In this journal, Joshua Blanchard (Reference BlanchardForthcoming) has recently identified a problem he calls the problem of unwelcome epistemic company. We find ourselves in unwelcome epistemic company (at least typically) when we realise that some controversial belief of ours is shared mainly or most prominently by people we regard as abhorrent. People who are sceptical about the value of masks to prevent respiratory illness (to use a topical example) might feel discomfort at the realization that this belief is most prominently espoused by fervent Trump supporters, if they are opposed to Trump and his views. Blanchard argues that this discomfort is justified: unwelcome epistemic company often provides higher-order evidence against our beliefs.
To my knowledge, Blanchard is the first to identify and discuss the problem of unwelcome epistemic company. His paper is therefore a significant contribution to epistemology. I aim to build on Blanchard's achievement. His programmatic paper does not provide us with either a test for when company is unwelcome or with a diagnosis of what makes it unwelcome. Focusing on the defeaters of the sort that are of central interest to epistemologists – alethic defeaters – I will provide both. I will show that unwelcome company provides higher-order evidence against the truth of our belief that p when it provides evidence that either our belief has the same content as a belief that expresses some of the properties that make our company unwelcome or that it was formed by an unreliable process (or, of course, both).
1. The need for a test
Worries about unwelcome epistemic company are familiar, often appearing in the form of an accusation (“you know who else believes that?”). As Blanchard notes, however, while the accusation is common, the fact that we have unwelcome company isn't always a reason for concern about the normative status of our beliefs. Consider these cases (Blue Sky, Refugees and Hitler are rewritten cases borrowed from Blanchard; the others are my own):
Blue Sky. Shi believes that the sky is blue. He subsequently reads that Osama Bin Laden also believed that the sky is blue.
Hitler. Arkady becomes a vegetarian to prevent suffering to animals. Her friend forwards her a link to a blogpost that alleges that Hitler was a vegetarian.
Drug. On the basis of reading an article in Newsweek, Seanan believes that a certain drug is an effective treatment for Covid-19. She subsequently discovers that Donald Trump has been promoting this drug for just this purpose and his followers have enthusiastically endorsed it.Footnote 1
Eggs. Liu argues that eggs should be stored in the fridge for longevity, against the passionate advocacy for the pantry of his friends. Liu subsequently discovers that the leader of a local white supremacist militia shares his belief about eggs.
Lockdowns. Jo believes that lockdowns to control COVID-19 are unjustified, on the basis that the economic harms lockdowns cause will result in a greater loss of life than the virus itself. But she is uncomfortably aware that this puts her in unwelcome epistemic company: the great majority of lockdown sceptics are also vocal Trump supporters, and she has no sympathy at all for their views.
Refugees. Cory believes refugees commit more crime than legal immigrants or those born in the country. He's upset to discover that this puts him in the company of white nationalists.
Blanchard claims that Blue Sky and Hitler should not trouble those who find themselves in this unwelcome company. But he doesn't provide an explanation for these judgments. We need such an explanation, to be able to identify when company is troublesome and when it is not. We need such a test because there are hard cases, in which our intuitions give us little guidance, and because there is disagreement even about some of the cases Blanchard takes to be clear. Should feminists who oppose sex work be concerned that they find themselves in the company of people whose conservative views they reject? Intuition does not provide us – or at any rate me – with a clear answer to this question. Is Blanchard right in thinking that vegetarians should not be troubled by their (apparent) agreement with Hitler? Many vegetarians and vegans do seem to find it troubling: I infer this from the fact that the most common response isn't to shrug off the bad company, but to deny that Hitler shared their belief at all. One advocate for vegetarianism went as far as to write a whole book devoted to debunking the claim (Berry Reference Berry2004).
2. Towards a test
When does unwelcome company provide higher-order evidence against a shared belief? Blanchard treats unwelcome company as the mirror image of peer disagreement. In a peer disagreement case, the fact that my epistemic peer comes to a conflicting conclusion to me is higher-order evidence that at least one of us is mistaken; many philosophers think we should reduce our confidence in our belief in such cases, because we have no more reason to think that our peer is mistaken than ourselves (Christensen Reference Christensen2007; Matheson Reference Matheson2015). In a disagreement case, the reliability of my peer provides me with higher-order evidence of error. In unwelcome company, the unreliability of my company, coupled with their agreement with me, once again provides me with higher-order evidence of error.
Not all unwelcome company provides us with even prima facie evidence against our beliefs. But that, too, is something it shares with peer disagreement. If my peer dissents from my verdict that 2+2 = 4, I have evidence to question his sanity or sincerity, rather than my arithmetic. Cases like this motivate Lackey's influential ‘justificationist’ view (Lackey Reference Lackey, Haddock, Millar and Pritchard2010). Blanchard notes that we are often able to invoke our justified confidence in our belief to dismiss unwelcome company. Perhaps Blue Sky is a case of this kind.
But unwelcome company may fail to provide evidence against a belief even when the agent lacks a particularly high degree of justified confidence in their belief. Consider Eggs. Liu's belief is controversial and he knows it. It might well be controversial enough to be subject to confidence-lowering peer disagreement; that is to say, justificationism provides no shelter to it. Nevertheless, the unwelcome company he finds himself in does not provide him with evidence against his belief.
Why not? Here's the beginning of a positive proposal: unwelcome epistemic company provides higher-order evidence against a shared belief when the unwelcome company's belief is counterfactually dependent on the very properties that make them unwelcome. To see the attraction of this proposal, contrast Eggs with Lockdowns. Suppose the cases as similar as possible in their epistemic features. Both beliefs are controversial and known to be so to those who hold them (suppose that Liu and Jo are each the only person in their group of friends who holds the belief). They are taken by each agent to be supported by the same proportion of relevant experts and are held to be justified to the same degree. Holding all these other features fixed, we can isolate the effect of the unwelcome epistemic company. Liu, I suggest, will not be concerned by his company whereas Jo will be worried by hers; plausibly, differences in whether their company's respective beliefs are counterfactually dependent on the very properties that make their company unwelcome explains the difference.
It is implausible that the properties (beliefs, vices, character traits) that make the company of the leader of the white supremacist militia so unwelcome play an important explanatory role in why he believes that eggs should be stored in the fridge for longevity. The content of this belief is too far removed from these properties to play such a role (matters might conceivably be different if he preferred white eggs to brown eggs). But it is very plausible that most, or the most prominent, lockdown sceptics hold that belief due to the properties that make their company unwelcome. They might hold their belief because they value the economy more than people's lives, or because they fail to care sufficiently about the deaths of older people (for example). Their belief is plausibly counterfactually dependent on the beliefs or values that Jo finds abhorrent in them. In other cases, a belief might be counterfactually dependent on other deplorable properties: epistemic vices (e.g. epistemic arrogance, manifesting in a failure to give proper regard to the opinions of experts), biases like the confirmation bias or something more visceral: for example, a desire to signal contempt for liberals. All of these are properties that make epistemic company very unwelcome to people like Jo.
Why would the fact that the unwelcome company's belief is counterfactually dependent on the properties that make their company unwelcome provide higher-order evidence against the belief? Belief formation and update is not transparent to agents like us. Our values and priors and traits are not easily open to inspection. When we recognize that a belief we share is in other agents counterfactually dependent on properties we find abhorrent in them, we have a reason to suspect that it is the product of such properties in ourselves. Jo may be concerned that she doesn't value older people as much as she should or as much as she thought; that she doesn't give due weight to expert opinion, and so on. Her degree of justification can make a decisive difference here: if Jo has a right to high confidence that her belief is justified independently of these properties, she can shrug off the unwelcome company. With controversial beliefs, we often can't be very confident of facts like these, and unwelcome company provides higher-order evidence against our belief.
However, the counterfactual dependence account is unable to explain all our cases. Consider Drug. Donald Trump's more fervent supporters share Seanan's belief, and their belief is counterfactually dependent on the very properties that make them abhorrent (by her lights). Were they not fervent Trump supporters, they wouldn't believe that the drug can cure Covid. But Seanan shouldn't be concerned by her unwelcome company, because it does not provide her with good reason to suspect that her belief is the product of the properties that make her company unwelcome. Her unwelcome company endorse the drug because Donald Trump does (and independent of any other properties it may or may not have). She (we're supposing) was unaware that Trump endorsed it when she formed her belief.
Counterfactual dependence is a good heuristic for when unwelcome company provides us with evidence against a belief, because a belief's dependence on the properties that make the company unwelcome is correlated with that belief expressing those properties. Cases like Drug show that this correlation sometimes fails to hold.
Seanan seems to be in the clear because, though her company's belief is counterfactually dependent on some of the properties that make them unwelcome, those properties are not manifested in the belief content. There's nothing intrinsic to the drug that makes it attractive to Trump supporters, after all, and had he endorsed a different one they'd embrace it instead. Cory, on the other hand, should feel discomfited at finding that white nationalists share his belief that migrants commit more crime than those born here because his company's belief might itself be an expression of their hateful views. But while such a match between belief content and the properties that make the content unwelcome is sufficient for the company to provide higher-order evidence against a belief, its absence is not always exculpatory.
Consider this variant of Drug. Nnedi, who is new to the country, comes to believe that the drug is effective against Covid-19 on the basis of the testimony of Donald Trump (about whom she knows nothing, other than that he is the President). Later, she learns that this belief is predominantly held by people who acquired it from the same source as her, and that they regard this source as reliable only because they hold values she regards as reprehensible. Nnedi's unwelcome company, unlike Seanan's, is higher-order evidence against her belief. But in neither case is there a match between the content of the belief and the properties that make their company unwelcome. The difference between the cases lies in how they were acquired: Nnedi's belief arises from the same route (coarsely individuated) as that of her unwelcome company (the testimony of the President), whereas Seanan's arises independently of his testimony. This is a difference that makes all the difference: Nnedi's unwelcome company is higher-order evidence that her source is unreliable. Seanan's belief is neither a match for the properties that make her company unwelcome nor does it arise in the same way in her case and in her company's; it is therefore in the clear so far as the problem of unwelcome epistemic company is concerned.
Our test for when unwelcome company provides us with a defeater therefore needs to be disjunctive. Unwelcome company provides us with a defeater when
(a) There are grounds to suspect that our unwelcome company's belief expresses some of the very properties that make their company unwelcome; or
(b) There are grounds to suspect that our belief might have arisen via the same belief-formation process as the belief of our shared company, and that process is unreliable.
The ‘or’ here is inclusive; nothing prevents a belief from suffering from both problems at once.
Disjunctive tests or accounts are often less satisfying than more unified ones, but this is a principled disjunction. The two arms of the test point to two different ways in which our beliefs might go wrong: they might have a content we should find worrisome or they might arise via a process that is unreliable. The second disjunct captures a defeater of a very familiar kind: it should be uncontroversial that if a belief arises from an unreliable process, then it is unjustified. The first disjunct requires a little discussion.
Cases like Refugees illustrate how we may be rightly concerned about unwelcome company, even in the absence of evidence that our belief has arisen via a shared (unreliable) mechanism. Cory's belief may have arisen via a route entirely independent of that giving rise to the same belief in white nationalists. He should nevertheless be worried by recognition that the belief has a content that expresses the very properties that make his company so very unwelcome (i.e. being racist).Footnote 2 There are several different reasons why having such a content might constitute a defeater or, more weakly, a reason for suspicion. Cory's sharing a belief that (in their case) expresses the properties that make his company unwelcome might be evidence that the belief expresses or is influenced by worrisome states of his (such as implicit bias), which interfere with his capacity to respond well to evidence. The match in content is evidence that though he doesn't satisfy (b) – his belief didn't arise via the same process as the belief of his unwelcome company – his evidence arose via a different unreliable process.
The match in content might also provide Cory with a reason to suspect that his evidence is incomplete or biased. Perhaps the (implicit) racism of others, or structural racism, ensures that evidence in favor of the belief is highly salient or easily accessible, while evidence against it is pallid or missing (this is surely a realistic hypothesis: cable news may tend to report a higher proportion of crimes by refugees than by other groups, and rarely features refugees who are going about their lives peacefully and productively). More weakly still, the match in content may simply provide him with a reason to look more closely or to consider raising his standards for justification. That his belief shares a content that (in their case) expresses some of the same properties that makes his company unwelcome is surely a reason to look again; to seek further evidence or to reassess the weight of the evidence he has.
This is a test that can readily be applied. Should vegetarians be concerned to discover that they're in some of the worst epistemic company imaginable? It's very difficult to see how vegetarianism can express the properties that make Hitler's company so very unwelcome. It is, moreover, unlikely that any causal route to belief formation shared by vegetarians and by Hitler (perhaps a response to animal suffering) might have the epistemic properties that made him a highly unreliable epistemic agent. This belief is in the clear, and the disjunctive test explains why.Footnote 3
