2.1 The Domain of Rationality

What kind of thing are we interested in when we’re making judgements about what is rational and irrational? We care about the mental and practical lives of human subjects, but not the whole of our mental lives, and not the whole of our practical lives. No one is squabbling over the rationality constraints of daydreaming about a holiday to Sicily, or how many cloves of garlic one might add to the pasta, or whether one wears blue or red socks on Wednesdays. So which states of mind or behaviours do we care about when we’re thinking about (ir)rationality?

The easiest category is those states identified as having a mind-to-world direction of fit, which is to say, states whose purpose is to match the world. Beliefs are the paradigm case. Things go right with my belief that the pub serves an excellent Riesling, when it is true that the pub serves an excellent Riesling. Beliefs are prime candidates for rationality assessments, with a standard approach being that when beliefs fail to match the world to an appropriate degree in the appropriate circumstances, we have candidate cases of irrationality. Of course, ‘to an appropriate degree in the appropriate circumstances’ is where the action is, lest we rule out the bare possibility of rational false belief or irrational true belief. We’ll say more about the rationality constraints on belief later (Sections 2.2 and 4).

Alternatively, we might think of those states which have a world-to-mind direction of fit, that is, those states whose purpose is to bring it about that the world falls into line with them. The paradigm case of this kind of attitude is desire. Things go right with my desire to have an excellent Riesling this evening when I do in fact have an excellent Riesling this evening. Whether desires can be rational or irrational is a contentious issue. It is easy enough to make a case for desires being instrumentally irrational, that is, being irrational insofar as they prevent one from achieving other goals. For example, I have a desire to go swimming tomorrow morning, but if I succumb to it, I’ll miss the deadline for submitting a reference for a student. Is my desire irrational? Well, perhaps, insofar as missing the deadline for submitting a reference would be pretty bad for me (if I care about my professional reputation for being reliable, organised, and trustworthy), and even worse for my student (who, presumably, would rather like that job). We might think that the role played by my desire to swim in this unhappy set of circumstances renders that desire instrumentally irrational.

But could my desire to go swimming tomorrow morning, or even my desire to do something less noble (stay in bed until midday?), be irrational by itself, irrespective of any unfortunate consequences of that desire? The case for non-instrumentally irrational desires is tricky, and indeed, some philosophers have denied that there are any such cases (most famously Hume Reference Hume1739, esp. Part 3, Sect. III; see also Hubin Reference Hubin1990).Footnote ² Although we’ll see a role for desire in many of the cases we discuss later, I won’t say anything further about whether desires can be non-instrumentally irrational. We’ll instead focus on less controversial cases.

Let us turn now to the relationship between rationality and irrationality. A tempting thought is that deviation from what is rational is irrational, but we should note that rational and irrational are (often) contraries not contradictories (which is to say, they can’t simultaneously characterise a mental state or action, but the possibility of arationality demonstrates that they can simultaneously not characterise a mental state or action). In standard conditions, we might resist the characterisation of my benignly imagining that I am the World Rubik’s Cube Champion as rational, but we’d probably also resist the idea that it’s irrational. My singing my favourite Elton John song whilst driving to work doesn’t look like a rational action, but it doesn’t look irrational either. These cases might be good candidates for arationality, perhaps because there’s not an imaginative activity or car-based action in which I ought to have engaged instead. The status of rationality and irrationality as mere contraries looks secure. But when it comes to other areas of my mental life, for example, my patterns of inference, there might be no arational way to proceed. If I believe that if p then q, and I also believe that p, the inference I draw from these beliefs can only be rational or irrational. I’m rational if I infer that q, and I’m irrational if I infer something else (or perhaps, even, if I infer nothing at all) (with the irrationality being more severe if I am occurrently aware of if p then q and p). In such contexts rationality and irrationality are contradictories.

As will become clear, many of the cases we’ll look at more closely resemble what we want to say about inference than what we want to say about my imaginative tendencies towards my Rubik’s cube credentials, or my singing along to Elton John. That is, our cases are ones where a more rational belief or action suggests itself. They are cases where a natural thought is that, speaking from a standpoint of rationality, our subjects could have done better. Given this, we’ll proceed by identifying what makes a person, mental state, or action rational and see the failure to meet such conditions as our route into irrationality.

2.2 The Standard Picture

Let us begin with what has been called the Standard Picture of rationality, which we will refer back to throughout the Element.

Adopting this Standard Picture naturally gives rise to what has been called the ‘pessimistic interpretation’ of the experimental results mentioned in the Introduction (Samuels, Stich, and Faucher Reference Samuels, Stich, Faucher, Niiniluoto, Sintonen and Woleński2004: 132). When experimental participants fail the Wason selection task, they fail to correctly apply the logic of conditionals. When experimental participants take it to be more likely that Linda is not only a bank teller but also one of the feminist variety, they misunderstand how probability works. With the Standard Picture in hand, we are left with ‘bleak implications’ when we turn to ‘the rationality of the man and woman in the street’ (Samuels, Stich, and Faucher Reference Samuels, Stich, Faucher, Niiniluoto, Sintonen and Woleński2004: 132). Of course, the Standard Picture also explains why we conclude that the experimental participants were selecting their cards or making their probability judgements in an irrational way; they were doing so against principles of reasoning we take to be constitutive of rationality.

Sometimes the relevant principles can’t be so straightforwardly understood as ones embodying proper appreciation of the formal rules of logic and probability theory.Footnote ³ How we understand the principles operative in such cases will depend on what kind of rationality we’re interested in, and what kind of thing it characterises (or not). It is common to distinguish two broad kinds of rationality: practical rationality and epistemic (or theoretical) rationality. Very roughly, practical rationality is concerned with our actions, whilst epistemic rationality takes as its remit our mental lives, and looks to the rationality of various cognitions – most obviously beliefs.

Let us turn to the kinds of cases that we’ll be focusing on – does the Standard Picture help us get a grip on what’s going wrong in such cases? Fred acts against his better judgement in leaving his umbrella at home. It is standard to think of cases like Fred’s as cases of practical irrationality. What ‘principles of reasoning’ does he fail to act in accordance with? A natural thought is that the practical reasoning underlying our action should issue in judgements concerning what to do which match our judgements about what it is best or better to do. Cases like Fred’s, where these things come apart, are said to be ones of weakness of will or akrasia.

Return to our cases of strange belief (Jack’s belief that there is a global cover-up of the fact that the Earth is flat, Katie’s belief that her wife is faithful, Glen’s belief that his daughter has been replaced by an imposter). Here the irrationality involved is of the epistemic variety. What principles are our protagonists failing to follow in forming their beliefs? Several rationality constraints on belief have been suggested, including consistency (one should not hold inconsistent beliefs) and inferential coherence (one should believe that which follows from beliefs already held). We’ll look at our cases in the light of these principles later. Perhaps the most obvious constraint on belief’s rationality concerns its having the appropriate relationship to evidence. Belief that is rational is belief formed in response to sufficient evidence, and which is sensitive to counterevidence. Jack, Katie, and Glen, we can suppose, all have beliefs which fail on both sides of the coin. Jack’s belief is based on poor evidence: testimony from scientifically illiterate friends and untrustworthy news sources, and is insensitive to the overwhelming evidence that the Earth is in fact a sun-orbiting sphere. Katie’s belief is based on her attending to the very thin evidence base for fidelity, perhaps her wife’s continued (albeit infrequent) expressions of love, whilst being insensitive to her coming home late, taking middle-of-the-night phone calls, a loss of interest in their intimate life, and so on. Glen’s belief is based on the very strange experience he has when looking at his daughter but is insensitive to the large body of evidence against his claim – the supposed imposter knows things only his daughter could know, no one else suspects her, and so on. Our cases of strange belief have in common, at the very least, an improper relationship to evidence.

What are we to say of Sue, some of whose actions look to be contrary to her firmly held and avowed egalitarian beliefs? In cases of implicit bias, the locus of irrationality is difficult to identify without first attending to what implicit bias is, and so I reserve that work for later (Section 5).

I’ve gestured at the kinds of things which might ground judgements of irrationality for our range of cases, of which we’ll say much more as we proceed. Before that, I’ll briefly address two distinctions regarding rationality, as well as two debates arising from the proper understanding and application of it.

2.3 Structural versus Substantive Rationality

Earlier I distinguished between epistemic and practical rationality. A different way of carving up the space of possibilities is suggested by the distinction between structural and substantive rationality. Benjamin Kiesewetter and Alex Worsnop understand structural rationality to concern combinations of attitudes (or the absence of combinations). This means that we can make assessments of structural rationality without knowing anything of the subject’s situation, and indeed, without even knowing the contents of her attitude(s). For example, we know that a subject who believes that p and believes that not-p is irrational. We can make that judgement without having to know about her circumstances, without having to know what p denotes, and indeed without having to know which of the two attitudes she ought to reject (Kiesewetter and Worsnip Reference Kiesewetter and Worsnip2023, section 1.2). Substantive rationality on the other hand, has none of these features, and is thought instead to relate to normative reasons, including those suggested by the weight of evidence. For example, we can denote the belief that grass is green and the delusional belief that my son is an imposter in the same way, namely, as the belief that p. Once we do so, though, we can no longer tell whether the belief is rational or not. In substantive cases then, the irrationality arises from what p denotes, whilst in structural cases it arises from relations between attitudes, regardless of what p denotes.

We’ll see that the irrationality attributed to subjects throughout our examples is of both the structural and the substantive kind. Sometimes our judgements of irrationality will be grounded in a subject having inconsistent attitudes, or their behaving in ways which diverge from their best judgement (structural irrationality). Other times they will be grounded in the subject’s attitudes having an inappropriate relationship to the evidence (substantive irrationality). As it happens, the project of this Element is not better served by explicitly carving these kinds of rationality up, particularly given that the distinction is not uncontroversial (Kiesewetter and Worsnip Reference Kiesewetter and Worsnip2023, section 1.3). For this reason, the distinction won’t play a role in the discussions that follow.

2.4 Ideal versus Bounded Rationality

So far we have sought to get a handle on what makes a mental state or action rational by appeal to the Standard Picture, which has it that rationality consists in reasoning in accordance with certain principles. What we’re doing in these cases is holding our target mental state or action up to the light of these principles and seeing whether it measures up. When it comes to assessing the rationality of belief, doing so in this way might be interpreted as a project of ideal epistemology, that is, a project ‘concerned with questions about what perfectly rational, cognitively idealized, computationally unlimited believers would believe’ (Carr Reference Carr2022: 1132). It is against this standard that we evaluate the subjects in some of our examples as irrational.

But is this standard a fair one? No one thinks that there is, in fact, a perfectly rational human agent, and so why would an imagined construction of one be the relevant standard against which to evaluate the rationality of our epistemic lives? This thought might make attractive a different approach, that of non-ideal or bounded epistemology. According to this framework, in assessing the rationality or otherwise of some cognition, we should recognise our cognitive limitations, or bounds.Footnote ⁴

David Thorstad (Reference Thorstad2024) suggests two ways of understanding the disagreement between ideal and bounded epistemologists concerned with rationality. A weaker way is to understand the two approaches as engaged in different but compatible normative projects, and the bounded epistemologist’s only plea is that we should accommodate facts about our bounds in order to get at a full picture of human rationality. A stronger reading of the disagreement has it that we have two incompatible descriptive projects, which, in some cases, in seeking to describe something as rational (or not), come to different judgements. There is only one type of rationality, and it is either ideal or bounded (397–8).

The project in the following sections is not one which opts for either approach. We’ll be looking at the etiology of particular kinds of cognition, and seeking to evaluate them against their typical contexts of manifestation. There may be some explanatory projects which play better with one approach over the other. For example, Jennifer Carr (Reference Carr2022) has argued that ‘[i]f we want a normatively robust theory of epistemic rationality, ideal epistemology is the only game in town’ (1131).Footnote ⁵ Such a project is not, of course, the focus of this Element.

As we proceed we will identify departures from ideal rationality. However, in our seeking to illuminate paradigmatic instances of irrationality, we will look at the broader context. Doing so is open to epistemologists of both stripes – the ideal epistemologist may well accept that, for example, a given belief was understandable or all-things-considered forgivable, but will just maintain that those evaluations ought not to move us from our judgement that it is nevertheless irrational. The bounded epistemologist may demur, having it that particular features of the broader context ought to make us reflect on our initial characterisation of our target belief as irrational. Sometimes our discussions take a bounded bent, but may be translated without much trouble into language friendlier to an ideal project.

2.5 Pragmatic Encroachment

Earlier (Section 2.2), in trying to capture what might make some beliefs irrational, we noted constraints including consistency, inferential coherence, and a belief’s relationship to evidence (and counterevidence). Whatever is said here about the rationality or otherwise of belief, it is coming firmly from epistemic quarters. That is, it is common to take it that non-epistemic considerations are not (and cannot be) reasons for belief, or things to which we might appeal in assessing the rationality of belief.Footnote ⁶

However, some philosophers have argued for the importance of recognising pragmatic encroachment in our epistemic lives. This is the idea that practical considerations should make a difference to our epistemic evaluations. Most commonly, the notion has been applied to knowledge ascription. This tends to be motivated by taking a low-stakes case to which we would happily ascribe knowledge, and then significantly raising the practical costs of being wrong, at which point, say the pragmatic encroachers, we should not ascribe knowledge after all (Kvanvig Reference Kvanvig2011: 77–8).Footnote ⁷

More saliently for us, it has been argued that there can also be pragmatic encroachment on the justification or rationality of belief. For example, Mark Schroeder argues that we can understand how there can be pragmatic encroachment on knowledge by understanding the practical factors we take to defeat knowledge to do so by defeating epistemic rationality (Schroeder Reference Schroeder2012: 266).Footnote ⁸ If the pragmatic encroachers are right, then we take too narrow a view of epistemic rationality when we understand it to be grounded purely in epistemic considerations.

We need take no stand on this issue, since in the cases we discuss the considerations which go into our assessments of rationality will be grounded only in the epistemic. There won’t be much by way of practical considerations which could plausibly encroach, even for those sympathetic to the idea of such a possibility (see fn. 8 for a notable exception).

2.6 Epistemic Innocence

Finally, let’s turn briefly to a recent research programme led by Lisa Bortolotti. It has been commonplace in the literature on epistemically imperfect cognitions to endorse what Bortolotti calls the trade-off view of practical benefits and epistemic harms (Bortolotti Reference Bortolotti2020: 125). That is, practical benefits come at the cost of epistemic ones. Bortolotti challenges this and offers a new way of thinking about epistemically imperfect cognitions, which is to use the framework of epistemic innocence. This framework promises to provide a more nuanced look at the relationship between epistemic and practical harms and benefits. To say that, for example, a belief is epistemically innocent, is not to say that it is thus rational, but only to say that it is due more epistemic credit than traditional frameworks might allow. Sometimes imperfect cognitions can be seen as emergency responses and, though epistemically irrational, can deliver epistemic benefits that would not be available were the subject to have the more epistemically rational cognition. A case for epistemic innocence has been made most robustly and generally in Bortolotti’s Reference Bortolotti2020 book, as well as specifically for delusions (Bortolotti Reference Bortolotti2015b, Reference Bortolotti2016; Sullivan-Bissett Reference Sullivan-Bissett2018), confabulatory explanation (Sullivan-Bissett Reference Sullivan-Bissett2015; Bortolotti Reference Bortolotti2018), psychedelic states (Letheby Reference Letheby2015), inaccurate social cognition (Puddifoot Reference Puddifoot2017), and clinical memory distortions (Bortolotti and Sullivan-Bissett Reference Sullivan-Bissett2020). I won’t discuss this notion any further, because it doesn’t go directly to our assessments of rationality. It does, though, serve as a reminder of the often complex and dynamic relationships held between irrational cognition and the contexts in which it occurs.

2.7 Concluding Remarks

We end our quick tour of rationality and irrationality here and move onto what are taken to be some of its paradigmatic instances. We might have taken a different approach which instead considered what various phenomena could tell us about the idea of irrationality, or even about the possibility of a perfectly rational being, or how it is that we might better live up to norms of rationality in our practical and mental lives. Thinking about instances of irrationality present in human psychology could be insightful to these broader, more abstract questions (for an example of this approach, see Bortolotti Reference Bortolotti2015a). This Element will come at the topic from a different angle. We’ll look closely at various kinds of cases and seek to come to an understanding of them by reflecting on their status as irrational. We’ll examine what it is about our cases that makes it proper to understand them in this way, and, where appropriate, we’ll question whether their reputation for being instances of irrationality is a fair one. In any case, my hope is that the discussions which follow will help the reader towards a better understanding of some of the more maligned aspects of our practical and mental lives.

3.1 What Is Akrasia?

As with so many phenomena to which the philosopher might direct her thinking, there is no uncontroversial characterisation of akrasia on which we can base the following discussion. The phenomenon of interest is also often called weakness of will, although some philosophers have argued that it is important to keep these two notions apart.Footnote ⁹ Before characterising akrasia in a minimally contentious way, let us, for the sake of contrast, say something about the non-akratic case. A natural (if over-simplistic) thought is that our actions are caused by some practical judgement concerning what to do, and that one’s practical judgement aligns with one’s better judgement (i.e. one’s judgement about what it is best or better to do).Footnote ¹⁰ What is meant by one’s better judgment? We can follow Nomy Arpaly (Reference Arpaly2000) in understanding it as referring to ‘the judgment that one reaches, having taken into account all the reasons one judges to be relevant, as to what would be best for one to do in a given situation’ (490). Now consider Figure 1 (where ‘=’ picks out that the same action is recommended by one’s practical judgement and one’s better judgement, and the downward arrow represents one’s practical judgement as the cause of one’s action).

Figure 1 The non-akratic agent.

In this highly simplified and idealised set-up, I form a practical judgement, of the form I shall do a, which aligns with my better judgement that it is better to do a. My practical judgement is the basis for my action, a, which is also the action recommended by my better judgement.

Let us turn now to akrasia, beginning with what Robert Audi identifies as ‘a common element’ in various accounts, that of acting against one’s better judgement (Reference Audi1990: 270).Footnote ¹¹ In cases of akrasia then, our schematic looks like this (Figure 2) (where ‘≠’ picks out that the action recommended by one’s practical judgement is different from that recommended by one’s better judgement).

Figure 2 The akratic agent.

On this picture, our action is caused by a practical judgement which does not align with our better judgement. What is it to act against one’s better judgement so understood? Here we can lay out four jointly sufficient conditions for so acting (adapted from Audi Reference Audi1990: 270).

For a subject S to act against her better judgement in performing action a, S:

1. 1. Performs action a intentionally

2. 2. Has judged that an alternative action, b, is better than a

3. 3. Has not abandoned the judgement that an alternative action, b, is better than a

4. 4. Is aware that she has judged that b would be better than a, and that she maintains this judgement.

Return to our earlier case of Fred, who judges that it would be better to take his umbrella. He has a big day at work schmoozing clients, involving taking them to lunch, followed by a walk along the canal, and (if all goes well), closing the deal over wine at a lavish bar. He thinks he will look credible and trustworthy in his new suit but takes this overall look to be significantly compromised should he carry his umbrella, and so resigns to take his chances, leaving the umbrella at home. However, a rainstorm is forecast, and the mildly deleterious effect on his overall look of carrying an umbrella would pale in comparison with the hugely deleterious effect of arriving to meet his clients dripping wet, or so negligently opening himself up to the elements partway through their encounter. He knows that carrying an umbrella is an annoyance, particularly if the forecast is wrong. He will need to remember where he puts it down as he goes about his day, and its bright pink colour really doesn’t set the suit off in a way one might hope of an accessory. Nevertheless, when considering the information available to him, Fred judges that it would be better to take his umbrella. Now imagine that Fred goes off to work without his umbrella. In doing so, Fred has performed an action against his better judgement (schematised in Figure 3).

Figure 3 The case of akratic Fred.

Let us turn to why the four conditions specified earlier are required for us to make good on this characterisation of Fred’s action.

Imagine that Fred went off to work without his umbrella unintentionally, that is, he simply forgot to grab it on the way out of the door. If the umbrella was left for this reason, we would withdraw our claim that Fred acted against his better judgement, although his action was of course inconsistent with it. We see then the necessity of condition (1), the action must be performed intentionally for it to be akratic.

Now imagine that Fred does not judge that it would be better to take his umbrella, and, indeed, does not take it. The case could be otherwise identical – a rainstorm is forecast, Fred knows this, but weighing everything up he fails to judge that taking his umbrella would be better than not taking it. Perhaps the colour clash with his outfit weighs on him, or perhaps he engages in motivated reasoning and comes to believe that it will not rain. It might well be that we, as neutral observers, take it to be obvious that Fred should take his umbrella, and we judge him harshly for not coming to this judgement. But if he really doesn’t come to this judgement, there is no akratic action in his failing to take his umbrella. Condition (2) is also necessary.

Now imagine that Fred did judge that it would be better to take his umbrella, but has since abandoned that judgement. Perhaps his evidence base has changed (rain is no longer forecast), or perhaps the reasons given previously for not forming the judgement in the first place now strike him, and he revises the judgement. On further consideration, prompted by the sight of himself in the mirror carrying what now seems comically ill-suited to accessorise his outfit, he revises his judgment that it would be best to take his umbrella – and with a knowing reckless abandon, invites the chips to fall where they may. Given this revision, Fred’s action of leaving the umbrella at home would not be akratic. Condition (3) is required.Footnote ¹²

Finally, condition (4) has it that Fred must be aware of conditions (2) and (3), that is, he must be aware that he has formed the judgement that it would be better to take his umbrella (condition 2), and he must be aware that his judgement has not changed (condition 3). Not being aware of (2) is easy enough – we can imagine that Fred judged that it would be better to take his umbrella, but then, through sheer forgetfulness or some more sinister memory-inhibiting event, simply forgot that he had judged this. And so, he leaves his umbrella at home. Doing so would not be akratic.

How should we understand what it would take to not be aware of (3), that is, for Fred not to be aware that he continues to judge that it would be best to take his umbrella? Fred not being aware that there is this judgement that he still holds is tricky, and gets us into territory about the nature of judgement. It is much easier to suppose that one could forget that one had judged, than that one could forget that one continues to judge. That is because it is commonplace to think that judgement is necessarily occurrent, or at the very least, that the judgement we’re interested in when thinking about akrasia is necessarily occurrent. If that’s right, then it would make little sense to suppose that Fred both (a) continues to hold the judgement and (b) is not aware that he continues to hold the judgement. This complication aside, we can see, I hope, even if it is hard to specify what it would mean for Fred not to be aware of a judgement he still holds, that Fred would need to be aware of his judgement, and of his maintaining it, to count as acting akratically.

Before turning to the rationality or otherwise of akratic action, I speak briefly to possibility. It might seem strange to ask whether akratic action is possible, having apparently (and thoroughly!) specified an example of that very thing. However, some authors have argued that akratic action is impossible, and, put simply, that’s because they see a very close relationship between judgements about what it is better for one to do, and decisions made about what is actually done (see, for example, Donald Davidson’s (Reference Davidson and Davidson1980) paradox of akrasia (23)). It might be thought that if I really judge that it is better to perform action b, then it follows from that judgement, all else equal, that I will perform action b.

Let us make explicit a presumption only implicit so far: akrasia stands up as an interesting phenomenon only to the extent that we take settling the question regarding what one will do as answered by settling what it is best or better for one to do. Without this, akrasia would be no more interesting than the fact that my judgement about what would be the most fun thing to do fails to bring about action in line with that judgement. Being the best action would simply take its place alongside other candidate actions recommended by their credentials in a given dimension, for example, that which is the most fun, or the most outrageous. Rejecting this implicit presumption would be to sever the link between judgements about what it is best to do with the practical reasoning that underlies what is in fact done. Akratic action would thus be possible, and more than that, not especially remarkable. Fred judges that it would be better to take his umbrella, but he also judges that it would be fashionable not to. Fashion simply won the day. This approach has been rejected by some philosophers on the grounds that severing the link between judgements about what is best with the practical reasoning underlying what is done is ‘too high a price to pay’ (Bratman Reference Bratman1979: 159).

A more nuanced route to establishing the possibility of akrasia involves modelling the relationship between two judgements: the action-prompting judgement which is the conclusion of practical reasoning, and the judgement about what it is best to do. Specifically, to make space for akratic action, we must be able to separate the two. Michael Bratman does this by casting the conclusions reached in practical reasoning as non-evaluative, and those reached by one’s better judgement as evaluative (Bratman Reference Bratman1979: 162). He has it that in failing to transition from the evaluative conclusion to the non-evaluative conclusion, one commits no ‘extreme logical error’ suggestive of impossibility (Bratman Reference Bratman1979: 164).

Another approach is to distinguish between the judgement that an action is prima facie better, and a judgement that an action is all things considered better (Davidson Reference Davidson and Davidson1980). In a case of akrasia where a subject’s better judgement is that x is better, but nevertheless she does y, the judgement that x is better is an all things considered judgement (one relative to the subject’s background beliefs and values she takes to be relevant to the case). A judgement of this kind is conditional (on the appropriateness of the beliefs and values that inform it). The competing judgement that causes the agent to y instead of x, is an unconditional judgement, and judgements of this kind are the ones that prompt action.

Our discussion of possibility has been extremely brief, but our end point represents what Sarah Stroud and Christine Tappolet (Reference Stroud, Tappolet, Stroud and Tappolet2003) have identified as ‘two general points of consensus’ in the contemporary literature on this topic: (1) that akrasia is possible, and it is so (2) ‘only as a species of practical irrationality’ (5). Let us turn then to assessing the (ir)rationality of akratic action.

3.2 Akrasia and (Ir)rationality

On one construal of what is going on in these cases, what is to be explained is why the akratic subject moves from considerations in favour of the akratic action a to the practical judgement ‘I shall do a’, given her recognition of considerations in favour of b, where the latter considerations are such as to recommend b over a as the best (or better) action. This is Bratman’s conception of the cognitive architecture, and he takes this transition to be perfectly possible, given his view of the nature of practical reasoning. However, in saving the possibility of akratic action, Bratman doesn’t say much about how we might explain this possible but irrational transition.

Similarly, Davidson’s theory of practical reasoning makes akratic action possible, but irrational. Davidson suggests that the rational subject will, in her application of practical reasoning, be guided by the principle of continence which has it that one should ‘perform the action judged best on the basis of all available relevant reasons’ (Davidson Reference Davidson and Davidson1980: 41). We are left not with questions of possibility, but with questions concerning why an agent would act in such and such a way, given the set of judgements involved. Why would anyone act in such a way as to be out of step with the principle of continence?

For Davidson, reason calls for Fred to carry his umbrella, the chances of rain are significant, arriving at the important meeting drowned will be a very bad look indeed and is likely to have undesirable professional consequences. But that pink umbrella really does look a bit silly carried against this rather expensive suit, it’s quite annoying having to carry it, weather forecasts are sometimes wrong … If Fred were not akratic, he would not give way to these latter considerations. But Fred is akratic, and he does. Why so? Davidson (Reference Davidson and Davidson1980) suggests that in seeking a psychological explanation of akratic action, we’ll find ourselves referring to otherwise familiar phenomena including ‘self-deception, overpowering desires, lack of imagination, and the rest’ (42).

The saving of akrasia from the grips of impossibility then does not wrap things up very neatly. We may have answered the question of whether akratic action is possible (yes), and we have done so by theorising about the architecture of practical reasoning, driving a gap between the judgements which motivate action, and those which represent one’s better judgement. Although we have found a gap, it is a small one, often bridged by what are taken to be principles of rational practical reasoning. These principles link together that which akratic action pulls apart – judgements that motivate action and judgements about how to act. That small gap has become what is to be explained, and that explanation will look different in each case because a whole range of standard resources are available to explain why we sometimes irrationally fail to have our actions align with what we judge it is better to do. Perhaps Fred convinces himself that today is the day the weather forecast will be wrong, or his desire to not offset his suit with a pink accessory is overpowering when he glimpses himself in the mirror. Our task now is to build explanations for why Fred allows for the gap exploited by akrasia in his practical reasoning.

There’s something mildly unsatisfying about this, if one starts from the idea that akratic action is a special form of irrationality demanding a particular explanation. If our explanation of the irrationality involved piggybacks on that involved in other areas of our mental lives, akratic action is not an especially interesting token of the type. The irrationality involved in akrasia might in fact strike us as rather pedestrian.

If one is in the market for something a bit more exciting to say, we might find what we’re looking for in some philosophers’ opposition to the idea that akrasia is always irrational, an idea which, as Audi (Reference Audi1990) observes, no philosopher ‘[had] seriously criticised’ (270). A decade later we hear a similar complaint from Arpaly (Reference Arpaly2000) who states that akratic action as irrational is largely uncontested: ‘it is almost a universal assumption in contemporary philosophy […] that acting against one’s best judgement is never an instance of rational action’ (490–1). Both Audi and Arpaly argue that at least some akratic actions are properly taken as rational, or at least, more rational than alternative actions, and they each take this idea to have not been given due attention.

One way of making good on this is to take a wider view of rationality, rather than having it that our better judgements are the ones that inform, or should inform, our action. On such a picture, action out of line with those judgements straightforwardly falls out as irrational. Audi (Reference Audi1990) offers the following counterexample:

Audi has it that John’s case is one of akratic – but not irrational – action. Although John’s better judgement – informed by his retributive desires – is that he should punish his daughter, it would be to take too narrow a look at the situation to characterise his not punishing his daughter as irrational purely for being out of line with that judgement. In John’s case, the rational basis for his better judgement is ‘outweighed by the larger rational considerations producing the incontinent action’ which also render the action rational after all (277).

Audi (Reference Audi1990) advocates pulling away from the model which grounds rationality on the appropriate transition between one’s better judgement and the conclusion of practical reasoning. Instead, a more holistic approach should inform our assessment of when an action is rational, taking into account ‘sufficiently good reasons of the agent’, even when those reasons are not part of the judgement concerning the action that would be best (or better) (278–9). Akrasia’s relative rarity is down to the fact that usually our overall reasons are reflected in our judgements about what is best, but sometimes such judgements pull away from our deepest beliefs and desires. If we give way to these deepest beliefs and desires, opting for the action recommended by them rather than our better judgement, we engage in akratic action which is not irrational.

Another way to separate akratic action and irrationality is suggested by Arpaly (Reference Arpaly2000), whose argument begins by drawing attention to cases of epistemically rational belief formation in the absence of deliberation, which she refers to as cases of dawning. It would take us too far afield to lay out detailed examples (as Arpaly does), so let us follow her in understanding these cases as ones in which ‘people change their minds, sans deliberation, as a result of a long period of exposure to new evidence’ (508). At least some cases of dawning are ones in which ‘we regard people as rational despite the fact that their rationality is not the product of deliberation’ (509). Arpaly asks:

We are invited to see akratic action as arising in a way akin to cases of belief change via dawning, which is non-deliberative, and sometimes rational. If non-deliberative belief change is not always irrational (and even sometimes rational), why not non-deliberative action? That’s a fine question, and Arpaly is surely right that we would need a reason to rule out the possibility of something rational arising out of a non-deliberative process in the action case if we allow it in the belief case. The plausibility of this for akratic action will depend on whether we think such cases are properly characterised as ones of action without deliberation.Footnote ¹³ Arpaly (Reference Arpaly2000) thinks they are:

There’s a move here from good deliberation to deliberation (as italicised), but the rational agent’s manual presumably doesn’t delineate all that is possible with us cognitively (lest we run together irrationality with impossibility). Putting that aside though, is it accurate to understand akratic action as not arising from deliberation? Personally, this strikes me as a rather revisionary take.

Our discussion so far is one that may well be interpreted as taking akratic action as the output of a process of deliberation over what to do, indeed, perhaps the interest of akrasia precisely stems from framing it as arising from deliberation. How can one’s practical judgement concerning what to do (which gives rise to action) diverge from one’s judgement about what it would be better to do? How can one decide to do a whilst at the same time recognising that it would be better to do b?

If akratic action does not arise from deliberation but is merely inconsistent with one’s better judgement, it starts to lose its notes of strangeness. As outlined earlier (Section 3.1), arguments for the possibility of akratic action proceed by constructing an architecture of practical judgement which allows for deliberation over what to do giving rise to actions divergent with one’s better judgement. Such talk is naturally heard as concerning the nature of deliberation over what to do.

Arpaly has it that the prescription to act against one’s better judgement is ‘flagrantly absurd’, and so doing so cannot result from deliberation. For me, this is precisely what’s interesting about akrasia. In seeking to make some akratic action rational, Arpaly takes action of this kind out of the deliberative realm. As a broader picture of the space of possibilities regarding rational action, non-deliberative action may well take its place. But acting inconsistently with one’s better judgement is not always to act against one’s better judgement. Thus, in conceiving of akratic action as a member of the class of non-deliberative actions, we may have changed the subject.

3.3 Concluding Remarks

Notwithstanding notable exceptions, there is consensus that cases of akratic action are those performed against one’s better judgement. We have seen that this is something which, with some creative modelling of the cognitive architecture, can be rendered possible, but irrational. Although often held up as a paradigm case of irrationality, such irrationality consists in failing to have one’s action guided by one’s best judgement, and that failure is attributed to a variety of causes, none unique to akrasia. And, in fact, there are ripples of dissent concerning whether akratic action is irrational after all. If we allow for a broader view of rational action, not taking it to consist solely in one’s practical judgements being appropriately related to one’s better judgement, we see that there are some cases where that appropriate relation does not hold, but where we might reasonably deny irrationality. Or, if we take akratic action not to be the output of faulty deliberation concerning what one should do, but rather as non-deliberative, we are welcome, as in the epistemic case of belief change via dawning, to see some cases as rational after all.

4.1 Doxasticism about Strange Belief

The commonality among the beliefs we will consider is that they all look a bit different from ordinary belief. Some philosophers have found these differences so compelling that they have argued that we are not in the domain of irrational belief after all, and that is because we are not in the domain of belief! Doxasticism about a given attitude type is the view that tokens of that attitude type are beliefs. Non-doxasticism about a given attitude type is the view that tokens of that attitude type are not beliefs. Non-doxasticists will often pair this negative claim with a positive one about how we should understand the attitude instead. That is, once the attitude type is cast out of the belief club, non-doxasticists will sign it up for membership to a different club, perhaps the club of imagination, or mere avowal.

Non-doxasticism about the cases to be considered later in the Element is typically motivated by characterising all beliefs (not just the rational ones) as, necessarily, having certain features. This might relate to their formation and revision (e.g. appropriate relationship to evidence), or to their downstream effects (e.g. that in concert with appropriate desires, beliefs motivate action). When attitudes fail to be formed or revised in line with these constitutive constraints, or when they fail to motivate the kind of action one would expect from a belief with a given content, it thus follows that those attitudes are not beliefs after all. Broadly speaking, when doxasticists seek to keep the target attitude in the club of belief, they have pursued two options. They will either loosen strict conceptions of belief on the grounds that a whole host of attitudes we standardly take to be beliefs also fail along some dimension of such strict conceptions, or they will argue that the target attitude has the features required of the strict conceptions after all.

Whether attitudes arising from conspiratorial ideation, self-deception, and delusion are beliefs matters for the project pursued in this section. That’s because what we are able to say about the rationality of any given attitude will depend on what kind of attitude it is. For example, some non-doxasticists appeal to imagination to model their target attitudes.Footnote ¹⁷ Whether an attitude is a belief or an imagining has clear implications for what can be said about whether or not it is rational. Consider Nikki and Sebastian, who both have an attitude with the content there’s a duck hiding in my office. They arrive at their respective offices one morning and spot a feather on the floor. This prompts Nikki to believe that there’s a duck hiding in my office, whilst it prompts Sebastian to merely imagine that there’s a duck hiding in my office. We can see that the contents of Nikki’s and Sebastian’s attitudes are the same, and we’ve even specified that they are prompted by the same event – perceiving a feather on their office floors. But neither of these things, singularly or combined, will settle the matter as to whether Nikki’s and Sebastian’s attitudes are rational (or not). We probably want to say that Nikki’s attitude is irrational. Why? In so believing, Nikki may have violated rationality constraints on belief. Perhaps her belief is inconsistent with other beliefs she has, for example, that there aren’t any ducks on campus, or that ducks are never properly described as ‘hiding’. Or perhaps she ought to have inferred that there is not a duck in her office given that that followsFootnote ¹⁸ from her beliefs that she locked her office door and ducks can’t pick locks. At the very least the evidence for her belief is insufficient, particularly as her office is not situated in a duck-dense environment. What about Sebastian? We probably don’t want to say that his attitude is irrational. Why not? To do so would make epistemology particularly stuffy. Surely Sebastian is free to smile to himself and allow the presence of the feather to prompt a fun imagining of a duck hiding in his office (this might be an instrumentally irrational thing to allow if it distracts him from his work all day, or if he has a duck phobia, but for a judgement of irrationality to stand up, we’d need to be creative in stipulating further features of the broader context).

In evaluating our attitudes with respect to rationality then, we are engaged in a project downstream of settling what kind of attitude we’re talking about. To keep things manageable, I will presume doxasticism throughout (and will refer readers to non-doxastic approaches). This will allow us to keep the goal posts stationary. Non-doxasticists are entitled to assess these attitudes as irrational (or not), but their case for such assessments will be very different from the parallel cases generated when working with a background presumption of doxasticism.

4.2 Belief and Bias

Before turning to our cases of interest, we need to say something about bias, since often it is by appeal to biases that philosophers and psychologists have explained the formation and maintenance of the beliefs we will discuss. To say that someone’s reasoning exhibits a bias, we need to have in mind what that bias is a deviation from. For our purposes, we can have in mind the Standard Picture noted earlier (Section 2.2), where we understood rationality as reasoning in line with principles based in the rules of logic, probability theory, and so on. To say of someone’s belief that it was formed or maintained in a biased way, is to say of it that it was formed or maintained in a way which departs from ideal belief formation and maintenance practices, and that it does so along a particular dimension.Footnote ¹⁹

Suppose I wanted to explain the presence of philosophical beliefs, or a subset of them, perhaps what Marga Reimer has labelled nihilistic philosophical doctrines. These might include external world scepticism, hard determinism about free will, and anti-realism about morality (Reimer Reference Reimer2010: 317). Now let us (perhaps implausibly) imagine that we take the endorsement of nihilistic philosophical doctrines to call for explanation, perhaps because there are a host of other (more mainstream) philosophical doctrines on the market that are more explanatorily fruitful and less gloomy. What we might do is take some nihilistic philosophers as our experimental group, and some non-nihilistic philosophers as our control group. We could then design an experiment which tests for reasoning biases outside of the philosophical context. Perhaps we’d ask our philosophers to select their preferred explanation for a range of imaginary events, where those explanations were subject to variables like plausibility or pessimism. If we found that nihilistic philosophers were more likely to opt for explanations which scored highly on our variable of pessimism, we would have evidence for the claim that nihilistic philosophers exhibited the pessimism bias.Footnote ²⁰ Attributing the pessimism bias to nihilistic philosophers would then be (at least part of) the explanation for why they endorse nihilistic philosophical doctrines over sunnier alternatives. This is (very roughly) how much of the psychological research on strange beliefs has proceeded, and we’ll discuss some of it in the following section.

One more point before we go on: it is useful to distinguish cold from hot biases. Cold biases are non-motivational and emerge from heuristics, whilst hot biases are those driven by emotions and desires. Examples of the former are the availability bias (our tendency to rely on examples which easily come to mind as representative), and the confirmation bias (our tendency to search for or interpret information that confirms our existing beliefs). Hot biases are those arising from motivation, most obviously, our desiring that p or our desiring to believe that p. This desire might, for example, bias the evidence we gather or the evidence we attend to.Footnote ²¹ Let us turn now to our three cases of strange belief.

4.3 Conspiracy Beliefs

The term conspiracy theory can be understood to pick out an explanation of an event that appeals to the intentional states of conspirators, who intended the event and kept their intentions and actions secret (Mandik Reference Mandik2007: 206). So far, so epistemically respectable, after all, some conspiracy theories are true, well justified, and so on. Most of us believe that Guy Fawkes conspired to blow up Parliament in 1605 with assistance from a handful of co-conspirators and thirty-six barrels of gunpowder. But discussion of that belief doesn’t belong in a book about irrationality, and endorsement of that belief doesn’t make one a conspiracy theorist, as that term is standardly used.Footnote ²² These are not the conspiracy theories which will interest us here.

The theoretical interest in conspiracy theories arises not just from their contents, but also from their being held in opposition to some received view or orthodoxy. The reason one doesn’t have to be a conspiracy theorist to endorse the Guy Fawkes conspiracy theory, is that doing so puts one in good company (that of the relevant epistemic authorities, in this case, historians). This is why some philosophers have defined conspiracy theories in a way that makes this feature explicit. To give just two examples of a more general trend: David Coady (Reference Coady2003) argues that ‘the proposed explanation must conflict with an “official” explanation of the same historical event’ (201) and Ichino and Juha Räikkä (Reference Ichino and Räikkä2021) have it that the explanation ‘conflicts with the received explanation of the said event, providing an alternative to the “official view” of that event’ (249). I’ll follow these authors and understand the phenomenon of interest to be those beliefs in theories concerning conspiracies which are in conflict with official explanations of events. I’ll use the term conspiracy belief to pick out belief in such theories.Footnote ²³

Let us introduce some examples. Jack believes that NASA and governments worldwide are covering up the fact that the Earth is flat. Joe believes that Covid-19 is a side effect of 5G towers. Romany believes that the ‘terrorist attacks’ on the Twin Towers in New York City and the Pentagon in Washington were an inside job. One interesting feature of conspiratorial ideation is that believing in one conspiracy theory is highly predictive of believing in others, indeed, ‘the single best predictor of belief in one conspiracy theory is belief in a different conspiracy theory’ (van Prooijen and Douglas Reference van Prooijen and Douglas2018: 898). Several explanations of this have been proposed, ranging from a broader underlying belief which supports the endorsement of conspiracy theories in general (Wood, Douglas, and Sutton Reference Wood, Douglas and Sutton2012), to various cognitive contributions (discussed in what follows) which characterise the conspiracy theorists’ reasoning, which might make attractive multiple conspiracy theories.

A related point is that conspiracy theorists may feel the burden of the rationality constraint of consistency,Footnote ²⁴ in such a way that their conspiracy beliefs become more elaborated. Often, for consistency, one has to accept further conspiracy theories in order to maintain existing ones. For example, if you believe that NASA and governments worldwide are covering up the fact that the Earth is flat, you’ll also have to believe that the moon landings were faked and that aircraft pilots and scientists are involved in the cover-up, and so on. From one belief, others sometimes must grow, otherwise the initial theory becomes unstable.

Conspiracy theories may mutually support one another, but they are all in conflict with received wisdom from the relevant epistemic authorities. So how do people end up with them? A huge amount of research in cognitive and social psychology has sought to answer this question, of which we can only scratch the surface. The first thing to note is that, despite many conspiracy beliefs being extremely strange in their contents, they are taken by most researchers to be a ‘normal’ phenomenon (Pierre Reference Pierre2020: 618). So, when researchers try to understand why, for example, Jack believes the flat Earth conspiracy, they’re not trying to identify unique irrationalities or pathologies in the way he forms and maintains his beliefs. Rather, they’re looking to identify particular normal range features of cognition or social position which contribute to the endorsement of conspiracy theories. As Joe Pierre (Reference Pierre2020) puts it with respect to the former:

Let us turn now to a discussion of the rationality of conspiracy beliefs.

4.4 Self-Deception

The literature on self-deception is vast, and as we have seen so far, often even pinning down the phenomenon of interest can be a controversial exercise. Let us begin with three examples which reflect at least some consensus.

In these examples, we can see what Neil Van Leeuwen (Reference Van Leeuwen2007) identifies as three ‘strands of consensus’ in the philosophical literature regarding what self-deception involves:

1. 1. A motivational component

2. 2. The presence of information which would justify a different belief

3. 3. The output of a truth-evaluable cognitive attitude

Let us identify each of these components in our examples. First, we can understand the requirement of a motivational component as an evaluative attitude (most often a desire) towards the content about which the subject is deceived. We can see that for each of our self-deceptive attitudes, there is a desire related to the content of the resulting belief. Our cases look rather different in this regard, and that is no accident. Rather they exemplify three kinds of self-deception, which differ with respect to how the motivational component relates to the content of the self-deceptive belief. Katie’s case is of the classic variety: wishful self-deception, in which the subject desires that p, and the self-deceptive belief matches the content of the desire – the subject believes that p. Eileen’s case is different. Although she desires to beat Jean, her route to a belief with that content is a little more circuitous. She recognises that her beliefs about the matter will make a difference to whether she wins or loses, and the desire in play in her self-deception is not the desire that p but rather the desire that she believes that p. This is the slightly odder case of wilful self-deception. Things look very different for Aaron, where he ends up believing that which he desires not to obtain. Aaron desires that he has not failed the exam and ends up with the belief that he has in fact failed the exam. This has been called dreadful (Van Leeuwen Reference Van Leeuwen2007: 425) or twisted (Mele Reference Mele2001: 4) self-deception.Footnote ²⁶

The second strand of consensus is that there is information available to our subjects which would justify a different belief. In all of our cases, we have described our subjects’ evidence (were they to give it its proper due) as supporting the opposite of the belief that they self-deceptively come to hold. For Katie, the evidence of Frankie’s likely infidelity is clear – Frankie is probably being unfaithful. For Eileen, the evidence of her likely defeat is also clear, it is very unlikely that she will beat county champion Jean, whose current form is good. For Aaron, the evidence strongly suggests that he has passed the exam – a fail would be the first on his academic record, and there are no reasons to think that his sustained excellence is about to come to an end. Notice that our subjects don’t believe anything that is demonstrably false, or for which the evidence in support of it couldn’t be explained in a way consistent with the self-deceptive belief. If Katie caught Frankie in an unambiguous act of infidelity, we wouldn’t take her continued belief in Frankie’s fidelity to be a case of self-deception, the kind to which we may all fall prey. Rather, self-deceptive belief relies on there being enough interpretative space that the evidence against the belief can be explained in a way that is nevertheless consistent with it.

The third strand of consensus is that the self-deceptive process ends in the generation of a truth-evaluable cognitive attitude. The specification of this component is loose enough to allow in non-doxastic theories of self-deception, which we have already put aside.Footnote ²⁷ For us, then, we can take self-deception to result in a belief in the content towards which one is self-deceived. Katie believes that Frankie is faithful, Eileen believes that she will beat Jean, and Aaron believes that he has failed the exam.

Let us turn now to the question of self-deception and (ir)rationality.

4.5 Delusion

Let us begin by putting aside the folk use of the term delusion, the kind one might employ in casual conversation. For example, I might tell my partner that he’s delusional if he believes that his football team will win a trophy this year. If you’re Richard Dawkins (Reference Dawkins2006), you’ll take anyone who believes in a religious deity to be properly characterised as delusional. These are not the uses of the term that will concern us here. Rather, I use the term delusion to pick out those strange beliefs of interest to psychiatry.

We can begin, as is common, with a definition of delusion drawn from the most recent Diagnostic and Statistical Manual of Mental Disorders. There delusions are described as:

Concerns have been raised with almost all components of this characterisation, specifically, with respect to the claims that all delusions are held despite clear or reasonable contradictory evidence, and that all delusions are beliefsFootnote ³² (for an overview see Sullivan-Bissett Reference Sullivan-Bissett and Sullivan-Bissett2024: 3–6). We won’t worry about these issues with the diagnostic criteria except where they arise with respect to our discussion of irrationality. Let us turn instead to some examples.

• Glen believes that his daughter has been replaced by an imposter (Capgras delusion)

• Marilyn believes that she has ceased existing (Cotard delusion)

• India believes that she has a second head (perceptual delusional bicephaly)

Let us also distinguish two types of delusion: monothematic and polythematic. Monothematic delusions concern a single theme and occur ‘in isolation in people whose beliefs are otherwise entirely unremarkable’ (Coltheart et al. Reference Coltheart, Langdon and McKay2007: 642). Polythematic delusions are more often elaborated and occur in the context of mental disorder. The relationships between monothematicity and circumscription on the one hand, and polythematicity and elaboration on the other are not exceptionless (Davies et al. Reference Davies, Coltheart, Langdon and Breen2001: 135), but this has nevertheless been taken to be a useful way to carve up the landscape. I’ll focus on monothematic delusions (our three examples fall into this category), because the debates which concern our question of rationality have taken place within this context.

One important feature of many monothematic delusions which is crucial for understanding their etiology is anomalous experience. For example, in the Capgras delusion, subjects experience a lack of affective response to somebody with whom they are close. In Cotard delusion, this lack of affective response may be generalised to the environment (Young et al. Reference Young, Robertson, Hellawell, de Pauw and Pentland1992: 200), or there is an experience of depersonalisation (see Gerrans Reference Gerrans and Sullivan-Bissett2024 for discussion of the Cotard experience). In perceptual delusional bicephaly, a subject may hallucinate a second head (see Ames Reference Ames1984 for a case of this kind).

A common approach to understanding delusions is empiricism, which is the view that these anomalous experiences play a causal role in the formation and maintenance of delusions.Footnote ³³ So, for example, the empiricist will say that Glen’s lack of affective response when looking at his daughter is part of the explanation of why he believes that his daughter has been replaced by an imposter. India’s hallucination of a second head is part of the explanation of why she believes that she has a second head. And so on. Broadly speaking, there are two ways of being an empiricist. Explanationist approaches understand delusions as explanations of anomalous experience, whilst endorsement approaches have it that the delusional content is present in the anomalous experience, and the subject endorses that content in her belief (Bayne and Pacherie Reference Bayne and Pacherie2004: 82). I’ll talk in explanationist terms in what follows, since such an approach lends itself more straightforwardly to the ideas I discuss.

4.6 Concluding Remarks

In general (notwithstanding noted exceptions), it is pretty uncontroversial that the beliefs we have discussed in this section are irrational. Why so? Because they all violate several plausible rationality constraints on belief. The debates concern how such irrationality arises. We have seen a range of possibilities – from cognitive biases which influence the processing of data, to motivational biases which might prompt one to prefer a belief with a certain feature. In the case of delusion – often taken to exhibit irrationality of an extreme kind – we saw that our usual resources for explaining the generation or maintenance of an irrational belief are often taken to be inadequate, and that a more severe, abnormal kind of irrationality must be posited. For my money, that’s a mistake, and delusions can perfectly well take their place alongside other strange beliefs. Where appearances suggest they involve more serious errors of reasoning, we can pay attention to the context in which they arrive and thrive (that of anomalous experience). More generally, the everyday irrationalities to which we are all prone can go a long way to help us understand strange beliefs of many stripes, even when at first blush they strike us as incomprehensible or, at least, to have arisen from such poor epistemic performance as to seem resistant to ordinary theorising.

5.1 Associationism

The standard view in philosophy and psychology is that implicit biases are associations. We all have stored associations, many in common with one another, some more unique to our personal circumstances. For most folk, when someone says salt, they think pepper. For many folk, when someone says Laurel, they think Hardy. For me, when someone says labour, I think of the British Labour Party and their electoral chances, whilst a midwife or a woman who had recently given birth may be put in mind of something very different. These examples demonstrate the genesis of associations as residing in the learning history of the subject (Levy Reference Levy2015: 803). To say that implicit biases are associations is to say something like this: Conor’s bias consists in a stored association between a particular concept (i.e. Muslim male) and a particular (negative) valence. In the presence of certain stimuli (e.g. seeing a Muslim male enter the train carriage), that stored association is activated. Some theorists have also explicitly allowed for associations between concepts (Mandelbaum Reference Mandelbaum2016: 630) (e.g. one’s concept of Black male and one’s concept of aggression, as we might ascribe to Sue).

5.2 Propositionalism

Although I identified associationism as the standard view, recently some authors have made the case for a radically different way of modelling implicit biases: propositionalism. States with propositional contents have satisfaction conditions, whereas states with only associative contents do not; instead, associative contents are ‘(relations among) mental representations that lack any syntactic structure’ (Mandelbaum Reference Mandelbaum2013: 199, fn. 1).

The case for propositionalism has been made by appeal to empirical work demonstrating that implicit biases are vulnerable to what Gabbrielle Johnson (Reference Basu, Beeghly and Madva2020) calls rational intervention, understood as ‘an attempt to intervene on a person’s implicit attitudes that relies on the informational content of the intervention (the reasons they present) rather than mere repeated exposure to the intervention’ (28–9). The case for propositionalism on these grounds has been most robustly made by Eric Mandelbaum (Reference Mandelbaum2016). He overviews empirical work that shows that implicit biases are modulated by logical and evidential considerations (see Toribio Reference Toribio2018: 41). In particular, they follow the logic of ‘the enemy of my enemy is my friend’ (Gawronski et al. Reference Gawronski, Walther and Blank2005), they are sensitive to argument strength (Briñol et al. Reference Briñol, Petty, McCaslin, Petty, Fazio and Briñol2009), and they are adjustable in light of peer judgement (Sechrist and Stangor Reference Sechrist and Stangor2001).

Of course, even if we find the case for propositionalism compelling, that doesn’t settle the question of what exactly implicit biases are. All sorts of mental states have propositional contents. Take the proposition implicit biases are difficult to theorise about. That could be the content of a belief, an imagining, a supposing, or even a (strange) desire. Propositionalism thus narrows the attitudinal possibilities, but there are still discoveries to be made.

Mandelbaum’s view is that implicit biases are unconscious beliefs. Sue’s implicit bias is best understood as the unconscious belief that Black men are aggressive, Tracey’s as the unconscious belief that women are not leaders, and Conor’s as the unconscious belief that Muslims are dangerous. Mandelbaum argues that this view can accommodate the empirical results supporting propositionalism.

A couple of other models which cast implicit biases as propositional are worth mentioning here. One comes from Levy (Reference Levy2015), according to which implicit biases are patchy endorsements. The endorsement part is that a subject commits to the world being as the proposition picks out, and the patchy part is in recognition of the fact that biases respond to only some sorts of evidence and feature in only some sorts of inference. On another, hybrid, view implicit biases are constituted by unconscious imaginings, at least some of which have propositional contents (Sullivan-Bissett Reference Sullivan-Bissett2019).Footnote ⁴⁴ Let us turn to what we can say about the (ir)rationality of implicit biases when they are understood as propositional.

5.3 Concluding Remarks

We began this section by characterising implicit biases as mental items that influence behaviours and discriminations which were not explainable by appeal to a subject’s explicit attitudes. From here the options open to us with respect to assessments of (ir)rationality depended on what we took implicit biases to be. According to the standard view, implicit biases are associations, and we saw a case for their being rational (insofar as they pick up on social realities). However, taken more broadly as items in an (egalitarian) subject’s overall cognitive economy, we saw that living in a society with inequalities leaves us only options of irrationality. Don’t encode the prejudiced biases which reflect social realities, or do encode them against one’s reflective judgement.

On the other hand, propositionalists take implicit biases to have propositional contents, and indeed, the most thoroughgoing defence of propositionalism took them to be unconscious beliefs. We saw that the background account of belief in play here made it illegitimate to refer to our earlier rationality constraints on belief, given that belief formation on this account is brute. We considered whether implicit biases might be rationalised on this model by appeal to their reflecting social realities, but saw that this was problematic.

Overall, implicit biases are perhaps the most obvious occupants of our minds which demonstrate that the rational isn’t always the good or the desired. Whereas it is natural to think that one should aim at rational action and rational belief formation, when it comes to implicit cognition shaped by social realities, things are considerably less clear.