. Levy’s account, which treats deference as rational merely because it accedes to the views of in-group authorities, seemingly cannot credit the rationality of this type of discrimination.
One might try to defend Levy’s position by carefully distinguishing several questions about epistemic norms and epistemic environments (thanks to a reviewer for suggesting this framing):
-
(i) How should individuals form beliefs when aiming at knowledge?
-
(ii) How should communities collectively form beliefs when aiming at knowledge?
-
(iii) How should individuals work to improve the epistemic environment?
-
(iv) How should communities work to improve the epistemic environment?
The thought would be that Levy’s radical deference principle applies only to questions (i) and perhaps (ii) – it tells us how beliefs should be formed within a given environment. When we turn to questions (iii) and (iv) about improving the environment itself, different norms might apply. On this view, there’s no contradiction in saying both that individuals should generally defer to authority when forming ordinary beliefs, and that some individuals need to exercise critical discrimination when working to establish which authorities are credible.
But this attempted defense actually just highlights the core problem with Levy’s account. The problem isn’t that Levy fails to distinguish these different questions – it’s that his defense of uncritical deference makes it difficult to maintain these distinctions in a principled way. If uncritical deference to perceived authorities is rational simply because they are perceived as authorities, then this norm would seem to apply equally to all contexts of belief formation. There would be no principled basis for saying “defer uncritically here, but think critically there.” The very attempt to carve out special contexts where more discriminating norms apply would require the kind of critical thinking whose rationality Levy’s account is purporting to deny.
What’s needed instead is precisely what this objection gestures at: an approach that recognises different epistemic norms operating in different contexts, in an ecosystemic way. But embracing such an approach means abandoning Levy’s far-reaching defense of the rationality of uncritical deference. The lesson here isn’t that Levy should have more carefully distinguished different epistemic questions – it’s that making such distinctions requires rejecting his central claim about the rationality of deferring to in-group authorities simply because they’re authorities.
Restructuring the epistemic environment to make default sources of information accurate necessitates a massive program of institutional discrimination. Countless judgments need to be made, and continually re-made, about which authorities and sources are credible. These judgments have to be responsive to evidence about the reliability of those sources. The people making the judgments can’t do this while uncritically deferring to authority. Instead, we need some reasoners capable of reliably discriminating the comparative expertise of putative authorities, even if they are not domain-relevant experts themselves. They need the critical thinking skills to spot potential biases, conflicts of interest, or methodological flaws that could compromise an authority’s credibility.Footnote 16
Part of what is going awry in Levy’s picture is that, having equated rationality (in matters of deference, discrimination, and critical thinking) with what conduces to collective epistemic success, he doesn’t put sufficient emphasis on the ways in which collective success can benefit from heterogeneity in people’s belief-forming practices. Footnote 17 We have significant evidence that a diversity of approaches and perspectives is required to optimise epistemic performance across a group.
To illustrate, consider a well-studied tradeoff in designing intelligent learning algorithms: the “explore-exploit” tradeoff (Cohen et al. Reference Cohen, McClure and Yu2007; Hills et al. Reference Hills, Todd, Lazer, Redish and Couzin2015). In a noisy environment where we can’t be sure that we have learned the entire truth, an epistemic agent must decide how to allocate their time between using the evidence they have already acquired (exploiting) versus engaging in experimentation to devise better beliefs (exploring). The optimal balance will depend on the agent’s uncertainty, and on the risks and rewards associated with new discoveries. Too much exploitation risks stagnation; too much exploration wastes resources on fruitless investigations. In a social context, the tradeoff becomes more complex, as an individual’s best strategy depends on what others in the community are doing. If everyone else is conformist exploratory critical thinking will be more valuable. But if others are already pushing the boundaries then exploiting existing knowledge may be better.
In a social environment where agents can learn from either social sources or from their own observation, the problem becomes even more complex again. In addition to intra-personal tradeoffs, there are now interpersonal tradeoffs. If the community already has many agents engaged in exploration, it may be best to simply watch and learn from them; and conversely if the community is highly conformist and only exploiting past discoveries, it may be best to undertake some experimentation.Footnote 18
The contrast between uncritical deference and independent critical thinking mirrors this explore–exploit tradeoff. The optimal balance, for the collective, between these two modes is not something that can be determined a priori, independent of actual context. Surely not everybody needs to critically examine every issue for themselves – that way lies a cacophony of idiosyncratic views. But nor is uncritical deference the answer – that runs a high risk of stagnation and dogmatism. A healthy epistemic community needs a mix of strategies, with some deference to keep the group anchored, and some independent questioning to keep it responsive to new evidence.
Levy might argue that even if some critical thinking is good for the collective, it could still be irrational for the individual. Some epistemic models show that individual and collective incentives can diverge, leading to epistemic tragedies of the commons (Mayo-Wilson et al. Reference Mayo-Wilson, Zollman and Danks2011; Zollman Reference Zollman2020). In these cases, individuals motivated purely by self-interest adopt strategies that are collectively suboptimal. Perhaps critical thinkers are like this: irrational martyrs sacrificing their own epistemic welfare for the greater good. But this argument requires abandoning Levy’s central claim that rationality is tied to collective success. The point of his account was to shrink the conceptual distance between individual rationality and collective optimality. Conceding that the two can come apart, far from salvaging his view of rationality, puts pressure on its core thesis.
5. Conclusion
The attractive features of epistemic environmentalism don’t require us to posit a tight link between individual epistemic rationality and group-level success. The core of the case for the environmental approach is that deferring to local authorities will always be an efficient means of transmitting knowledge, so the best strategy to promote true beliefs is to ensure those authorities are espousing true and well-evidenced beliefs.
Levy gives a supercharged justification for this environmental approach: if uncritical deference to local authorities is rational, then individualistic interventions seem totally futile. There’s little to be gained trying to help people be rational social epistemic agents if people’s natural impulse to defer to in-group authorities is already rational. But the environmental approach doesn’t need such a supercharged justification. Individualistic approaches can still be wrongheaded, even if they aren’t utterly futile. As we see it, there is something to be gained in trying to help people be more rational, it’s just that there’s generally more to be gained – vis-à-vis the realisation of our collective epistemic aims – in prioritising environmental interventions.
We have argued that Levy’s radicalism is not just unnecessary, but counterproductive. In §3 we argued that it ties the environmental approach to an unattractive view of rational deference – endorsing uncritical deference that actually hurts individual and group epistemic aims. In §4 we argued that it undermines some of the judgements that are involved in the construction of a favourable information environment. Some non-experts need to critically assess expertise, and doing this should, contra Levy’s view, be credited as rational, provided that it’s done in a way that’s critically minded and consistently responsive to evidence.
Acknowledgements
Thanks to the editors and a referee for helpful feedback on an earlier version of this paper. Thanks also to Michael Hannon and Elise Woodard for the discussion on this topic. The authors’ work on this paper was supported by an Australian Research Council Discovery Project (DP190100041) on ‘Governing the Knowledge Commons’.
Open access
