1. Key concepts

2. The motivation for experts as preemptive authorities

What is the rationale behind the claim that whenever someone has identified an expert, this expert has the status of an authority for them? Suppose Sara is a pupil who does not know anything about relativity theory. One day, she learns from her teacher that the expert physicist Einstein discovered that it is not true that two events are either simultaneous or they are not. This is clearly a reason for Sara to defer to Einstein. But what happens to Sara's background reasons regarding this simultaneity claim? What happens to her everyday experience of simultaneity? If Sara is clever, she will understand the situation in the following way: As an expert physicist, Einstein clearly has superior reasoning skills in the field of physics, but he has – most probably – also considered all the relevant evidence that is available to laypeople such as me. Moreover, it is highly likely that Einstein responded rationally to the total available domain-specific evidence because he has excellent domain-specific reasoning skills. However, it is also true that when Sara considers domain-specific evidence on her own and aggregates it with the evidence of Einstein's judgment, she will deviate (at least slightly) from his judgment. This becomes more explicit when we talk about degrees of belief or credences. When adding the evidence of the expert judgment to the domain-specific evidence (either publicly available data or Sara's private experiences), it is highly likely that Sara will end up with a credence that differs (at least slightly) from the expert's credence. After all, the pool of total evidence Sara relies on differs from Einstein's, due to the evidence she adds to his judgment. All these basic facts are fully disclosed to Sara. But then she can see that by weighing further domain-specific evidence with the evidence of the expert judgment she will deviate from what she herself takes to be the clearly superior judgment of Einstein who has already considered all of Sara's own evidence. Hence, she recognizes that considering these further pieces of evidence would lead to a judgment that is not fully rational. She might factor in evidence that has already been taken into account by the expert (this would be a case of double counting), or she might use evidence that is clearly irrelevant or false from the expert's perspective, or she might make a mistake in weighing the evidence on her own. In any case, she has evidence that the expert's judgment is the result of the rational weighing of all the relevant evidence. At this point, the pupil should realize that her domain-specific reasons are not rationally usable because they would lead to a judgment that is irrational to the extent that it deviates from the expert judgment. So, taking the expert's judgment seriously does not only generate a reason to follow his lead but also generates a reason that preempts the pupil's further domain-specific reasons.Footnote ¹¹

Before I present this line of thought as a semi-formal argument, I must highlight some relevant background assumptions. The identified expert judgment generates a preemptive reason for the pupil's reasons with respect to p under the following three necessary conditions: (a) ONLY IF there is no reason to believe that the pupil has domain-specific reasons on her own that are not shared with the expert, (b) ONLY IF there is no reason to believe that the expert judgment is based on domain-independent reasons, and (c) ONLY IF there is no reason to believe that different equally rational answers to the total evidence are possible in this particular case. It seems obvious that the expert judgment will not determine the rational response to the relevant evidence, if Sara has gathered new, so far unknown evidence (e.g., she might have made a relevantly new observation that is not shared with the expert), or if the expert judges about propositions outside his area of expertise, or if different responses to the same evidence can be equally rational (as permissivismFootnote ¹² claims). However, these would be rather atypical cases. Typically, when a layperson confronts an expert, these conditions will be satisfied, and the layperson will know this.

Here is, then, the argument for the preemptive character of a recognized expert judgment:

Therefore,

This argument is clearly valid. It has the form:

1. (1) p → q

2. (2) q → r

   ------------------------

   (C) p → r

We can use a conditional proof to show its validity:

1. 1 p   assumption

2. 2 p →q   premise (1)

3. 3 q   →Elim : 1, 2

4. 4 q → r    premise (2)

5. 5 r     →Elim: 3, 4

6. 6 p → r    →Intro: 1-5

Now, let me clarify the premises. (P1) appears particularly plausible if one thinks about judgments in terms of graded beliefs or credences. If the layperson considers further evidence apart from the expert judgment, then this will have at least a minimal weight that, in the aggregation, will make at least a minimal difference. (P2) claims that the expert judgment (most likely) sets the standard for a rational response. This is only true, if the three conditions explicated above are satisfied. It is true (a) only if there is no reason to believe that the expert judgment is based on domain-independent reasons, (b) only if there is no reason to believe that the layperson has additional domain-specific evidence that is not shared with the expert, and (c) only if there is no reason to believe that permissiveness holds for a rational response in this case. This qualification has two significant consequences: First, what is preempted are the domain-specific reasons of the layperson only. Second, these domain-specific reasons are preempted, only in so far as the layperson has no reasons to believe that they are not considered by the expert. In this case, the evidence of the layperson is a subset of the expert's evidence. All pieces of the layperson's evidence that are not taken to be considered by the expert do not get preempted. This means that preemption through the expert judgment is always restricted in the indicated ways.

The argument sketched above is just one among other argumentsFootnote ¹³ that have been delivered to support the preemption view.Footnote ¹⁴ Without going into further details here, it seems to be the most promising argument. Its main advantage is that it explains preemptive reasons as undercutting defeaters – a phenomenon that everybody should be happy to accept.

3. Why Zagzebski's attempt to cope with the dismissal of outrageous expert judgments fails

On Zagzebski's view, any reasonable view on the nature of expert judgments must cope with the challenge from outrageous expert judgments. For her, it must be possible to reject outrageous judgments. She also claims that the authority view has the resources to cope with this challenge.

What does Zagzebski's response to the challenge look like? She believes that trust in experts is based on reasons and that we can use these resources to reject the thesis that experts are authorities, if they hold outrageous beliefs.

So, Zagzebski claims that preemptive authority is justified and can also be defeated. But how is this going to work in detail? One natural worry here is that you cannot use preempted reasons as a defeater for the outrageous expert judgment. This is so because preempted reasons are not rationally usable. However, Zagzebski seems to claim exactly this, namely that preempted reasons can be used against their defeater (as defeater-defeaters), if the authority judgment is too absurd in light of the preempted reasons. There is another option here for the proponent of the authority view, namely to claim that preemption is limited to domain-specific reasons and that we can use the domain-independent reasons that are untouched by preemption to reject outrageous judgments and authority. But Zagzebski does not take this route. This becomes clear when she addresses the 4,000-pills case. There, she argues that the reason why we (as laypeople) can reject the judgment is not that it is physically impossible to follow the expert's recommendation (as in a case with 40,000 pills) but that the judgment is outrageous (Zagzebski Reference Zagzebski2012: 116, fn. 12). She thinks that reference to the 40,000 pills case blurs this point. On my interpretation, when it becomes physically impossible to take all these pills, we have domain-independent reasons (which are not preempted) to reject the judgment. In contrast, Zagzebski seems to argue that you can even use your domain-specific reasons to defeat preemptive authorities. It is this claim that I find hard to swallow.

Why is it impossible to defeat the authority judgment on the basis of the defeated domain-specific reasons? There are two main objections to this. First, that a reason is preempted means that it is rationally unusable and thus counts for nothing. It seems obvious to me that you cannot use a reason that has literally zero evidential weight to overrule a judgment. In order to do that the reason would still need some minimal weight, which preempted reasons do not have. The evidential role of such reasons can be illustrated by the epistemological mechanism of undercutting defeaters that are paradigm cases of preempting reasons. Consider again the case of red impressions under red-light illumination. When we are informed that the widgets that appear red to us are illuminated by red light, the evidential weight of the red impressions is undermined. In such a situation, red impressions are preempted reasons. Can we still use these impressions to defeat the defeater? No. It will not help, if we ourselves and other people experience the redly illuminated widgets as red again and again. Even an enormous amount of extremely vivid red impressions would not overrule the undercutting information about the red lights. The explanation for this is that the red impressions have zero weight and thus cannot make a difference to what is rational or justified to believe on the matter.Footnote ^15, Footnote ¹⁶ A defeater of the undercutting defeater would have to be source-independent. Such a defeater-defeater might, for example, be that another person informs us that our original testifier was lying to us about the red lights.

Second, when we use preempted reasons to reject authority judgments, we confuse them with reasons that were overruled but not preempted. When a reason is overruled, it still counts. It is simply weaker than the contrary reason. But it might be aggregated with further reasons such that, as a result, it overrules its defeater again. In the case of rebutting defeaters there is a continuing aggregation process. But treating reasons like this differs fundamentally from treating them as preempted (for this objection, see also Lackey Reference Lackey, Benton, Hawthorne and Rabinowitz2018: 237–8).

The lesson to be learned here is that even the authority view can explain how outrageous judgments of experts can sometimes be defeated. But this explanation must always refer to domain-independent reasons that are not preempted. This also means that the expert's outrageous judgments cannot be dismissed simply in virtue of being outrageous. Everything depends on the right reasons for their outrageousness.Footnote ¹⁷ (More on this in the next section.)

There is one further concern about the authority view that needs to be addressed here. Jennifer Lackey objects to the authority view that it does not have the resources to identify outrageous beliefs as such. As she puts it:

I think that there is a clear answer to the question she poses. On the authority view, the “replacement” of reasons is an exclusively normative issue. The “replaced” reasons do not vanish metaphysically but are rendered rationally unusable. Now, if these reasons are still there, they also can explain why the expert judgment appears outrageous to us. Judgments may feel decidedly false to us on the basis of reasons that are already defeated. Something like this happens for example in the case of sensory or cognitive illusions where the truth is very hard for us to believe. In the definitional part of this paper, I defended a non-doxastic reading of outrageousness. On that reading, reasons that are preempted can still create the impression of obvious falsehood.

4. How outrageous judgments by preemptive authorities can be rationally rejected

One thing that can be learned from the discussion of Zagzebski's view is that if we accept the authority view, we cannot use domain-specific reasons to reject an experts’ outrageous judgment. In this section, I will explore to what extent there are further resources available to a layperson that may justify the rejection of these outrageous judgments. I will proceed by first investigating whether there are some general structural limitations to preemption. Then, I will discuss the cases listed above one by one to find out whether there are specific resources available in each of these cases to reject the expert's outrageous judgment.

One principled limitation of preemption is First Person Authority. When a physician, after a brain screen and further testing, tells you that you don't have pain although you feel terrible pain yourself, the physician's judgment is outrageous for you and you seem to have every right to reject this judgment. Interestingly, this verdict can be accepted by the proponent of the authority view. This is because the physician's judgments about your pain are not based on the excellent method on which your judgments about your pain are based upon. Whereas the physician only has third-personal access to your pain, your own judgment relies on first-personal introspective access. This methodological difference explains why your reasons in this case are domain-independent reasons, since domains are, as I have argued above, methodologically rather than metaphysically individuated. So, if your physician tells you “You don't have pain” when you feel terrible pain, you can rationally reject this outrageous judgment, and you can do this even if you are a proponent of the authority view.

A second limitation of preemption concerns expert judgments that are undermining trust in experts. Suppose someone whom you regard as an expert in the field of psychology tells you that she is no expert, that there are no experts or even that persons do not exist. Or suppose an expert philosopher tells you that there are no preemptive reasons or that there are no meaningful judgments. In these cases, the expert tells you something about her field of expertise that, if true, would undermine your trust in her preemptive authority. You cannot, at the same time, accept her as an authority and believe her preemptively. The class of judgments characterized by this feature thus cannot, for structural reasons, be accepted on authority.

It is tempting to claim that there is a further class of reasons that cannot be preempted through authority – a priori reasons. There is hardly any current rationalist who believes that a priori justification is generally infallible or incorrigible.Footnote ¹⁸ For instance, if we do mathematics in a non-empirical way, there is always the possibility of making a performance error or of further progress through the revision of some axioms. However, some philosophers hold the view that a priori justification cannot be defeated by empirical reasons.Footnote ¹⁹ If this were true, a priori reasons could never be preempted by the testimony of experts. This would have the welcome consequence that one could use one's own a priori insights to reject outrageous mathematical expert judgments such as “2 + 2 = 5” or Paul's judgment in the Extreme Restaurant Case. But there is good reason to doubt that a priori justification is empirically indefeasible and thus cannot be preempted by expert judgment. There are three main reasons that support the empirical defeasibility of a priori justification: First, our beliefs can be justified by more than one source. That, for instance, my pencil lies on my desk is something that I can justify by looking at it, by remembering that I put it there a short time ago, or by relying on a trustworthy testifier. But if beliefs can be justified by different sources why shouldn't they be defeasible by more than one source? This general feature of epistemic overdeterminationFootnote ²⁰ is a prima facie reason for believing that a priori justified beliefs can be defeated by a priori as well as empirical reasons. Second, suppose that you try to make a complicated calculation. You are a bit unsure about its result and, hence, you ask an expert mathematician for advice. She tells you that you are almost right but have to revise the numbers a little bit. To me, this testimony looks like a perfect rebutting empirical defeater of my a priori reasoning. Third, there are examples of empirical undercutting defeaters for a priori justification. Suppose that neuroscientists discover that a certain pattern of neurological activity NA is correlated with erroneous mathematical proofs. In a specific situation, you write down a certain mathematical proof and the neuroscientists tell you that NA has recurred. Intuitively, your belief based on the proof has been undercut.Footnote ²¹ In light of these reasons, one should accept that a priori justification is empirically defeasible. It is, therefore, not excluded from preemption. A more specific case of rationally indefeasible beliefs would be beliefs that are meaning-constitutive. According to this view, it is impossible to believe that Archibald is a bachelor without believing (or being disposed to believe) that Archibald is unmarried. If one were to give up the latter belief one would cease to believe that Archibald is a bachelor. If there are beliefs that are constitutively related to concepts, these beliefs could not be defeated by expert judgments. In contrast, we could always use these beliefs to defeat the conflicting beliefs of experts. However, whether there are such beliefs involved in conceptual competence is highly controversial.Footnote ²² Footnote ²³

What are, apart from these structural limitations of preemption, the domain-independent resources for rejecting outrageous expert judgments? As far as I can see, there are three major sources of evidence typically available: social evidence, analytic evidence, and evidence from domain-independent disciplines. Let me address these sources in due order. First, consider social evidence. We often do not only know what one expert judges in her field of expertise, but also know how many other experts in the field agree or disagree with this judgment. This “social” information can be a domain-independent basis for a rational rejection of an individual expert judgment. If, for instance, the majority of relevant experts rejects the view of one individual expert, then this is a good reason for the layperson to reject this judgment as well. Sometimes, even social information about the agreement among laypeople can justify the rejection of expert judgment. Suppose there are experts and laypeople on mental math. Let us assume that the expert has a truth ratio of 90% and that the laypeople are still minimally competent such that they have a truth ratio slightly above 50%. If 1,000 laypeople all do a given calculation independently and if they fully agree about the result, then information about this fact will outweigh a diverging expert judgment.Footnote ²⁴ Second, consider analytic evidence. Suppose an expert physicist writes a book about physics which contains a number of fallacies, contradictions or conceptual inconsistencies. According to the authority view, we as laypeople cannot reject her views on the basis of our physical reasoning. However, as long as she is not an expert in logic or linguistics, we can use our domain-independent logical or conceptual reasons to challenge her view. This is clearly true, if the lay physicist is an expert logician. Third, consider evidence from domain-independent disciplines. Some expert judgments have implications for other disciplines. For example, if a weather forecaster predicts that it will rain potatoes tomorrow, her claim has also, for example, physical implications regarding the mechanics of potatoes and weather. According to the authority view, we are not permitted to use our domain-specific reasons (about meteorology) to dismiss this claim. But we can nevertheless draw on our knowledge of physics to do so.

Let me apply these abstract categories to the eight examples on the list of outrageous expert judgments and see to what extent we can reject these judgments on the basis of the remaining domain-independent reasons. The first example is the case of your trusted physician who tells you that you should take 4,000 pills an hour for the rest of your life. This judgment seems outrageous. What are the resources available to the proponent of the authority view that may allow her to claim that a layperson can reject it? Whereas taking 40,000 pills an hour might be physically impossible, taking 4,000 pills may not be. But there is social evidence available to most laypeople that permits rejecting this judgment. After all, you would have heard about prescriptions of such a large number of pills, if any had been given out.Footnote ²⁵ This indicates that your physician represents a minority view among the relevant experts. And this is, in turn, a domain-independent reason to reject her prescription. However, we can revise the case such that you live in an informationally isolated community in which there is no access to further medical research. Then, one can no longer argue that there are indicators that your physician represents a minority view. Under these conditions, you would not have any reason to reject his judgment. It thus would be rational for you to accept that should take 4,000 pills an hour for the rest of your life. Of course, one might argue that this is a case in which the practical stakes are high. Your very survival might depend on the right decision here. One might then argue that in a high-stakes case the threshold for justification is so high that a single expert opinion cannot meet it. You would then need to ask additional physicians for advice to pass this threshold. Alternatively, one might argue that, even if the threshold for justification is met, the threshold for action (namely, following the physician's prescription in practice) may not be. But this applies equally to all medical judgments and not specifically to the outrageous ones. Even in the case of a medical treatment that is perfectly in line with our expectations we then would have to ask additional physicians for advice. In conclusion, one has to admit that there may be variants of the 4,000 pills case in which it would be rational to follow the physician's outrageous prescription.

How should we think about the second example in which your expert calculator displays: “2 + 2 = 5”? A straightforward response would be that the belief that 2 + 2 = 4 is constitutive to the meaning of basic numeral concepts and hence, in principle, not empirically defeasible and we thus can use our own belief to dismiss the calculator's calculation. But as we have seen above, the view that some kind of justification is empirically indefeasible is controversial. However, there are further domain-independent reasons to reject the calculator's result. We have social knowledge of the majority view among mathematicians on whether 2 + 2 = 5. We have learnt in school and from the textbooks that 2 + 2 = 4, which further indicates this fact. And we may rely on this domain-independent social knowledge when we reject the calculator's result as too exotic. One may also argue that the expert calculator is superior to us with respect to complicated mathematical problems rather than simple ones, such as “2 + 2 = 4”. If this is correct, the calculator cannot be taken to be an expert with respect to these kinds of simple problems, even relative to mathematical laypeople. Then, however, laypeople are not rationally prohibited to use their own mathematical reasoning when solutions to simple mathematical problems become controversial.Footnote ²⁶

What about the forecaster's judgment that it will rain potatoes in the Rhineland tomorrow? First of all, one might find this judgment outrageous because, strictly speaking, it involves a conceptual inconsistency. It is analytically true that it rains raindrops, i.e., aggregations of water or at least liquids, when it rains. So, we might just use our – undefeated – conceptual competence to reject the forecaster's claim. But one can also understand her use of “rain” metaphorically. It rains potatoes in this sense if and only if potatoes will fall from the clouds, and that is conceptually possible. It also seems meteorologically possible that a strong tornado lifts up all kinds of objects from the surface that will fall down shortly after that.Footnote ²⁷ An expert forecaster might be able to exactly predict such an event. Nevertheless, it is physically highly improbable that the wind will pick up nothing but potatoes and it is also physically highly improbable that the wind picks up any potatoes since potatoes are objects that grow underneath the surface and are thus not a primary target of strong updrafts. So, this is a clear case in which a layperson can use her domain-independent physical and general knowledge to reject the forecaster's judgment.Footnote ²⁸

The next example is the case of a locally highly trusted moral expert such as a pastor. Suppose such a “moral expert” tells you: “Women are morally inferior to men.” Should you, as a “layperson”, adopt this attitude? The case is supposed to be set up in such a way that we take the pastor's judgment to be false. Would it still be rational for a woman living in the relevant community to defer to this moral authority without thinking about the moral issue herself? If the intuitive verdict is “no”, then there are still resources available to the authority view that can be used to dismiss this judgment. The most natural suggestion in this vein would be that we have social knowledge of a majority of other moral experts that disagree with the pastor. However, we can revise the case in such a way that the community is, again, informationally isolated. In this version, no member of the community has access to disagreeing moral experts. The proponent of the authority view would thus be forced to admit that laypeople then should follow the false judgment of the pastor. One simply has to admit that rational belief is fallible and sometimes misleading.

When an AI-scientist tells you that knowledge has turned out false, you may rely on your – unpreempted – analytic (conceptual) competence to reject this judgment. It is simply an analytic truth that knowledge requires truth.

What about the case of an expert philosopher who advocates an error theory about morality, numbers or even experts? If the expert disputes expertise, we have the case of a judgment that undermines trust in authority and thus in that very judgment. You therefore cannot believe what the expert judges preemptively. Error theories about numbers or morality must be treated differently. Here, again, we have social knowledge that tells us that we should not follow this exotic minority view in philosophy.

Next, consider the Extreme Restaurant Case. In this case, Paul diverges with his judgment to an extreme extent from our intuitive judgment and from all other judgments that are within the range of minimally plausible results. His answer is thus clearly outrageous. According to the authority view, we cannot rely on our mental calculations and the credences related to it. But there are clearly further resources available that allow the proponent of the authority view to reject this judgment. A proper subset cannot be larger than the set. This is a generally accepted set-theoretical truth. But Paul's judgment conflicts with this truth (a truth which he most probably accepts himself). So, one might argue that we can detect a logical inconsistency among Paul's mathematical judgments. And this is a domain-independent reason to reject his judgment. Of course, Paul might restore consistency by giving up the set-theoretical judgment. But we could object to this move by referring to the majority view among the mathematical experts.

In the final example, a biologist believes a theory that involves an obvious contradiction. As laypeople, we can easily reject this theory by relying on our domain-independent logical skills.

To sum up, one can say that six out of eight cases of outrageous expert judgments are such that we can reject them without relying on the preempted, domain-specific reasons. In the other two cases, we were unable to reject the expert judgment on the basis of domain-independent reasons, although in both cases the expert judgment is probably false. But this need not be an overall bad result since we have to keep in mind that rationality is not truth-guaranteeing.

5. Outrageousness and rationality

Initially, there seems to be a strong conflict between the core intuition that outrageous expert judgments should be rejected, simply because they are outrageous on the one hand and the authority view of experts according to which our own reasons are preempted by the expert judgment and hence cannot be used to reject the outrageous expert judgment on the other hand. By now, it has turned out that the alleged conflict is much weaker than expected. First, identified expert judgmentsFootnote ²⁹ do not preempt all our reasons but only our domain-specific reasons. Second, the reasons that are not preempted are sufficient to justify the rejection of outrageous expert judgments in many, though not in all cases. In this section, I want to focus on the remaining conflict and will come to a rational assessment.

It is true that the authority view recommends accepting outrageous expert judgments even in some cases in which this judgment is false. So, you might think that it would be advisable to reject all outrageous expert beliefs and, as a consequence, also the authority view. However, there are also outrageous expert beliefs that strike the laypeople as obviously false but that are nevertheless true (and well justified). That it is not true that two events are either simultaneous or that they are not is something that follows from one of the best confirmed theories in physics, relativity theory, but it is very hard for the layperson to accept this implication. Something similar is true for ontological indeterminism and non-locality within quantum physics.Footnote ³⁰ Here are two further, more mundane examples. The first is about the collapse of the World Trade Center at 9/11. According to the data on this collapse, it took the Twin towers 14 to 16 seconds to collapse. In comparison, it would take something to free-fall from the top of this building 9 seconds. The difference is so small that laypeople do not understand how the buildings could collapse so quickly without a controlled demolition, i.e., without controlled explosions on many of its floors at the same time. Keep in mind that the World Trade Center was a very solid construction. The outrageousness of this fact is clearly expressed by Rosie O'Donnell, one of the “Truthers”, who characteristically express incredulity:

However, according to the community of experts, simulations have proven that a time like the actual time of the collapse was to be expected under the given circumstances. Here we have a clear case of an expert judgment that appears outrageous to some laypeople although it is correct.

My second example is the Monty Hall Puzzle. Suppose you are in a game show. Behind one of three doors is a brand-new car that the player will win if she picks the correct door. The player picks, say, door number one. Then the host opens one of the other doors, say door number three, showing that this is not the correct door. The host now offers the player the option to switch from door 1 to 2. Should she accept this offer to raise her chances? A number of scientific studies has shown that an average of 85% of people do not switch (Bruns and Wieth Reference Bruns and Wieth2004). For them the chances seem equal and hence there is no reason to switch. However, as expert mathematicians have proven and also simulated, the player's chances are raised from 1/3 to 2/3 if she switches.Footnote ³² This is an outrageous but true expert judgment.

Often but not always, outrageous expert judgments turn out true. So, outrageousness is not a guarantee of falsity. On the other hand, the psychological studies by Kruger and Dunning (Reference Kruger and Dunning1999) and Dunning (Reference Dunning2005) suggest that the attribution of competence to oneself and the attribution of incompetence to others is strongly biased and not competently produced when the relevant person is incompetent herself. As Dunning shows, incompetent people are blissfully unaware of their own incompetence because the metacognitive cues of competence (such as fluency or the availability of reasons) are not reliably exploitable as evidence by incompetent people. Hence, incompetent laypeople tend to assess themselves as competent. When they are highly confident in this self-assessment they will classify most opposing views as outrageous.

The impression of outrageousness is clearly not a mark of falsity (too often outrageous expert beliefs turn out true). Moreover, outrageousness itself cannot be treated as a reason to reject the outrageous judgment. However, epistemic feelings probably have an adaptive epistemological function (Hookway Reference Hookway2003). They may serve as a heuristics for cognitive regulation. For example, we select good inferences because they feel compelling, we choose among alternative explanations the one that feels most appropriate, or we assess relevance on the basis of immediate salience. Under appropriate conditions the spontaneously occurring feeling of outrageousness may indicate that the considered judgment is massively irrational. What triggers this emotion may be a strong cognitive dissonance with the agent's background beliefs or natural appearances (McCauley Reference McCauley2011) or massive disfluency in cognitive processing (Alter et al. Reference Alter, Oppenheimer and Epley2007; Thompson et al. Reference Thompson, Prowse Turner and Pennycook2011; Levy Reference Levy2017). Outrageousness may thus provide us with a reliable heuristics of selecting strongly irrational judgments when conditions are normal. However, this heuristics may easily turn into bias if one confronts expert judgments. This anti-expert bias results from the fact that expert judgments are most likely in conflict with laypeople's background beliefs (since they are innovative or even revolutionary) and processed less fluently (because they are unintuitive). As I have already argued, being in conflict with laypeople's (prima facie rational) background beliefs does not render expert judgments irrational, if one keeps in mind that they preempt the laypeople's domain-specific reasons. Given the anti-expert bias in the mechanism behind the outrageousness impression, we should not expect rational guidance from them when we confront expert judgments. If we really want guidance in these cases, we have to look at the reasons behind the impression of outrageousness.

6. Objections and replies

In this section, I will discuss three objections to the view defended here. According to the first objection, preemption generally requires deference to the judgment of epistemic superiors, even if their judgment is only slightly more competent. Such an epistemic behavior may easily have epistemically terrible consequences. Consider Bernard Williams and Philippa Foot, and imagine that one of them – it needn't concern us who – is slightly more reliable in her judgments about moral philosophy than the other. It would be clearly disastrous for philosophical practice if Williams or Foot simply deferred in their judgments to their counterpart since it would be the end of all fruitful debates in philosophy. This looks like a reductio of the preemption view.Footnote ³³

In reply, two things can be said. First, what is epistemically required and what is instrumentally valuable with respect to certain epistemic ends are two completely different things that may conflict under specific conditions. For example, conciliatory views on peer disagreement rationally require that the parties suspend judgment under fully disclosed peer disagreement. This rational requirement obviously conflicts with the fact that only sticking to one's guns throughout intellectual controversies promotes intellectual progress. However, this does not show that there is no rational requirement to suspend judgment in the face of peer disagreement in the first place. So, having epistemically bad consequences does not automatically speak against being rationally required. Second, expert controversies among near-but-not-perfect peers are often such that the opposing parties do not fully share their evidence. Typically, the parties rely on their own insights, original arguments or different perspectives on the topic under discussion to support their views. This keeps the debate going. If that correctly describes their epistemic situation, the conditions for preemption are not fully satisfied. It was argued above that expert judgment generates a preemptive reason ONLY IF there is no reason to believe that the pupil has domain-specific reasons on her own that are not shared with the expert. However, the debating experts do have such a reason since they can reasonably expect of one another to possess not yet fully shared evidence. If, on the other hand, all arguments and pieces of evidence have been fully disclosed and no further dynamics are missing, deference to the epistemic superior, if she is identifiable, seems fully appropriate.

Here is a second objection: The argument that is intended to support general preemption does not really succeed since it supports non-aggregation rather than preemption. What is really prohibited is that we use our own domain-specific reasons over and above the reasons provided by the expert testimony. Now, consider a situation in which the layperson's judgment that is based on domain-specific evidence is confirmed by the expert judgment. In that situation, the layperson may argue that she has good reason to trust her own judgment. Since aggregation is prohibited, she may not use the expert judgment as additional evidence. Hence, when the layperson agrees with the expert, non-aggregation can be implemented without preempting the layperson's use of her own domain-specific reasons.Footnote ³⁴

Although the situation looks symmetrical when the expert agrees with the layperson, it is not. For the layperson, it is much safer to rely on the expert judgment than on her own domain-specific reasoning, even in cases of agreement. This is so because even if the layperson correctly assesses the evidence she may do so accidentally or in an incompetent manner. Since the layperson takes the expert to be much more competent than herself, it is always safer for her to follow the expert's lead.

Finally, there is the worry that preemption and subsequent deference to the expert abstracts away from the temporal setting of a conversation between layperson and expert. Typically, the layperson will not only request ready-made verdicts about some issue but also will ask for further explanations in order to improve her understanding step by step.Footnote ³⁵ One might even argue that the layperson should suspend judgment unless her judgment about the issue is based on firm understanding.

It is true that laypeople often do not only strive for correct answers to given questions but also demand understanding which involves, among other things, the ability to base judgments on domain-specific reasons. However, it is not true that one should not form a belief unless one has some degree of understanding. In the face of expert testimony, just the reverse is true. In a first step, the layperson just follows the expert's lead and is bracketing the rational use of her own domain-specific reasons. At stage two, the layperson uses her off-line abilities to simulate understanding of what she already believes on the basis of expert testimony. If her simulated domain-specific reasoning leads her regularly to what the expert has already told her, she has become an expert herself and can now base her judgments on her own understanding. Correctly interpreted, striving for understanding is an epistemic goal over and above forming true judgments about the world and does not exclude the temporal priority of the latter.