1 Introduction

Lisa Bortolotti’s (2016, 2020) assessment that elaborated delusions can be epistemically innocent is stimulating both from an epistemological standpoint and a practical, therapeutic one. Although I partly disagree with this claim, I can fully agree with the general motivations behind Bortolotti’s work, which are to remove the stigma from delusions and psychosis (Ritunnano and Bortolotti 2021) and show how these delusions function in such a way that could be helpful from a therapeutic point of view (Bortolotti 2020). I also believe Bortolotti is correct in her quest to explain that such delusions are not meaningless; on the contrary, they can give meaning and fuel creativity (Ritunnano and Bortolotti 2021). Moreover, I think that the concept of epistemic innocence has already been proven to be helpful and can work well for many different kinds of imperfect cognitions, such as motivated delusions or confabulations.

A restrained interpretation of Bortolotti’s claim would be that at least some elaborated delusions can be judged epistemically innocent in the right circumstances on the basis that they are epistemically beneficial in some way (despite being epistemically costly at the same time) and that they satisfy the “no alternatives” condition. I think this claim is technically true, but it is essential from both philosophical and therapeutic perspectives to explore the details. Moreover, I will argue that most elaborated delusions in most circumstances bring an intense type of epistemic harm, not comparable to the harms caused by imperfect cognitions that had so far been qualified as epistemically innocent. I believe that the definition of epistemic innocence should not allow us to categorise such delusions as innocent, but it does not exclude them in its current form.

My goal is, therefore, twofold. First, I will explain what is unique about elaborated delusions in schizophrenia, what makes most of them so epistemically harmful most of the time, and which of them can be judged as epistemically innocent and at what point. I will also examine them in the light of the extant definition and show why the current conditions for epistemic innocence might not be enough to provide judgement. Second, I will suggest how the conditions for epistemic innocence can be modified to exclude beliefs that cause such radical harm but still apply to other kinds of delusions and imperfect cognitions that deserve to be absolved. Finally, I will argue that Bortolotti’s original motivations behind the claim that elaborated delusions are epistemically innocent can be preserved, and fruitful practical implications can be drawn from the analysis of elaborated delusions offered.

2 Section 1: What are the Elaborated Delusions in Question

Defining delusions, especially those that are elaborated and typically encountered in people on the schizophrenic spectrum, is by no means uncontroversial. For reasons of space and clarity, I must leave aside the philosophical debate on whether delusions, in general, can or should be viewed as beliefs (for an overview, cf. Bortolotti, 2012). Let me assume a loose understanding of “belief” as a type of conceptual cognition without committing to any specific claims concerning its propositional or linguistic nature and its relation to either actions or other kinds of cognitions. I am making this assumption to be able to employ the most widely accepted definition of delusion, formulated by the American Psychiatric Association in DSM-5 (2013, p. 87), which tells us that it is a fixed belief that is not amenable to change despite conflicting evidence. However, the Appendix still features an older, yet more controversial definition, according to which delusions are “false beliefs based on incorrect inference about external reality” (p. 819). By now, it has become universally accepted that being false is not an essential feature of any delusions.

As is usually the case with technical terms, the way the term “delusions” is understood is varied and embedded in psychological, philosophical, and cognitive theories, a full review of which lies beyond the scope of this paper. However, I will refer to specific ideas in the course of my argument.

2.1 What is Special About Elaborated Delusions in Schizophrenia

It is, unfortunately, difficult to pinpoint the exact kind of delusions I would like to consider. Both Bortolotti and I wish to refer to elaborated, systematised delusions that typically occur in people who have schizophrenia. It is not clear whether those form a distinct type of delusions and, if so, on what grounds. In the DSM, delusions are divided into subgroups according to their contents or topic. Many of them can be said to be typical of schizophrenia: those about the self, misattributions of agency, delusions of grandeur, persecutory delusions, and inserted or broadcasted thoughts. Philosophers of psychiatry usually employ Coltheart’s (Coltheart 2013; Coltheart et al. 2010) helpful distinction between monothematic and polythematic delusions, as it grasps one essential feature of delusions: the extent to which they are entangled with other beliefs. Monothematic delusions are such that concentrate on a particular subject and are typically circumscribed: not fully integrated with the rest of the patient’s belief system. In contrast, polythematic delusions can refer to multiple topics and are usually elaborated and systematised – incorporated into the belief system (for details of this distinction, see (Coltheart 2013).

The three dominant approaches to delusions – the two-factor theory, the predictive processing models and the phenomenological approach – offer slightly different perspectives on the specificity of elaborated delusions.

The two-factor theory treats delusions as irrational beliefs that respond to or explain very strange perceptions or experiences, typically caused by a neurobiological pathology. According to this theory, both the belief and the experience are pathological, which is why they are both factors contributing to the emergence of delusions. This approach is best suited to monothematic delusions such as the Capgras delusion, especially if we can gain empirical insight into the first factor – the neurobiological pathology causing the strange experience. The two-factor approach has not been as fruitful when tackling the other kind of delusions; the not-circumscribed, polythematic, systematised, and elaborated delusions. From the point of view of the two-factor theory, elaborated delusions in schizophrenia may, of course, be treated as a special subtype, still, the theory itself does not point to their distinctive features, apart from the fact that they do not focus on one specific pathological experience with a clear neurobiological cause.

In contrast, the predictive approach to delusions is unified as predictive approaches usually are. In the family of predictive processing (predictive coding, predictive error) accounts of delusions, delusions are treated as a single phenomenon without making any far-reaching categorisations (see e.g. Corlett 2009; Fletcher and Frith 2009; Fineberg and Corlett 2016; Griffiths et al. 2014; Parrott 2021; Sterzer et al. 2018; Erdmann & Matys, 2021). The generative model is a multi-level hierarchical structure of constantly updating predictive mechanisms comparing priors to inputs in order to generate more adequate posteriors. The system strives to remove any discrepancies – errors of prediction. Most explanations of delusions within the paradigm suggest that they result from disturbances in the process, especially concerning the priors. The pathological experience is extremely challenging for the system and generates strong error signals, possibly because the prior hypotheses are too vague. At the same time, the delusional explanation becomes a disproportionately strong prior that cannot be then correctly updated and, surprisingly, trumps any more mundane explanations that should normally emerge as more likely.

Although many predictive models of delusions have been offered so far, it must be noted that, as concludes Parrott (2021) in his review, there is no deeper agreement as to what delusions are, what causes them, and to what extent different types of delusions should be explained differently within the predictive framework. Litwin and Miłkowski (2020) fiercely criticise the predictive models of psychosis (including delusions), pointing to numerous inconsistencies between the models themselves and between those models and the general commitments of the predictive processing approach. Haarsma et al. (2020) also enumerate no fewer than four different ideas for the source of disruption in the hierarchical predictive mechanism in psychosis. Moreover, none of the predictive analyses offers any details about whether precisely the same predictive mechanisms fuel different subtypes of delusions. In sum, there is no clarity on how (and if) the predictive models could explain the more specific features of delusions encountered in schizophrenia, specifically if and how these delusions differ from monothematic and circumscribed delusions.

In the phenomenological and embodied tradition, schizophrenia in general and all the symptoms associated with it are usually treated as a unique disorder, the disorder of the self (Sass 2003; Sass et al. 2011). Most authors working within the phenomenological paradigm emphasise the need for a deeper understanding of the experiential dimension of any psychiatric phenomenon, including delusions (Sass 2004; Sass and Byrom 2015). This is visible in the three phenomenological questionnaires: EASE, EAWE, and EAFI, which allow psychiatrists to examine their patients’ anomalous experiences in detail (Parnas et al. 2005; Rasmussen et al. 2018; Sass et al. 2017). Within the EASE questionnaire, in particular, much attention is given to the many specific ways that the self is disturbed and how it affects self-awareness, presence, distorted bodily affectivity (including somatic depersonalisation), inadequate bodily demarcation, confusing self for other, and solipsistic experiences leading to an existential reorientation (Parnas et al. 2005, 2021) also observe, following Kępiński (1974), that the content of delusions in schizophrenia is typically “coloured by” ontological, eschatological, and charismatic themes. Let me explore this description, because it will be playing a role in the following sections of this paper.

Kępiński, an outstanding Polish psychiatrist and philosopher of psychiatry, was also grounded in the phenomenological school, although he tended to distance himself from it. He treated the contents of schizophrenic delusions with the utmost respect because he valued how they rip the patient out of a mundane, everyday world with its petty concerns and direct their attention towards ethical, philosophical, and religious considerations (Kępiński 1974). In doing so, he was probably influenced and inspired by his own experience of psychosis. This approach informed his characteristic, deeply humane, respectful, and humble attitude towards his patients. Kępiński described the world of his patients as richer, grander, more colourful (which is something frequently remarked upon in the context of the creative feats accomplished by people with schizophrenia), belonging to a different, higher plane of existence. This is not a description that could be applied to just any type of delusions.

Parnas et al. (2021) observe, referring to the German psychiatrist Mueller-Suur, that we cannot analyse the content of delusions in abstraction – their content reflects their nature and is dialectically interrelated with their form. Therefore, we can expect delusions encountered in schizophrenia, with their very particular themes and topics, to be substantially different to other kinds of delusions, also in their mechanisms or structures.

In the later sections of this paper, I will focus on one such unique feature of delusions encountered in people with schizophrenia that had been brought up in the phenomenological tradition: the phenomenon known as double bookkeeping (Bortolotti and Broome 2012; Gallagher 2009; Parnas et al. 2021; Sass 1995). It sometimes seems that certain patients can exist simultaneously in two worlds – one of them is the world of their delusions, and the other, the world they share with other people. Even though some delusional beliefs conflict with the shared world, the patient does not let go of their claims and does not perceive the tension. For example, the patient may be convinced that they are on a mission to convert and save all humanity, but this would not prevent them from interacting with medical personnel in a regular way (Cermolacce et al. 2018). On the one hand, this phenomenon resembles the circumscription of monothematic delusions, which typically fail to influence any actions. On the other hand, in the case of schizophrenia, double bookkeeping seems to be facilitated by the fact that elaborated delusions often do not even concern the empirical world within which any actions could be taken. Thus, no evidence against them can be found within the bounds of the shared empirical reality. I would like to draw attention to the similarities and differences between those two phenomena. Thus, I propose to think of double bookkeeping as a “broad” form of circumscribing a whole web of delusional beliefs. In contrast, by a “narrow” circumscription, I will be referring to the isolation of a single monothematic delusion or a tightly knit group of delusional beliefs about a single subject.

The main purpose of Sect. 1 was to introduce the crucial features of elaborated delusions that require attention when we assess their epistemic innocence. What I wish to discuss – and what Bortolotti is also referring to – is delusions that are polythematic, that result from the disturbance of the basic self, that are not circumscribed in the narrowest sense of the term, although they are in the broad sense, that mostly have the specific subjects I elaborated on, and that are typically encountered in people with schizophrenic or schizoaffective disorders.

In the following section, I will recount and then challenge Bortolotti’s line of argument for the claim that such delusions are epistemically innocent.

3 Section 2: When could Elaborated Delusions be Epistemically Innocent?

3.1 Bortolotti’s Original Case for the Epistemic Innocence of Elaborated Delusions

Bortolotti’s updated conditions for epistemic innocence are as follows:

  1. 1.

    Epistemic Irrationality. Belief b is epistemically irrational.

  2. 2.

    Epistemic Benefit. The adoption, maintenance, or reporting of b by agent A delivers some significant epistemic benefit to A.

  3. 3.

    No Alternative. The adoption, maintenance, or reporting of a less epistemically irrational belief than b is not available or would fail to deliver the same significant epistemic benefit to A as b. (Bortolotti 2020, p. 13)

The first condition emphasises that “epistemic innocence” is a notion designed to absolve such imperfect cognitions that are treated with suspicion in the first place. In the final section of this paper, I will delve deeper into the possible understanding of “irrationality” in this context. For now, I leave this condition out of my analysis, as it is satisfied by all types of delusions by their definition.

The second condition (from now on, I will call it EB) is to be understood by pragmatic lights. Bortolotti emphatically does not wish to reduce the notion of epistemic benefit to something like “the number of true beliefs.” On the contrary, she proposes that to bring benefit is to restore or improve epistemic functionality, which, in turn, is defined as the capacity to pursue and attain epistemic goals (again, without narrowing this notion to any particular understanding of what those goals should be). The crux of Bortolotti’s argument for the epistemic innocence of elaborated delusions is that they are beneficial in this pragmatic sense.

If reducing anxiety increases concentration, and both the adoption of the delusion and the good rest reduce anxiety, then they both support epistemic functionality to some extent. However, the contribution of the delusion is qualitatively different and more interesting for my purposes here. Assuming that elaborated and systematised delusions serve as default explanations for anomalous experience, the adoption of the delusion puts an end to the uncertainty caused by hypersalience, which some interpret as a consequence of inappropriate prediction-error signals, using the framework of the prediction-error theory. Therefore, in this explanatory framework, the formation of delusions is an imperfect solution to the anxiety caused by hypersalience instead of a short-term relief source, such as a good night’s sleep. (Bortolotti 2020, p. 84)

For Bortolotti, it is crucial that elaborated delusions can be considered an imperfect solution and not a problem in itself. However, the author emphasises that delusions of this kind may prove epistemically costly at the same time and that their positive assessment does not entail judging them as epistemically good or justified.

The third condition (henceforth: NA) can be satisfied in three ways. Relevant alternatives can be unavailable in the strict sense, in the motivational sense, or in the explanatory sense. In the case of elaborated delusions, Bortolotti claims, relevant alternatives that would bring equal benefit are strictly not available.

I believe, however, that the application of the above conditions does not allow us to fully grasp the epistemic status of elaborated delusions. In the following sections, I will re-examine this type of delusions from the epistemic perspective as well as propose an updated version of the conditions for epistemic innocence.

3.2 The Unique Epistemic Harms and Benefits of Elaborated Delusions

Let me first consider Bortolotti’s claim that elaborated delusions can satisfy the EB condition, even though they can also be harmful at the same time. I think Bortolotti is right in that she does not believe that a belief needs to be true, justified, and rational to be epistemically beneficial. I would also like to maintain this general pragmatic way of thinking about epistemic benefits. I believe, however, that the way epistemic benefit is defined now (as something that facilitates pursuing and attaining epistemic goals) may be a little too open to interpretation.

My background assumption for this claim is as follows: the concepts of epistemic innocence and epistemic benefit are only truly practically useful if they help us grasp something specifically epistemic, something that goes beyond psychological or emotional benefits. Bortolotti’s motivations for defending specific pathological cognitions as “epistemically innocent” stem partly from the conviction that this could help remove the stigma that simple arguments from psychological and emotional benefits could not remove. When we are told that certain beliefs simply make someone feel better, we tend to remain sceptical and critical – the classic ideals of rationality require us to be courageous in accepting difficult truths. We also typically value gaining true beliefs and other epistemic activity above just feeling well. All this makes “epistemic benefit” sound even more valuable than just “emotional benefit”. However, as it is now, Bortolotti herself directly compares the benefits granted by delusions to those of a good night’s sleep – not because the latter is known to help us to organise knowledge, but because both offer psychological relief. In both cases, the psychological relief is then said to translate to epistemic benefit. I think categorising specific cognitions as epistemically beneficial could be more profitable if the concept of epistemic benefit was clearer and focused on the distinctly epistemic aspects of the benefit.

Bortolotti’s argument for the initial intuition that elaborated delusions are epistemically beneficial (and, consequently, also epistemically innocent) is based on the general line of thinking that this author and her collaborators have successfully applied to other pathological cognitions, and, in particular, to motivated delusions, irrational beliefs that typically serve to defend oneself from emotionally difficult truths at the cost of self-deception. Although Bortolotti acknowledges the specificity of elaborated delusions emphasised by the phenomenological tradition, she also applies the same general philosophical (two-factor) and computational (predictive) frameworks to both elaborated and motivated delusions. In her view, elaborated delusions can also be treated as beliefs that initially do more good than harm because they attempt to alleviate an even more difficult epistemic situation. Specifically, they provide relief from anxiety, for they explain some of the most disturbing experiences caused by psychosis and introduce new meanings into the patient’s life. From a purely psychological perspective, the analogy between the benefits offered by both types of delusions seems sound. However, from the epistemic perspective, I think it underestimates the severe difference between the workings of those two kinds of delusions, perhaps made easier by the fact that the distinction is also at best blurry within the predictive framework.

Motivated delusions are also usually monothematic and, as such, circumscribed in the narrow sense. We can indeed think of them as a way of dealing with a problematic, pathological, or emotionally difficult experience – it is not the perfect way, it is in itself epistemically dubious, but it helps to get back on the epistemic track and proceed with other epistemic endeavours in an unaffected way. For example, in the case of the Reverse Othello syndrome or anosognosia, the delusional belief is similar to self-deception; it protects the feelings of the patient from a difficult truth, which helps them deal with the rest of their lives (see Bortolotti 2020, pp. 106–108). Their primary role is to ultimately become a building block for other sound beliefs.

This is not the primary role of elaborated delusions, though. Such delusions are only broadly circumscribed – they open up a whole different delusional reality and become a building block for the entire system of delusional beliefs about that reality. They are the first entries in a double bookkeeping system and they are the reason such a system has to be developed. True, they may exert an overall good influence on the patient’s wellbeing and help them resume clearer thinking in the shared reality. However, this is in the best case only a possible side effect that becomes quickly overshadowed by their inevitable primary outcome, which is to draw the patient’s interest away from their shared reality, to engage their attention and cognitive powers elsewhere.

Moreover, as Bortolotti also acknowledges, most delusions encountered in people who have schizophrenia are in themselves disturbing and ultimately psychologically and emotionally challenging. This challenge also translates directly into further epistemic harm: many delusional beliefs undermine the epistemic agency of patients, their sense of control of their thoughts, and their most profound sense of self. It is no surprise that delusions may even increase the risk of suicidal ideation. I would like to draw particular attention to the suggestion made by Parnas and his collaborators (Skodlar et al. 2008) that the primary cause of suicidal ideation in schizophrenia might be the experience of social isolation and self-alienation – I will be focusing on those consequences of elaborated delusions in the later sections of this paper.

For now, let me observe that from this perspective, any epistemic benefit that elaborated delusions in schizophrenia could provide is demolished. True, in the typical case of an imperfect cognition judged as epistemically innocent, epistemic benefit and epistemic harm co-occurr and their balance may be dynamically changing. In the case of elaborated delusions in schizophrenia, however, the possible epistemic benefits are usually directly causing harms. Moreover, as I will argue in greater detail in the following sections, the harms typically caused by elaborated delusions are such that even the fleeting co-existing benefits should not convince us to judge those beliefs as generally beneficial and epistemically innocent. This could even undermine some of the excellent work that has been achieved by defending – from the standpoint of epistemic innocence – other cognitions that deserve much more tolerance in the clinical and everyday setting. My tentative conclusion is, therefore, that the concept of epistemic benefit – and the EB condition – should enable us to provide a more nuanced (and perhaps harsher) judgement of this type of delusions.

3.3 The Harms and Benefits of Elaborated Delusions in the Light of the New EB Condition and the Concept of Sensus Communis

In Białek 2021, I have proposed an enhanced version of the EB condition, inspired by the phenomenological notion of sensus communis, interpreted by Fuchs as a “common space of intermodal integration that enables us to share the same social realm” (Fuchs, 2018; Thoma and Fuchs, 2018). My goal was to rule out categorising confabulations in gaslighting as candidates for epistemically innocent beliefs. This condition is as follows:

A cognition is epistemically beneficial for the agent if it is necessary for the agent’s epistemic functioning, by which it is meant that the agent is able to participate in the standard practices of exploring the physical and social environment shared by their community. (Białek 2021)

Before I discuss some of the general ramifications of this change, let me first show how this condition is helpful with respect to the matter at hand: in differentiating between motivated delusions and elaborated delusions in a way that does not undo any of the great work done so far by the concept of epistemic innocence and its proponents. As for motivated delusions (and some of the other types of imperfect cognitions), I have already shown why the new definition still allows them to be categorised as epistemically beneficial (2021). Let me now draw on this analysis to apply it to elaborated delusions, bearing in mind the theoretical and meta-theoretical complexities I have already discussed.

In the case of confabulations in gaslighting, the main reason why the new version of the EB condition would not grant those cognitions the status of epistemic innocence is that they are prying the agents away from their sensus communis. I also refered in 2021 to the concept of double bookkeeping to explain what happens when the gaslighter is forced to keep up simultaneously with the confabulated reality of their relationship with their victim and the real world. From here, my way forward is simple: systematised and elaborated delusions are the original type of imperfect cognitions that inspired psychiatrists and philosophers to employ the notion of double bookkeeping. What occurs in a mild form in the case of gaslighting is a full-blown mechanism in the case of delusions in schizophrenia.

However, the case of delusions in schizophrenia requires a more nuanced approach, for two reasons.

Firstly, I believe that the presence of double bookkeeping is not enough, by itself, to say that certain delusions cannot be epistemically beneficial. As I will shortly explain in more detail, despite being generally costly, double bookkeeping in itself can also have certain beneficial effects. What makes systematised and elaborated delusions in schizophrenia especially harmful is how they work within the double bookkeeping mechanism.

Secondly, and more importantly, I would not wish to contradict the claims grounded in the phenomenological school and reinforced by Bortolotti and her team (especially in Ritunnano and Bortolotti 2021): that delusions in schizophrenia can be meaningful and are never reducible to just some false beliefs that need to be weeded out of the patient’s system. In this respect, elaborated delusions are not comparable to the comfortable lies that a gaslighter may confabulate to preserve their self-esteem. Instead, I propose that elaborated delusions be understood as an attempt to open up a new sensus communis – one that is prohibitively difficult for other people to participate in but one that essentially creates a new realm of senses. Let me elaborate on the concept of sensus communis to make this claim clearer.

I propose two ways of thinking about what sensus communis is, how it relates to delusional beliefs, and what protects our ability to engage in participatory sense-making. It is apparent – since, at least, Wittgenstein’s (1953) inspirational portrayal of our language games and ways of life – that we do not function in a single sphere of rationality created by a single community within a well-defined environment. Instead, we participate in different language games appropriate for the different ways of life in which we engage. We share different games with different interlinking communities, but we do not share all of them with anyone. This first way of thinking leads us to say that there are multiple, independent, interlinked sections of a single, general, all-human sensus communis in which we all function as a community of epistemic agents. This way of thinking would make us consider the world created by elaborated delusions as an attempt to create a whole new sphere, one that cannot fit within the bounds of the one general sensus communis. Participating in this section would hamper the ability to engage in any part of the shared sphere.

The second way of thinking is to acknowledge the existence of a web of multiple interrelated sensus communes that are, however, also at least connected by the Wittgensteinian family resemblance relation. This would make the delusional world just a part of this great web, albeit one with particularly weak ties to any other parts. This idea is closely related to Gallagher’s concept of multiple realities (Gallagher 2009), which has also been used to conceptualise delusional reality as one of many in which we engage (Cermolacce et al. 2018). This approach, I believe, is superior as it allows for a clearer understanding of the relation between the delusional sensus communis and the other sensus communes and helps formulate the reason why partaking in the former may inhibit the ability to engage with the latter. The delusional world is so distinct, so different from the other sensus communes concerning both scope, topics, and possible activities, that it cannot be treated simply as just another language game in everyday life. At the same time, however, it is deceivingly similar – it is not something like a fictional world that can be wildly incomprehensible and disturbing but, at the same time, is obviously fake and can be ignored altogether. It poses as a part of the shared reality but turns out not to be shared by anyone. It is experienced as both something real and, at the same time, impossible to share, which is never the case with any of the other sensus communes. In the short term, it encourages the creation of the double bookkeeping system. In the long term, it weakens the intuitive connection between what is real and what is not shared.

Usually, we find no difficulty in navigating the different sensus communes because they are all grounded in the same basic shared sense of reality mediated by our bodily self-awareness (which is a point developed by Fuchs). This makes them “common” to us and other sentient beings. However, the reality of elaborated delusions arises from the disruption of this basic shared sense of reality and basic bodily self-awareness. They attempt to confer sense on experiences that are nothing like any regular experiences. Bortolotti, along with the phenomenological tradition, is, of course, right in claiming that they succeed in doing so and that they are, at first, a solution to a problem and not the underlying problem. Nevertheless, this delusional reality is disconnected from any shareable experience. It is out of this world and, by its nature, it is only made possible by the grave disturbance of the self of the patient and their relation to their external reality. It could only be compared directly to motivational delusions if it could be claimed to facilitate maintaining ties with the other shared realities. It does not seem to do this; on the contrary – the eschatological themes, hyperreflexivity, and delusions of control further disrupt the whole system of sensus communes and weaken the threads connecting all the realities. The most basic assumptions about the patient’s self, agency, and environment are weakened. Systematised and elaborated delusions can undermine everything the patient has ever believed about themself. Thus, they can damage the very possibility of there being a web of shared sensus communes. In Wittgenstein’s terms, we could say that they erase or distort some of the basic grammar that makes this web possible.

In this light, double bookkeeping can be understood as the mechanism that protects the patient from falling too deeply into a single, incompatible, delusional, isolated reality. We need to remember, however, that at the same time, double bookkeeping unfortunately helps maintain contact with the delusional reality, as it does in the case of the gaslighter. As I have mentioned when describing the epistemic harms of elaborated delusions, they are the first entries in the double bookkeeping system and the reason for its existence, and this is something costly in itself: maintaining this system is energy-consuming and ultimately epistemically detrimental. The existence of the broad circumscription around delusional reality is only beneficial if it helps, at least to a degree, to navigate the web of “regular” sensus communes. However, what makes typical elaborated delusions so suspect is that they hinder this ability, as they draw the patient deeper into the circumscribed, alienated reality. In the following section, I will examine a case study to contrast the harmful effects of most elaborated delusions with the beneficial effects of some beliefs that help put double bookkeeping to good use – for example, such that exempt certain parts of reality from the delusional projections. Such beliefs can indeed be called epistemically beneficial, even if they are false or delusional.

An instructive example of how double bookkeeping works and how it can be used to draw the patient back to the web of shared sensus communes is provided by Cermolacce et al. 2018. Cermolacce et al. refer to Gallagher’s concept of multiple realities to present the idea that there are objects – the so-called hybrid objects – that could belong to multiple realities simultaneously, thus serving as anchors helping people remain in communication despite the plurality of realities. The authors discuss an illustrative case study of KS, a 52-year-old man diagnosed with schizophrenia, who is convinced he is a religious saviour fighting a war against communists and Muslims. At the same time, KS can participate in his family’s life and the life of his community, proving that double bookkeeping tactics can successfully enable people with schizophrenia to remain in the system of sensus communes, at least to a degree. Unfortunately, KS’s realities sometimes mix, and the patient has once approached a secretary with a knife and a rosary to convert her. At the same time, KS and his psychiatric team can use an object belonging to the shared reality to facilitate communication and bring the patient back from the delusional reality. In this case, it is a telephone. KS sometimes claims he is talking on the phone with Joseph Stalin – but asked to “tell Mr Stalin that you are busy and he must call back later,” he can end this conversation, and the delusion seems to partially dissipate. “He [Stalin] must have hung down the receiver,” concludes KS (Cermolacce et al. 2018). Ending the phone call with Stalin is the first step to concentrating on the real people in KS’s life and the conversation with his psychiatrist. The telephone works both as an object in KS’s delusional reality, his hot-line to the Soviet government, and a mundane object in the shared reality.

The KS case study serves me to illustrate three points. The first is that double bookkeeping may be generally useful to an extent, but it never eliminates all the possible harm that delusions can cause. The delusional reality is also preserved and protected by this mechanism. It is crucial that the patient retains the ability to use double bookkeeping to maintain contact with the shared system of sensus communes. The second one is that it is helpful to seek anchors that facilitate shifting from one reality to another and, in particular, letting go of the delusional reality to concentrate on what is happening in the sensus communis. The final point is that although elaborated delusions that constitute the delusional reality may bring many psychological benefits and confer some meaning on the delusional reality, and they even initiate the double bookkeeping mechanism, ultimately they move the patient away from the shared reality. They cannot be epistemically beneficial.

Partaking in a reality that resembles a sensus communis but is not shared with any real people surrounding the patient cannot be beneficial for their overall ability to juggle the real sensus communes. It raises fears and suspicions about those real sensus communes, the intentions, and the sincerity of the people around, thus hampering the patient’s ability to participate in their community’s epistemic practices and achieve any kind of epistemic goals.

Even should we count explaining pathological experiences as an epistemic goal in itself (which Bortolotti doesn’t do), this goal is not realised in a way that would grant the long-term benefit of any kind. Its fulfilment may even strengthen the patient’s delusional attitude and cause more pathological experiences, as delusional beliefs are used to shed light on more and more of the shared reality. What is genuinely explained at the moment when the delusions are first adopted, and the delusional reality begins to develop is not the shared reality, nor is it any conjunction between the shared reality and the pathological experiences – it is the experiences alone. If we considered building a separate, delusional reality as a possible epistemic benefit, this would undermine any conceivable understanding of what “epistemic” means.

This does not preclude granting any merit to engaging in the delusional reality – on the contrary. We can still admire the creativity, the vibrance, and the out-of-this-world quality of the delusional realities and how they may influence our everyday outlook on our sensus communes, thanks to how they are distanced from it. This, however, does not require us to grant them epistemic benefit. It is more akin to what abstract art can offer – and we would never demand any direct epistemic benefit from this type of art.

This analysis calls for a more nuanced appraisal of the benefits of elaborated delusions. I must underline that any benefit can only be sought at the moment of adoption because, going forward, the very existence of a delusional sensus communis and the need for double bookkeeping should be considered harmful. However, according to the proposed understanding of epistemic benefit, specific delusions could still be considered initially epistemically beneficial. That would be such delusional beliefs that facilitate merging or shifting between sensus communes. An example of such a belief would be what KS thinks about particular hybrid objects, especially the telephone – “It is enough to end the call to disconnect from the conversation with Stalin”. This delusion allows the patient to focus more on the sensus communis he shares with the psychiatric team, which is beneficial. In this particular case, it turned out that some other delusional beliefs about Stalin, for example, that the Soviet dictator is concerned about the health of KS’s father, served as similar transitional beliefs that pulled KS back to the shared reality. The critical characteristic of this kind of delusional beliefs is that they bridge the gap between KS’s purely delusional reality and the shared reality, at least to some degree. They also serve as an anchor, similar to hybrid objects such as the telephone. Therefore, I would like to conclude that it is rare that systematised and elaborated delusions bring epistemic benefits. If it happens, it is only when they are infused with such elements that facilitate leaving the delusional reality in favour of the system of sensus communes. Most of them, however, should be considered detrimental in this sense.

I must acknowledge that Bortolotti is aware that it may be considered surprising to argue for the benefits of delusions as beliefs that help the patient “remain in vital connection with his/her environment” (Mishara and Corlett 2009) because they used to be “described in the philosophical literature as a departure from reality or a failure in reality testing” (Bortolotti 2016). However, it is not clear from Bortolotti’s analysis that the unique harms and the relatively rare benefits of elaborated delusions are different to practically any other of the imperfect cognitions discussed in (2020).

I believe that part of the reason why it is the similarities between motivated and elaborated delusions that come to the fore in Bortolotti’s analysis, and not the differences, is that they are treated so uniformly within the predictive paradigm. The author explicitly refers to the predictive models when she claims that elaborated delusions are something akin to a shear pin or a dam (cf. McKay and Dennett 2009): they stop the relentless propagation of error signals and thus enable the system to resume the automated learning process. However, as I have discussed in Sect. 1, it is not clear to what extent the predictive models can be applied to this type of delusions. There still seems to be much disagreement surrounding the interpretations and details of the predictive models. From what I have shown in Sects. 1 and 2, it is clear there are features of elaborated delusions that are not fully grasped by the uniform predictive approach – and exactly those features make us question their benefits.

The limits of applying the predictive models to elaborated delusions will also require our attention in the context of the NA condition, which I will be tackling in Sect. 3. At least some interpretations of the predictive account of elaborated delusions can convince us that they meet the NA condition (or some version of it). This would seem to aid their case. However, as I will show, it should cast even more doubt over their claims to be epistemically beneficial and innocent. I will now discuss the possible interpretations of the NA condition and how they are met by elaborated delusions, both in the light of the uniform predictive models and the more nuanced phenomenological ones.

4 Why the NA Condition Does Not Help

The role of the NA condition is to serve as something similar to a “mitigating circumstance.” Bortolotti acknowledges that all delusions are costly, but, as she argues, in some cases we should excuse them because, prior to their adoption, there were no available alternatives that could provide the same benefit at a lesser cost. This role of the NA condition also has important clinical implications – one of the reasons why the psychiatric team should not be hasty in undermining or negating the patient’s delusions is that they have nothing better to offer. Delusions are a temporary crutch – we should not take it away when we cannot provide better support.

Bortolotti approaches the NA condition carefully: not only does she distinguish three ways in which it can be satisfied, but she also considers whether we check for available alternatives prior to adopting the delusional hypothesis or afterwards. In our particular case, Bortolotti stresses that alternative beliefs are unavailable to the patient (for different reasons) both before and after the delusional hypothesis is adopted.

I will argue that:

a) Elaborated delusions can satisfy the “after” version of the condition;

b) We cannot ascertain whether they ever satisfy the “prior” version of the condition;

c) Regardless of (a) and (b), the NA condition in either form is not a useful criterion for epistemic innocence;

d) The NA condition can be eliminated (in favour of strengthening the EB condition); And finally, that.

e) Many of Bortolotti’s observations about the clinical implications of her analysis arevalid, although valuable modifications can be suggested.

5 A: Elaborated Delusions can Satisfy the “after” Version of the NA Condition

I would like to start with (a) because it is the most straightforward point. What happens after the adoption of the delusion? In the case of elaborated delusions, it is uncontroversial to say that they have no more alternatives after their adoption, either better or equally good. Delusions have it in their nature – by almost all accounts, including the official DSM definitions – that they make all rational counterhypotheses unavailable after being adopted. Therefore, as far as the accessibility of alternatives after adopting a belief is concerned, delusions satisfy this criterion by definition – analytically and trivially. We do not call them delusions unless they do. A fortiori, there are no better alternatives available because delusions make any alternatives unavailable. This is also reflected in the predictive processing account of delusions: the delusional belief is pathologically reinforced by the hierarchical system; it becomes an overly strong prior which is immune to updating. However, as Bortolotti agrees, this very process is in itself toxic and harmful.

To understand why, it is instructive to compare delusions to regular beliefs with regard to this version of the NA condition. Are any imperfect regular beliefs such that no better alternatives are available after the beliefs have been adopted? We should answer “no”. After all, there are plenty of epistemically beneficial alternatives to almost anything we think. Our beliefs can always be improved, and we are usually at least vaguely aware of that. We are imperfect agents that have to act in real-time in a dynamically changing environment. We adopt simplified heuristics, employ stereotypes, and accept as true such statements that we should be more inquisitive about – otherwise, our cognitive life would be impossible. It is unclear what it would even mean that alternatives to any regular belief become “not available” in an empirical sense. Let’s suppose I am in a hurry. I find a quick but oversimplified strategy to perform a specific task and cannot correct myself afterwards, which prevents me from adopting more precise and ultimately more beneficial beliefs. Can I say that, in this case, better alternatives become unavailable? Well, no. It seems they are never fundamentally beyond my reach. I could change any imperfect belief at any time if only someone presented me with a good reason to do so, and my ability to change it is beneficial to me as a cognitive agent. Being stuck with a belief unamenable to change – not so much.

I submit that the “none after” variant of the “no alternatives” condition is satisfied without any doubt only in the case of delusions (because it is in their definition). We can be sure that no alternatives are available after adopting the belief in question because the definition of a delusion entails this. At the same time, we need to note that delusions can only satisfy this version of the NA condition due to their toxic nature. Meeting this version of the NA condition makes them harmful.

6 b: We Cannot Ascertain Whether Delusions Ever Satisfy the “Prior” Version of the Condition

Point (b): the accessibility of alternatives before adopting the delusion is more complex.

We don’t know how we could ever ascertain that specific delusions adopted by the system to make sense of pathological experiences “could not have been less costly.” It is even more challenging to discuss whether or not they could have been at all different. From an empirical point of view, we don’t know enough about how it happens that we adopt particular beliefs, let alone what else could or could not have happened. And the leading accounts of delusions – the two-factor theory and predictive models – do not shed as much light on this issue as we would like them to do. As I have mentioned, after surveying the different predictive processing accounts of delusions, Parrott (2021) concludes that, so far, there is no agreement not only as to how delusions arise but also, crucially, on why they are adopted. We can sketch an account of why people with schizophrenia do not ascribe correct priors to the “obvious beliefs” that could explain their strange experiences. However, we can’t pinpoint exactly why the beliefs they end up accepting are often “explanatory nonstarters,” absurd and disturbing claims that should never have been included in the pool of candidate explanations. This is partly because, as of yet, predictive processing models cannot explain how such pools are generated. Therefore, they don’t tell us much about developing alternative beliefs (or not doing so). Parrott himself offers no fewer than three mutually exclusive suggestions. The only thing we know is that, according to predictive models, the process of belief selection, of assigning weights and correcting the whole hierarchy of predictions, is widely distributed both horizontally and vertically and remains chiefly below the level of conscious awareness.

In line with Parrott’s findings, none of the predictive models cited previously can help us understand precisely what happens before the delusional prior becomes pathologically strengthened and disjointed from other beliefs and perceptual contents. There is, however, a crucial congruence between the predictive models and the phenomenological accounts of delusion formation.

Parnas et al. (2021) emphasise that the essential aspect of the delusional belief is the absolute conviction it brings, not the actual content. The belief is not the result of typical inferential reasoning or rational deliberation – it is an immediate revelation, often called, after Klaus Conrad, “the A-ha moment.” It is originally “an affective, pathic experience with only vague meaning, but carries with it an absolute affective conviction that precedes the concretisation of the delusional content” (Parnas et al. 2021, emphasis mine). The conviction, the level of certainty, is somehow prior to the emergence of a contentful belief. This is in line with Erdmann & Mathys’s (2021) predictive model, in which the essence of the delusional prior is its warped, overly high expected certainty. Erdmann & Mathys concentrate on the later stage of delusion formation when the delusional belief becomes a strong prior, one that cannot be overcome. Their model shows why delusions are fixed and cannot be updated correctly (as other perceptions/beliefs are). The critical difference between regular and delusional cognitions is that the latter have a too high expected precision of explanations, leading to overaccomodating the data and a bias against disconfirmatory evidence. We could say, that according to this model, the single most defining characteristic of delusions, regardless of their aetiology, is the conviction with which they are held. Considered as individual beliefs, delusions are isolated and immune to updates.

On the one hand, as discussed, this is precisely what is responsible for the fact that delusions do not allow for alternatives after their adoption. On the other hand, the specific content of the delusion does not seem to be significant at all. It is chiefly the attitude, the – pathological, harmful – level of conviction that makes the delusion a delusion.

The above observation makes it all the more difficult – if not empirically and theoretically impossible – to say whether any alternatives are available before adopting the delusion. From one point of view: yes, there are because it is easy to imagine that the initial vague meaning of the delusion could have crystallised into a less disturbing belief without losing any of the conviction. From another: not really, because no alternatives are actually considered or rejected; the process of arriving at the delusional belief is incomparable with any regular type of reasoning. The second answer could allow for schizophrenic delusions to still satisfy Bortolotti’s NA condition, although, I would like to point out, in a counterintuitive way. The NA condition was meant to work similarly to how we excuseFootnote 1 specific controversial actions, such as harming someone in self-defence. Usually, when we wish to excuse such actions, we try to prove that we could not have achieved the same goal (for example, saving our life) by doing anything different or less drastic (for instance, without shooting the assailant). However, in such cases, we are still physically capable of doing something differently; we are at least half-consciously deciding to act in a certain way. In the case of adopting a delusion, we can only say that we objectively could not have chosen otherwise; no other choices have ever been available to us. Things might have ended better (we might have adopted a less costly belief), but we had no real, conscious ability to change anything. This line of defence would be akin to simply pleading insanity – which is, of course, not surprising in our case. And yet, the whole point of judging specific pathological cognitions as epistemically innocent has always been to say something more for them than that they can be tolerated because of the underlying disturbances and illnesses.

A more straightforward interpretation would be to claim that delusions satisfy the “none prior” variant of the condition purely on the trivial assumption that, if a particular delusion was adopted, then it must have been the best or the only available choice the cognitive system was ultimately capable of. We know painfully well that it would be less costly for us if our brains worked better in many respects. Unfortunately, they don’t – they are simply wired as they are and not as they are not.

However, if we followed this crude course of reasoning, it would also be prima facie impossible to prove that any belief we adopt had ever had better available alternatives or any alternatives at all. Therefore, all possible beliefs would trivially meet the “none prior” variant of the “no alternatives” condition. Yet both the philosophical and folk-psychological tradition of understanding rationality and at least some of our theoretical attempts at explaining how belief systems work, including the predictive models, suggest that this is not the way we would like to think about this. Beliefs that are false or otherwise epistemically costly are typically not acquitted of their epistemic guilt precisely because there had been better alternatives available. If having no alternatives, especially prior to adoption, is to be a requirement for achieving epistemic innocence, meeting it cannot be a trivial matter – at best, it would not be a helpful condition at all.

The tentative conclusion from the above discussion of point (b) is, therefore, threefold: depending on the interpretation, either the “prior” version of the NA condition is too trivial and too easy to be satisfied, or we cannot assess whether elaborated delusions meet it, or we can, and they do, but it seems to be a vice rather than a virtue – it undermines the point of classifying them as epistemically innocent. This already raises suspicions about the NA condition and provides some grounds for my claim that (c): it should be eliminated. Therefore, I will now develop my argument for (c).

7 c: Regardless of (a) and (b), the NA Condition in Either Form is Not a Useful Criterion for Epistemic Innocence

I believe that there is a good idea behind the NA condition: it offers us a way to accept that various beliefs should be considered beneficial despite being obviously imperfect, simply because we could not have done better. However, I think this particular way of expressing this idea goes against the deeper goals behind establishing the concept of epistemic innocence. In particular, I will now argue that satisfying the NA condition is a sign of being epistemically harmful and thus at odds with the EB condition.

The case of delusions in schizophrenia is particularly instructive. It clearly shows how the conditions are intertwined and why satisfying NA undermines the possibility of satisfying EB at the same time. Elaborated delusions can only meet the “prior” and “after” versions of the NA condition because of their most epistemically harmful quality: the fact that they carry a pathological level of conviction and (for this reason) their contents are not amenable to change, despite being partly random. Moreover, the benefit they indeed can be claimed to offer – the psychological and emotional relief brought on by a sense of meaning and clarity – is directly tied to why they are epistemologically harmful: this same pathological level of conviction. We may even suspect that the level of emotional relief is proportional to the level of epistemic harm because both are predicated on the strength of conviction. This, in turn, leads to the conclusion that there is a trade-off between the level of epistemic harm and any possible epistemic benefit that the emotional benefit could mediate.

I will now elaborate on why satisfying NA is epistemically harmful not only in the case of delusions in schizophrenia but also in general – and, thus, why NA can and should be replaced in the definition of epistemic innocence.

The critical observation is that we tend to distinguish delusions from nonpathological beliefs precisely because the former are inescapable, inevitable and based on insufficient evidence (this is engrained in the DSM definition). This puts them in contrast with regular beliefs, which are presumably rational and coherent. In the case of delusions, we are causally forced to believe certain things by a disturbance of the mind, not convinced by reasons to do so. The proclivity for delusional thinking has often been associated with a phenomenon called “jumping to conclusions” (JTC), where decisions are made despite insufficient evidence, and the connection has been empirically confirmed to some extent (see, for example, (Dudley et al. 2016; So et al. 2012). The similarity between the two phenomena has also been reflected in predictive models (Parrott 2021). The JTC phenomenon, similarly to delusional thinking, is the opposite of the careful consideration of alternatives required by our ideals of rationality.

This is because the process of choosing and eliminating alternatives is the essential mechanism of achieving epistemic benefit – it is the process of providing reasons for adopting our beliefs (and rejecting others), including their relation to existing evidence, of seeking coherence in our belief system, aiming for truth, precision, clarity and understanding. And if any translation between the traditional notion of belief and the predictive one is possible, the Bayesian models validate this view. One of the most intuitive elements of the predictive processing models of beliefs is that the system is constantly seeking the best predictive hypothesis (out of many alternative ones), one that best predicts/explains the sensory data. Specific predictions are more adequate than others – the whole hierarchical system of top-down and bottom-up error propagation is designed precisely to help us generate and assert the best possible ones out of the candidate set. At the end of the process, we adopt predictions we no longer question (or, at least, such predictions that only a strong error signal would force us to revise). We restrict the candidate pool as much as possible and as much as we need. Thus, ideally, we have no (or very few) alternatives left, but not because we had had none to begin with. It is crucial that we have carefully eliminated them in the process of often interactive, evidence-based, empirically grounded reasoning, or, to describe it in the predictive framework: we have been able to perfect our predictions in such a way that they are as precise as possible: we only have a limited set of hypotheses with relatively high probabilities and future perceptions do not cause much surprise. At the same time, we can modify those hypotheses accordingly, should something unpredicted occur. Despite being assigned low probabilities, our alternatives remain at least partly available. We can discuss them with others, explain why we rejected them, and, should new evidence emerge, re-examine them and ultimately even change our minds. Naturally, this is a highly idealised view of cognition – we don’t need to actually meet this kind of robust rational requirement on an everyday basis whenever our predictive mechanisms generate simple explanations of perceptual input, and our mechanisms implement various shortcuts and simplifying strategies. In addition, an enactivist would strongly emphasise the role of constant sensory interaction with the environment, the so-called active inferences, and engaging with other agents. For my argument, it is essential to note that the process of eliminating worse beliefs and adopting better ones is crucial to our system of reasoning or, at least, to our understanding of it.

Furthermore, this ability to engage with other epistemic agents in the process of choosing and reassigning probabilities to our beliefs is, of course, essential to our ability to participate in the sensus communis and, even more importantly, in the whole system of sensus communes. Retaining at least some alternatives to our beliefs is an obvious way to leave various open gateways between different sensus communes. Whenever we adopt a belief without having any available alternatives, we undermine our ability to participate in the epistemic practices of our community because they typically require us to be able to consider alternatives. True, each sensus communis is built around certain shared beliefs that are not typically undermined. Still, they are not immune to being questioned because they had never had any alternatives, but because the community has somehow agreed on which alternative beliefs are the most likely. Whoever wishes to undermine any of those beliefs or introduce a new controversial or counterintuitive one needs to be prepared for a long debate. I am not claiming, by any means, that the epistemic activity of a typical community is rational in the traditional, idealised, philosophical sense and that alternatives are being accepted or rejected on the sole basis of the strength of arguments. I am only observing that participating in any epistemic activity with our peers requires at least some rudimentary cognitive flexibility, some basic ability to discuss why we accept or reject certain beliefs. This, in turn, requires us to have at least some notion of the alternatives, even if we are highly confident that they are wrong.

Aside from specific religious contexts (which are typically firmly grounded in a concrete type of sensus communis), presenting any belief as having no conceivable alternatives at all raises doubts rather than generates understanding. This is particularly poignant in the case of delusions because their content is typically challenging to accept for anyone in our community.

In Sect. 2, I have already explained how elaborated delusions force the agent to engage in double bookkeeping and isolate them from their sensus communes. I have now shown how this is connected to the fact that they satisfy the NA condition. I am not claiming that every belief that satisfies the NA condition must have such dire consequences – I would only like to point out that meeting the NA condition should raise our suspicions that the belief is epistemically harmful and possibly not epistemically innocent. Any belief without alternatives disrupts our ability to participate in a sensus communis – it is just that some are more harmful than others. In the case of monothematic, narrowly circumscribed delusions, the benefits can outweigh the costs because the disruption is isolated. In the case of elaborated delusions, whose content influences our whole web of beliefs to a higher degree and typically concerns the very basis of our understanding of ourselves and the world, the costs are skyrocketing.

This is why I propose to (d): leave the NA condition aside.

8 d: The NA Condition can be Eliminated

Do we need something in place of the NA condition? I believe that, in fact, not: the revamped EB condition is enough to do the job. As I remarked, the sound idea behind the NA condition was that we sometimes need to appreciate even very imperfect cognitions simply because we could not have done better. Sometimes a very harmful cognition is the only possible way of achieving a higher epistemic goal – or the least harmful way. I will now show that this idea is preserved by my version of the EB condition proposed in (Białek 2021).

The new version of the EB condition encourages us to treat all cognitions as beneficial if they improve our ability to participate in the shared epistemic practices of our community. Still, it also requires that they be necessary. This second part of the condition is precisely what we need. Any harmful cognition could still qualify as beneficial if it turns out that it has genuinely been necessary to support our epistemic functioning.

It may seem that this definition is too strong to be applied in general to non-pathological beliefs, such as do not satisfy Condition 1. I want to emphasise that this definition does not force us to classify any such beliefs that are not necessary for our epistemic practices as harmful – they may be neutral. In fact, there are many redundant or useless beliefs that do not do us any epistemic harm (possibly apart from the toll they might take on our cognitive system). They are, however, also not beneficial – for any beneficial belief, we should be able to find at least a single epistemic practice for which it could be considered necessary, any fruitful exchange within our sensus communis.

Finally, in assessing epistemic innocence, we no longer specify that no alternatives had been available. Instead, we determine that this cognition cannot be harmful in the worst way imaginable: it cannot, on the whole, alienate us from our sensus communis.

I have shown both why I believe that elaborated delusions are the epitome of epistemic guilt and what changes can and should be made to the conditions of epistemic innocence in light of this observation. In the last part of this section, I would like to argue that this diagnosis does not undermine the clinical usefulness of Bortolotti’s claims about elaborated delusions. However, I believe that some practical suggestions can be framed differently.

9 e: Many of Bortolotti’s Observations About the Clinical Implications of her Analysis are Valid, Although Valuable Modifications can be Suggested

The first issue to tackle is how this changed appraisal of delusions in schizophrenia – as being neither epistemically beneficial nor innocent – affects any suggestions for dealing with them in clinical practice, as this was one of the most important motivations behind Bortolotti’s original work. Diagnosing elaborated delusions as innocent could help promote a more positive attitude towards them in clinical practice. In particular, it could discourage both medical personnel and patients’ loved ones from treating delusions as simply false beliefs that could and should be challenged and dismissed. I would not like to undermine Bortolotti’s general suggestions that a flexible approach to delusions is needed, that it is not advisable to attempt to correct them and that, in some cases, they should be acknowledged as genuinely helpful. I only intend to nuance it a little.

As I argued, there is one crucial reason why elaborated delusions in schizophrenia should be considered harmful even despite the psychological and emotional benefits they may provide: they undermine the ability to shift between our various realities and participate in our sensus communes. However, this is caused not by their specific content but by the toxic level of certainty they bring about. This directly harms the patient’s ability to function in their community, including their therapeutic alliance with the medical staff. Therefore, even though this certainty is also responsible for emotional benefit, it can be safely assumed that it is generally detrimental and any pharmacological and therapeutic means that can enhance the flexibility of patient epistemic functioning, their ability to participate in multiple sensus communes and to choose between many available alternatives, can only be encouraged.

Nevertheless, my interpretation would not undermine in any way Bortolotti’s suggestion that the specific claims made by the person suffering from delusions need not necessarily be challenged and, if at all, not in a dismissive manner. This, in itself, seems not only futile but possibly detrimental to the process of restoring the patient’s epistemic freedom to consider alternatives and shift between realities.

On the contrary, let us go back to the clinical case study of the patient who claimed, among other things, that he was in contact with Stalin. His doctor was able to bring him back, at least partially or temporarily, to his sensus communis not by challenging the idea of talking to Stain but by not doing this. Instead, the doctor built a bridge over which the patient could cross back to their shared reality. In this particular case, the doctor used a hybrid object, which was in itself an accessible doorway between the realities. Another practical suggestion that could be devised based on my changed assessment of elaborated delusions is to seek out such helpful ways of restoring the patient’s ability to partake in multiple sensus communes. They are not confined to physical or abstract objects that exist simultaneously in many realities, although those are of particular use. Any way to restore the patient’s ability to consider alternatives and shift between realities could be helpful, even if it does not undermine any of the key delusional beliefs of the patient. The above analysis of the case study reinforces Bortolotti’s earlier intuition that many delusions can actually be clinically helpful. The idea that a conversation with Stalin can be ended is as delusional as the idea that it has ever started – the former, however, is clearly beneficial, even in the purely epistemic sense. It has turned out to be the necessary condition of possibility for the patient’s epistemic functioning. As shown in (Białek 2021), the new EB condition can still be satisfied by motivated delusions – and I have now argued both in Sect. 2.3 and above that it can even apply to some of the elaborated delusions encountered in people with schizophrenia, albeit such cases are very rare. The new condition helps us see why only some very specific elaborated delusions can be deemed epistemically innocent, while most are purely detrimental.

10 Do we have Anything Better than Innocent Beliefs? The Irrationality Condition

In the final section of this paper, I will briefly touch upon the first condition of epistemic innocence proposed by Bortolotti – the condition of epistemic irrationality. Initially, the concept of epistemic innocence was meant to absolve such cognitions that are evidently epistemically blameworthy for some of the traditional reasons (because of the way they are impervious to evidence, groundless or somehow detached from other beliefs and behaviours, for example). It is interesting to ask whether we really should – or even could – separate irrational, apparently flawed cognitions that could only qualify as “innocent” from those that could aspire to be beneficial without having to go through the trial at all. This is especially interesting in view of the fact that many authors, Bortolotti and myself included, seem to be leaning toward the view that there are no clear cut lines between “irrational” and “rational” beliefs and that we should expect at least some blurriness, if not continuity, between the two categories. I believe, however, that retaining at least a sketchy distinction between them is intuitive and valuable for Bortolotti and her colleagues’ practical purposes. There are good reasons why delusions or confabulations are considered with much suspicion and seem to require absolution, while everyday imprecise or slightly incoherent beliefs are not even put on trial. Although ultimately maybe only very few beliefs are perfectly rational in the traditional sense, and most have some faults, certain imperfect cognitions such as confabulations or delusions are so imperfect that they are particularly costly for the epistemic agent, and should invoke mistrust. Therefore, I do not wish to argue that the first condition should be dropped altogether or that the term “irrational” is too vague to be used at all. Nevertheless, I would like to observe that examining the concept of epistemic innocence without confining our interests to obviously suspicious beliefs which we easily categorise as “irrational” leads to exciting results.

There are two tempting issues. First: if there is a continuity between strongly “irrational” beliefs such as delusions and regular imperfect beliefs, “innocence” may be the best that any regular belief could achieve. And second: can any actual human belief ever achieve anything more than epistemic innocence?

Regarding the first issue, another analogy with our legal and moral practice springs to mind. We usually only judge people when they are in the wrong (or suspected to be so). The best we can typically achieve in our everyday life is, therefore, to be innocent. Or as innocent as possible. Sometimes fate places us in such dire circumstances that we are forced to do something objectively wrong, for example in self-defence – but we can still maintain our innocence if we can prove that this was the best we could have done. This does not make the deed morally good; it makes it excusable. At the same time, most of our actions are not judged at all, they do not draw anyone’s attention because they have no spectacular consequences, particularly bad ones. Most of them are nothing special, though. They bring some slight benefits and minor costs and harms, they are also usually imperfect at least to some degree, and, in the grand scheme of things, they may be regarded as neutral. “Neutral”, and not “innocent” only because they haven’t attracted enough attention to be judged. They did not raise any suspicions and thus both their minor faults and the minor benefits they bring go unexamined. The simple fact is that they did not raise any suspicions because they did not cause too much harm.

In this light, let us now think of delusions and confabulations and their relation to regular beliefs. The former are markedly more suspicious than the latter – just like our violent behaviour in self-defence attracts much more attention than our regular conduct. This is why delusions and confabulations become subject to trial and may be judged as innocent only if they satisfy our conditions. Most of our regular beliefs, however, are also imperfect; they do not bring any spectacular benefits, they are faulty in many ways, and they always incur some costs. We assume their neutrality and do not ask about their innocence only because those costs and faults are not as prominent as in the case of the suspicious beliefs, the ones that begin their trial categorised as “irrational”. We should hope, however, that most of the regular beliefs could also fulfil the EB condition for epistemic innocence – we just don’t examine if they do because we do not feel they fulfil the irrationality condition. This analogy between actions and beliefs inspires an attractive, minimalist way of thinking about rationality: the blurry line between what is classified as “rational” or “irrational” has chiefly to do with “behaving suspiciously”. In the case of actions, our suspicions are raised whenever we notice harm being done – in the case of beliefs, it is epistemic harm. Our everyday beliefs do not raise suspicions simply because they do not cause too much harm.

A modest understanding of rationality would be, thus, in perfect symmetry with our legal and moral approach to actions. We do not demand much of any agent or belief: just that they do not do anything obviously harmful. Until that happens, there is no trial, and we do not examine the innocence of anyone. We do realize, however, that our practical distinction between “suspicious” irrational beliefs and the ones that we choose not to put on trial is just a rule of thumb. Should we look closer, any belief is, in fact, imperfect and faulty at least to a degree. This is why, should we look closer, we might wish to examine any and all beliefs just as we examine confabulations and delusions. Then, the best they could ever do would be to fulfil our condition of epistemic benefit and achieve the status of innocence.

Finally, this modest understanding of rationality inspires my second question: can any belief do better than to be declared innocent? As far as our actions go, we sometimes feel the need to praise certain deeds as good, or even heroic and supererogatory. Some actions seem to bring significantly more benefits than costs and harms, and this can obviously also be true of beliefs. We can use the EB condition to gauge the benefits they bring and some may prove themselves much more fruitful in maintaining and building new, shared sensus communes than others. And perhaps we can even enjoy rare “heroically rational” beliefs – such that have been given more careful consideration than usual, that have required special epistemic effort, a profound revolution in our conceptual scheme, a thorough updating of our entire belief system, a painful revision of deeply engrained opinions and meticulous control of all kinds of bias. It is simply essential to realise that such heroic beliefs are as rare as are heroic deeds. Even though most traditional ideas of rationality would require us to perform at this level every day. I think that this might be a heretofore underappreciated lesson from the epistemic innocence literature: perhaps achieving epistemic innocence is the best most regular beliefs can hope for.