Belief and Prediction: Wittgenstein’s Philosophy of Religion in Dialogue with Predictive Processing

Abstract

The present article aims to offer a new perspective on later Wittgenstein’s notion of belief by comparing the ideas in two well-known works of his, namely, On Certainty (OC) and Philosophical Investigations (PI), with ideas from a widely recognized theory of brain function within the field of neuroscience known as predictive processing (PP). The purpose of this comparison is to demonstrate how Wittgenstein’s notion of belief transcends the conventional cognitive/non-cognitive dichotomy when examined alongside the PP framework—a model that, by integrating action with perception, already obscures the sharp boundary between these two categories. Employing a comparative philosophical approach, this study brings together insights from PP theory and Wittgenstein’s ideas on belief, exploring how PP could support and, in certain respects, shed light on Wittgenstein’s ideas about religion. One central point of the present analysis is to challenge the claim that Wittgenstein’s philosophy of religion isolates religious discourse as uniquely evidence-resistant and/or evidence-repellent. By reading Wittgenstein’s ideas on world image and form of life through the lens of PP—with a special focus on the notion of self-evidencing—it is shown that the notion of belief, whether religious or not, in both Wittgenstein’s later philosophy and PP, is as much about being as it is about believing.

1. Introduction

In his analysis of Wittgenstein’s ideas on religious language, Michael Scott, mainly focusing on Wittgenstein’s Lectures on Religious Belief and Culture and Value, addresses the question of why Wittgenstein’s insights into religious language, up until the present time, “have endured but not prevailed” (Scott 2018, p. 164). Scott finds the appeal of Wittgenstein’s ideas on religious language in its highlighting of the distinctiveness of religious discourse in comparison to other areas of discourse, or, more specifically, to scientific and historical discourses. According to Scott’s reading of late Wittgenstein, whereas propositions from scientific and historical discourses are grounded in evidence-based facts and, as such, are truth-apt, propositions from religious discourse, on the other hand, resist face-value assessment (assessment according to face value theory1) due to their indifference to empirical evidence, or, in other words, due to the irrelevance of empirical evidence to the enterprise of religious faith. Regarding the latter point, Scott argues by referring to quotes from Wittgenstein’s Lectures on Religious Belief, including the following (ibid., pp. 157–58):

“The point is that if there were evidence [for religious belief], this would in fact destroy the whole business.

Anything that I normally call evidence wouldn’t in the slightest influence me.

Suppose, for instance, we knew people who foresaw the future; make forecasts for years and years ahead; and they described some sort of a Judgement Day. Queerly enough, even if there were such a thing, and even if it were more convincing than I have described but, belief in this happening wouldn’t be at all a religious belief”.

(Wittgenstein 1967, p. 56)

According to Scott, the appeal of Wittgenstein’s philosophy of religion for many scholars lies thus primarily in its assertion that religious language is not subject to face-value evaluation. This approach to religion—which is also labeled as Wittgensteinian fideism (Nielsen 1967; for critical engagement with this view, see for example, Addis 2001; Schönbaumsfeld 2023)—has been very well received by non-cognitivist scholars of religion, who treat religious discourse as less about propositional beliefs and cognitive states than about contingent forms of expression of human motivations and commitments manifested in their way of living (see, Scott 2018, pp. 159–60).

Nevertheless, one might say that claiming that Wittgenstein’s philosophy necessarily entails a non-cognitivist approach to religion contradicts the merely descriptive task of his philosophy claimed by Wittgenstein himself with the following words: “Philosophy simply puts everything before us, and neither explains nor deduces anything” (PI sct. 126). Scott, however, views this as contrary to what Wittgenstein’s philosophy actually does. According to him, Wittgenstein’s ideas on religion in the two aforementioned works reflect his contentment with the idea of the particularity of religious language as characterized by judgments that stand at odds with empirical evidence or, as he puts it, that are both evidence-resistant and evidence-repellent. At the same time, Scott argues, the lack of direction and theorization in Wittgenstein’s remarks on religion as well as the fact that neither evidence resistance nor evidence repellence is unique to religious discourse—being instead common characteristic of all kinds of commitments resulting from “a broader phenomenon of cognitive dissonance”—has led his philosophy of religion, despite its appeal, to be less persuasive than it initially appears (Scott 2018, pp. 163–64).

Nevertheless, in his reconstruction of Wittgenstein’s later philosophy of religion, Klaus von Stosch, drawing on the general context of his later philosophical works, views in Wittgenstein’s reluctance to formulate a formal theory of religion a coherent philosophical approach that neither reduces religious language to its regulative, non-cognitive aspect nor burdens it with truth claims isolated from their practical context, as often expected by analytic philosophers of religion. Von Stosch argues that, for Wittgenstein, the referential character of certain religious beliefs, that is, their referring to external realities, was not only obvious but also trivial. What makes Wittgenstein’s account of religious belief remarkable is its emphasis on “the expressive and regulative components of religious belief without forgetting its cognitive elements” (Von Stosch 2010, p. 122; also see, pp. 127–31). According to von Stosch, Wittgenstein does not simply advocate “a cognitive or a non-cognitive approach to religious belief, but he shows that you always have to be aware of both these elements of belief if you want to understand something” (ibid., pp. 122–23).2 Already, John Churchill has shown that the conception of belief, in general, in Wittgenstein’s later philosophy transcends the simple cognitive/non-cognitive distinction. Churchill regards the difficulty in categorizing Wittgenstein’s views on religious belief within a cognitive/non-cognitive dichotomy as lying less in an inconsistency in Wittgenstein’s treatment of religious discourse than in his general philosophical project that aimed at challenging—or, one might say, with Churchill, attacking—the very concepts of belief that held such a sharp distinction between cognitive and non-cognitive (Churchill 1984, p. 139; also see, Schönbaumsfeld 2023). I tend to agree with Churchill, primarily because determining whether Wittgenstein’s approach to religion is cognitivist or non-cognitivist is not derivable from his own philosophy. This is due to the fact that conventional categorizations as such do not apply to his fluid, unsystematic account of cognition, behavior, and belief, the complex structure of which he explains through his riverbed metaphor (OC sct. 96–97).3

Now, if this is the case, one could argue that there is nothing essentially distinctive about religious discourse that would make it immune, in contrast to other areas of discourse, from philosophical criticism, whatever is meant by “philosophical” in each and every language game.

The present article aims to offer a new perspective on later Wittgenstein’s notion of belief by comparing the ideas in two well-known works of his, namely, On Certainty (OC) and Philosophical Investigations (PI), with ideas from a widely recognized theory of brain function within the field of neuroscience known as predictive processing (PP). The purpose of this comparison is to demonstrate how Wittgenstein’s notion of belief transcends the conventional cognitive/non-cognitive dichotomy when examined alongside the predictive processing framework—a model that, by integrating action with perception, already obscures the sharp boundary between these two categories. Although predictive processing is ultimately categorized as a theory of cognition itself, its conceptualization of cognition, as we will see, resonates very well with Wittgenstein’s emphasis on the regulative-expressive dimension of belief.

Already, Judith Wolfe, in her analysis of the role of perceptual and imaginative processes in our engagement with the world in general and our theological understanding of reality, has highlighted the relevance of the predictive processing paradigm for studying religious faith and religious language. She introduces predictive processing as a paradigm that has had the potential to concretize some of her phenomenological observations “by describing perception as a constant negotiation between ‘bottom-up’ input and ‘top-down’ priors, which is operative at all levels of engagement with the world” (Wolfe 2024, p. 10). She then proceeds to suggest a link between predictive processing’s account of cognition and later Wittgenstein’s remarks on aspect perception (ibid., pp. 76–79)—a topic to be discussed also in the present article. Finally, I should add that my approach to predictive processing theory throughout this article, similar to that of Judith Wolfe’s, focuses less on its scientific and technical aspects than on its descriptive philosophical function. This philosophical approach to predictive processing allows for a comparative philosophical analysis that the present article undertakes by bringing together insights from PP and Wittgenstein’s ideas on belief, providing a foundation for meaningful dialogue between the two discourses, and ultimately offering a new perspective on Wittgenstein’s philosophy of religion through the lens of the predictive processing paradigm.

2. The Predictive Mind and Its Language Games

In his significant work of 2016, Surfing Uncertainty, Andy Clark has described predictive processing (PP) as an account of brain function that grounds cognition in “a complex looping interplay between perception and action” (Clark 2016, p. 63). In our normal everyday engagement with the world, it is this complex interplay between action and perception that shapes the model of the world (generative model) that we develop in our minds. By “looping interplay”, Clark means to highlight the fact that, within the PP framework, our cognitive engagement with the world is not understood as a unidirectional process in which sensory input is simply assimilated into a generated model of the world (bottom-up processing). Instead, our model of the world actively generates expectations and forms predictions around sensory input, and the top-down imposition of these predictions on our experience shapes our actions in the world—actions that would either align with and confirm the model or revise it through error correction, thereby establishing a dynamic feedback loop between perception and action. Clark speaks of this interplay also in terms of “the delicate dance between the top-down and the bottom-up” (ibid., p. 57; Clark 2013, p. 189). Philosophers engaging with PP often trace its philosophical roots back to Immanuel Kant, mediated through the theory’s direct influence from Hermann von Helmholtz. It is very common among these philosophers to compare PP’s account of the active, predictive mind with Kant’s conception of cognition grounded in the synthesis of the spontaneity of concepts (top-down processing) and the receptivity of sensory impressions (bottom-up processing) (Swanson 2016; Wiese and Metzinger 2017; also see, Hohwy 2013, pp. 5–6). Nevertheless, despite this comparison with Kant, it is important to note, as Thomas van Es and Erik Myin have argued, that PP does not necessarily entail a representationalist approach to cognition. On the contrary, they have shown that this model of cognition, with its emphasis on the significance of the cognitive agent’s sensitivity to the facts of the environment and reliance on the statistical regularities, aligns more with a non-representationalist account of cognition (Van Es and Myin 2020). With this brief introduction to PP, I will now turn to the philosophy of later Wittgenstein to explore the possibility of a meaningful comparison between his philosophical project and PP.

In a passage from his Philosophical Investigations, Wittgenstein presents the example of a disappearing and reappearing chair to emphasize the consistency of experience and the predictability of the world as the conditions for the possibility of a meaningful application of concepts to objects of experience. The passage reads:

“I say ‘There is a chair’. What if I go up to it, meaning to fetch it, and it suddenly disappears from sight?—“So it wasn’t a chair, but some kind of illusion”.—But in a few moments we see it again and are able to touch it and so on.—‘So the chair was there after all and its disappearance was some kind of illusion’.—But suppose that after a time it disappears again—or seems to disappear. What are we to say now? Have you rules ready for such cases—rules saying whether one may use the word “chair” to include this kind of thing? But do we miss them when we use the word “chair”; and are we to say that we do not really attach any meaning to this word, because we are not equipped with rules for every possible application of it?”

(PI sct. 80)

The point of the cited remark is that the attribution of a meaningful word (concept) to an object of experience depends on the possibility of present as well as future applicability of that concept within a coherent context in which the experience takes place. So, when Wittgenstein speaks of fetching the chair as an act based on the conviction that “there is a chair”, the predictability of the successful performance of this act of fetching (and subsequently sitting on the chair) serves as a regulative condition for the meaningful reference to the object as a chair. If our predictions as such consistently fail, the regulative frame of reference collapses as we are not equipped with rules in the absence of regularities.

In another passage, Wittgenstein refers to the normality of cases of experience, which, as his description of normality suggests, means the predictability of those cases, as the condition for the possibility of our language games to function:

“It is only in normal cases that the use of a word is clearly prescribed; we know, are in no doubt, what to say in this or that case. The more abnormal the case, the more doubtful it becomes what we are to say. And if things were quite different from what they actually are—if there were for instance no characteristic expression of pain, of fear, of joy; if rule became exception and exception rule; or if both became phenomena of roughly equal frequency—this would make our normal language-games lose their point.—The procedure of putting a lump of cheese on a balance and fixing the price by the turn of the scale would lose its point if it frequently happened for such lumps to suddenly grow or shrink for no obvious reason”.

(PI sct. 142)

Wittgenstein’s above remarks, which tackle the question of the conditions for the possibility of “our normal language-games”—conditions that are closely related to cognition and certainty—raise the question of whether his philosophical investigations on language games share parallels with Kant’s transcendental philosophical account of cognition. This question is particularly significant for the purpose of the present article due to the fact that, as already mentioned, predictive processing (PP) is often conceived as carrying Kant’s legacy.

To address this question, I would like to begin by revisiting section 126 of Philosophical Investigations. In this passage, as previously noted, Wittgenstein emphasizes the purely descriptive role of philosophy, framing it as a business that “neither explains nor deduces anything”. Following this assertion, Wittgenstein provides a second definition of philosophy that is less frequently cited than the preceding one: “One might also give the name ‘philosophy’ to what is possible before all new discoveries and inventions” (PI sct. 126). Based on this second description, one could consider within the framework of Wittgenstein’s thought an adequate space for philosophizing which makes the talk of language games possible in the first place. Moreover, in the same work, Wittgenstein describes his own philosophical investigation—to which he also gives the title “grammatical investigation”—as being mainly concerned with the question of the possibility of the phenomena, rather than with the phenomena as such (PI sct. 90). All this brings into mind a type of transcendental philosophical approach to the question of human experience as tackled also by Kant, who also searched for the conditions for the possibility of knowledge, prior to any concrete example of knowledge. In his discussion on the plausibility of comparison between Kant and Wittgenstein, Klaus von Stosch, also referring to section 90 of Philosophical Investigations, has pointed out that as far as Wittgenstein’s philosophical investigations into the limits and conditions of human knowledge are concerned, one could recognize some transcendental philosophical tendencies in his treatment of the related topics (Von Stosch 2001, pp. 169–75)—if transcendental philosophy is defined, above all, by its central theme, namely, its focus on the question of the possibility of knowledge as such, rather than the type of knowledge in question and the responses given to that. Leslie Stevenson even went so far as to attribute a sense of transcendental deduction to Wittgenstein (Stevenson 1982). In any case, it is important to note that the connection between Wittgenstein and Kant is evident when considering that Wittgenstein addressed similar foundational questions to those that shaped Kant’s critical philosophy, namely, questions concerning the conceptualizability of experience and the limits of knowledge. This Kantian link offers a gateway for exploring the extent to which Wittgenstein’s responses to these questions correspond with those of PP with its presumably Kantian roots.

Perhaps, as Judith Wolfe’s original observation also suggests (see, Wolfe 2024, pp. 78–79), nothing could demonstrate the parallels between Wittgenstein’s conception of perception and PP as effectively as his famous duck–rabbit example. In Philosophical Investigations, Part II, Wittgenstein presents an illustration of a duck–rabbit, that is, a figure that can be seen as either a duck or a rabbit. He then provides the following scenario to show how our recognition of the figure as any of the two is shaped by the contextual framing in which it is presented:

“I see two pictures, with the duck–rabbit surrounded by rabbits in one, by ducks in the other. I do not notice that they are the same. Does it follow from this that I see something different in the two cases?—It gives us a reason for using this expression here”.

(Wittgenstein 1953, p. 195^e)

The phenomenon that Wittgenstein describes in the above passage is explained4 by predictive processing (PP) in terms of a mechanism known as surprise minimization or, more technically, prediction error minimization (PEM). According to PP, cognition by any intelligent being involves minimizing the discrepancy between that being’s internal model of the world and the external world (Parr et al. 2022, p. 25f), or more particularly, the error between the incoming sensory data (bottom-up signal) and the top-down predictions based on prior hypotheses (Hohwy 2013, p. 42ff).

Wittgenstein describes this process inversely, noting that surprise arises when, after having recognized the figure as either a duck or a rabbit—depending on the context, one subsequently becomes aware of the alternative aspect of the figure:

“…If I later see the aspect change—can I say that the duck and rabbit aspects are now seen quite differently from when I recognized them separately in the tangle of lines? No.

But the change produces a surprise not produced by the recognition [Erkennen]”.

(Wittgenstein 1953, pp. 199, 199^e)

One might say, with an eye on predictive processing, that the surprise was not “produced” by the recognition, perhaps, because the very act of recognizing the figure was concomitant with minimizing surprise as effectively as possible. Nevertheless, one should be aware of the fact that Wittgenstein’s main point of concern is not the unconscious activities of the brain in a mere act of seeing, as he makes it very clear that the type of seeing that he is interested in is “noticing an aspect” (Wittgenstein 1953, p. 193^e). One could, however, think of Wittgenstein’s notion of “seeing” discussed here as a more complex case of cognition (i.e., recognition/erkennen) that is already included in the predictive processing scenario. Moreover, the very core idea of predictive processing is that perception, even at its most basic level, is always concomitant with the projection of prior beliefs onto the context in which perception is taking place and thus always involves an aspect of the perceived that corresponds to the generative model. This is done by the agent either through selecting aspects that fit the generative model or through acting upon the context to change it in a way that makes the bottom-up signal fit the top-down prediction—an act that, ultimately, could also result in the correction of the generative model if necessary. This two-fold process of cognition, in which perception and action work together toward the same goal of surprise-minimization, is referred to as active inference in predictive processing (Parr et al. 2022, p. 24f).

That said, it is important to remember that when Wittgenstein speaks of concepts, he is primarily concerned with concepts as linguistic units, namely, meaning-bearing words (see, for example, PI sct. 208). For Wittgenstein, thus, the above-explained act of Erkennen involves recognizing objects under specific concepts known to us through language. Regarding object recognition, he specifically writes: “…what is at issue is the fixing of concepts. A concept forces itself on one” (Wittgenstein 1953, p. 204^e). This statement, one might say, captures then the unconscious business of the mind of receiving concepts that force themselves upon us through our exposure to certain contexts. But what about the action part? Does this mean that we are simply passive receivers of impressions that force concepts upon us? In order to respond to this challenge, one simply needs to recall one of the most significant aspects of Wittgenstein’s philosophy of language, that is, his description of meaning as use:

“For a large class of cases—though not for all—in which we employ the word ‘meaning’ it can be defined thus: the meaning of a word is its use in the language.

And the meaning of a name is sometimes explained by pointing to its bearer”.

(PI sct. 43)

Or, as put in his On Certainty:

“…A meaning of word is a kind of employment of it. For it is what we learn where the word is incorporated into our language”.

(OC sct. 61)

For Wittgenstein, meaning is thus something that the mind actively gives to words in specific contexts based on previous experiences of how those words have been used in those contexts. It is a rule that has to be followed if one is to play a language game properly:

“That is why there exists a correspondence between the concepts ‘rule’ and ‘meaning’”.

(OC sct. 62)

“If we imagine the facts otherwise than as they are, certain language-games lose some of their importance, while others become important. And in this way there is an alteration—a gradual one—in the use of the vocabulary of a language”.

(OC sct. 63)

“When language-games change, then there is a change in concepts, and with the concepts the meanings of words change”.

(OC sct. 65)

This shows that in Wittgenstein’s conception, the human person is not in an unalterable fixed relationship to the language games accessible to her, but the power of imagination plays a significant role in shifting the meaning of the vocabulary and altering the language games.

Wittgenstein’s description of meaning as use applies not only to words and concepts but also to propositions. It is important to remember, at the same time, that in Wittgenstein’s philosophy—particularly in On Certainty—propositions are usually treated as expressions of convictions. And this treatment expands the scope of his philosophical investigation to include debates on doubt and certainty.

It is in the same work that Wittgenstein introduces his concept of world image (Weltbild, also translated as “picture of the world” or “world-picture”). For Wittgenstein, the world image consists of the convictions that shape one’s orientation in the world; it is the handed-down (überkommene; also translated as “inherited”) background against which one distinguishes between true and false (OC sct. 94). It is important to note that world image, according to Wittgenstein, does not have a merely epistemic function. More importantly, as emphasized by his riverbed metaphor, it serves—similar to the generative model in PP—a regulative role (top-down processing?) in guiding human behavior and thought. It operates, as his game metaphor suggests, without requiring explicit awareness of the underlying convictions informing those behaviors:

“The propositions describing this world-picture might be part of a kind of mythology. And their role is like that of rules of a game; and the game can be learned purely practically, without learning any explicit rules”.

(OC sct. 95)

This statement is significant because it directly links Wittgenstein’s discussion of certainty to his philosophy of religion. If the propositions in mythology or religion(?) serve as the criteria for truth and falsity from the outset, how can we evaluate the plausibility of those propositions in the first place? Does this approach to world image not lead to the notorious Wittgensteinian fideism? Especially when he further describes the propositions defining the world image with the following words:

“That is to say, the questions that we raise and our doubts depend on the fact that some propositions are exempt from doubt, are as it were like hinges on which those turn”.

(OC sct. 341)

Moreover, considering the fact that Wittgenstein describes world image as inherited or handed-down, it seems that in his philosophy, there is little room for the individual to play a significant role in shaping one’s own world image. He emphasizes that one does not get or have one’s picture of the world by having convinced oneself of its correctness (OC sct. 94). Instead, it is handed down to the individual, which makes it a fundamentally social construct. This has indeed been considered one of the most crucial points of contrast between Wittgenstein’s philosophy and predictive processing. One can see this explicitly stated in the last pages of Hohwy’s Predictive Mind, where he writes:

“In Western intellectual history there is wide agreement that conscious experience is private, subjective, first person perspectival, shielded from the view of others. There are philosophers, beginning mainly with Wittgenstein (1953), and other researchers who for various reasons disagree with this. I think they are wrong and thus side with the historical majority view”.

(Hohwy 2013, p. 249)

The actual landscape of Western intellectual history, however, may reveal less consensus on this issue than suggested above. This becomes apparent when considering, for instance, that already in the 18th century, Fichte had viewed the concepts of summoning (Aufforderung) and recognition (Anerkennung) by other human beings as playing a pivotal role in establishing the foundation of individual (self-)consciousness. The existence of other human beings is thus central to his transcendental deduction of human conscious experience, as he asserts:

“The finite rational being cannot ascribe to itself a free efficacy in the sensible world without also ascribing such efficacy to others, and thus without also presupposing the existence of other finite rational beings outside of itself”.

(Fichte [1797] 2000, p. 29)

He goes even so far as to say that,

“The human being (like all finite beings in general) becomes a human being only among human beings; and since the human being can be nothing other than a human being and would not exist at all if it were not this—it follows that, if there are to be human beings at all, there must be more than one”.

(ibid., p. 37)

So, if predictive processing is to be considered a candidate for a “grand unifying theory of mind and brain function” (Corlett et al. 2020, p. 529), as some of its proponents usually claim, it ought to account for a fundamental aspect of human life: the necessity of recognizing, and being recognized by, other human beings as active agents equal to oneself. In addition, it will be attempted to show in the following section, through an exploration of the notion of self-evidencing, that there is indeed the potential to include this aspect of human consciousness into the predictive processing paradigm. The same examination, taking place in the playground of PP, reveals that Wittgenstein’s remarks on the world image, as well as his association of language games with forms of life, do not necessarily lead to fideism when these remarks are applied to religious discourse. This analysis requires first, as said above, an exploration of a widely discussed concept in PP, namely, self-evidencing, which is a task undertaken in the following section.

3. The Self-Evidencing Agent and Its Form of Life

If prediction error minimization has to do with minimizing surprise, self-evidencing, as the other side of the story, mainly involves maximizing evidence in support of one’s generative model (or internal model) of the world. For any living being, this process is essential for survival: too significant a discrepancy between the internal model and sensory input could pose life-threatening risks. Consider, for instance, a hypothetical case in which, although a living organism’s internal model prescribes refraining from harmful stimuli, such as fire, the organism fails to respond accordingly, ignoring its generative model by continually exposing itself to the sensation of burning. This results in sensory input that not only contradicts the internal model but also jeopardizes the organism’s very existence. The more efficiently the selected sensory input serves as evidence for the internal model, the greater the organism’s chances of sustaining life. Due to the life-preserving function of this “evidence” selection mechanism, it is aptly termed self-evidencing.

Although the term was coined in the context of PP for the first time by Jakob Hohwy (Hohwy 2016), it was in Karl Friston’s formulation of his free energy principle that a sense of co-dependence or fusion between the “self”, or the agent, and its generative model was first introduced:

“…[M]inimizing surprise is the same as maximizing the sensory evidence for an agent’s existence, if we regard the agent as a model of its world”.

(Friston 2010, p. 128)

The idea of “the agent as a model of its world” is crucial as, according to Friston, and as explained above, it is the very life of the agent or any self-organizing, self-regulating (biological) system that depends on the reliability of its predictions, hence, on the surprise minimization mechanism. Friston associates surprise in his theory with the more technical concepts of free energy and entropy. A “fish out of water”, for example, according to this framework, is in a high state of surprise—in Friston’s own words, “both emotionally and mathematically” (ibid., p. 127). He writes further in this regard:

“Note that both surprise and entropy depend on the agent: what is surprising for one agent (for example, being out of water) may not be surprising for another. Biological agents must therefore minimize the long-term average of surprise to ensure that their sensory entropy remains low”.

(ibid.)

Hohwy elaborated upon this idea within the predictive processing framework, focusing on the brain as the most complex self-organizing system (Hohwy 2016). He further distinguished between conscious and non-conscious self-evidencing, attributing to the former specific properties commonly associated with human consciousness, such as first-person perspective, metacognition, introspection, a sense of self, and an awareness of being (Hohwy 2022). As the foundational process underlying these properties, Hohwy proposed the concept of counterfactual active inference, which he defines as active inference in the face of counterfactual scenarios, each of which requires the self-evidencing system to enact a particular policy for its action: “Now active inference becomes a race between possible policies, or an inference to the policy that best augurs uncertainty reduction” (Hohwy 2022, p. 817).

Nevertheless, as previously noted, if predictive processing claims to be a universal theory of brain function, it should address in its account of self-evidencing, when it comes to the human species, the decisive role of communication and the uniquely human modes of interaction with a world that naturally includes other human beings. As Michael Tomasello explains in The Cultural Origins of Human Cognition, the evolutionary development of uniquely human cognitive skills—enabling the formation of cultures and institutions—occurred within a relatively short time span (approximately six million years), too short, from an evolutionary point of view, to be fully accounted for by the “normal processes of biological evolution involving genetic variation and natural selection” (Tomasello 1999, p. 2). The rapid evolution of human cognition, Tomasello argues, could be explained by the particular human ability of cultural and social adaptation through learning. Central to this is the human being’s unique capability to perceive other human beings as intentional agents like the self, from whom she/he would willingly learn. He later developed his thesis by linking it to the concept of shared intentionality, which he defines as follows:

“…any behavioral phenomena that are both intentional and irreducibly social, in the sense that the agent of the intentions and actions is the plural subject ‘we’”.

(Tomasello 2008, p. 72)

According to Tomasello, thus, the cognitive evolution of the human species owes its particularity to the human capacity to share psychological states with one another, enabling collaborative interactions among them as well as second-personal thinking, that is, a form of thinking that takes the perspective of the other into consideration (Tomasello and Carpenter 2007; Tomasello 2014, p. 68ff). Similarly, Vojtěch Kolman sees in Wittgenstein’s philosophy the notion of a “transcendental we” as the condition for the possibility of any language game and, subsequently, any (human) form of life (Kolman 2019, p. 234).

What makes Tomasello’s thesis an interesting case to be brought into discussion with the predictive processing paradigm is the inferential aspect of the process of understanding others and incorporating the information communicated by them into our own “intentional states”. He writes in this respect:

“[H]uman cooperative communication involves a set of special intentional and inferential processes… Human communicators conceptualize situations and entities via external communicative vehicles for other persons; these other persons then attempt to determine why the communicator thinks that these situations and entities will be relevant for them. This dialogic process involves not only skills and motivations for shared intentionality but also a number of complex and recursive inferences about others’ intentions toward my intentional states”.

(Tomasello 2014, p. 3)

The question now is whether human engagement in dialogue, as described above, can be conceptualized as a form of active inference—one that generates unique modes of development in human generative models due to the complex cognitive strategies required for minimizing surprise when interacting with other human beings, who are obviously less predictable than natural phenomena. I do not see any reason that predictive processing should resist this suggestion (see Andy Clark’s remarks on the significance of sociocultural structures and practices for the particularity of the case of the human mind in the PP paradigm in Clark 2016, pp. 275–94).

But what does this mean for human self-evidencing? Should not maximizing evidence for the agent’s existence include seeking dialogue mainly with those whose generative models closely align with the agent’s own—or, in Wittgenstein’s terms, engaging with those whose world image or form of life corresponds to that of the agent? An overview of human history that testifies to sharp tribal divisions, distinct cultural patterns, religious borders, and nationalist aspirations, all in all, speaks for an affirmative response to this question. But if considered through the lens of predictive processing, this is only half of the story! The concept of maximizing evidence for one’s existence encompasses seeking such evidence beyond familiar zones, often driving a tendency to explore the unknown. Clark writes in this regard:

“But we humans are also expert at deliberately manipulating our physical and social worlds so that they provide new and ever-more-challenging patterns that will drive new learning”.

(Clark 2016, p. 277)

If this description, rooted in predictive processing theory, holds true, it suggests that human beings naturally tend to engage in communication also with those whose generative models differ to varying degrees from their own. Engaging in dialogue with such partners, as long as the agent does not perceive it as life-threatening, can refine and reform the human agent’s generative model in unique ways.5 As previously noted, the dual nature of human behavior—predictable in some aspects yet unpredictable in others—combined with the familiarity of one’s form of life contrasted against the unfamiliarity of those of others, renders prediction-error minimization in interpersonal relations more complex than in instances where cognition involves non-human entities.

In fact, as Wittgenstein’s various examples and imagined scenarios in Philosophical Investigations suggest, the predictability of certain aspects of other human beings’ actions is crucial for the possibility of sharing language games with them. It is a common understanding of the rules of action and trust in the intersubjective realm of information exchange based on shared rules of grammar that make human communication and participation in language games possible. Wittgenstein speaks of this common understanding in terms of an agreement in the form of life (PI sct. 241). But Wittgenstein’s philosophy also relies on an element of unpredictability. The gradual alteration of language games, as stated in the above-cited remark from On Certainty (OC sct. 63, also see, PI sct. 23), and the dynamism within the world image described by Wittgenstein with his riverbed/flux metaphor (OC sct. 96–97) already show that there is an element of contingency in the rules of grammar required for practice in accord with a form of life (see, Von Stosch 2019).

One should bear in mind, however, that for Wittgenstein, forms of life and rules of grammar are not mere conventions consciously agreed upon at a collective level. Instead, he associates human forms of life with what he calls the “natural history” of humankind (PI sct. 25). In her explanation of this remark, Toril Moi, drawing on Stanley Cavell’s reading of Wittgenstein, has referred to the broad sense of the term “natural” in Wittgenstein’s philosophy. According to Moi, the notion of “form of life” in Wittgenstein encompasses not only biological facts—with the human body as the fundamental condition for the possibility of acting and behaving in the world—but also cultural practices:

“Cutting across the nature/culture divide, the concept of “forms of life” is elastic enough to range from the purely biological to the completely cultural, to encompass the human body as well as the finest distinctions of the practices in a specific culture”.

(Moi 2017, p. 56)

Mikel Burley also, underlining the lack of a sharp division between the natural and the religious in Wittgenstein’s philosophy, rejects attributing to Wittgenstein what he calls “a full-bloodedly reductive form of naturalism” (Burley 2018, p. 58)—one that reduces the natural to its scientific sense, prioritizing scientific discourse over all other discourses. According to Burley, Wittgenstein’s philosophy neither adopts a purely scientific view of nature nor reduces religious belief to an over-imposition of religious discourse to an otherwise scientifically understood realm of nature. He shows that to read Wittgenstein in this way would be exactly in direct opposition to Wittgenstein’s own critique of James George Frazer:

“What Wittgenstein chiefly objects to in Frazer’s treatment of ritual behavior is the very idea that such behavior could be explained by reference to some belief or theory that the participants supposedly hold… Instead of assuming that rites must be explicable in terms of beliefs or opinions, Wittgenstein urges us to look for analogous forms of behavior in our own lives and to thereby come to see resemblances” (ibid., p. 60).

As an example of such forms of behavior, Burley refers to Wittgenstein’s own reference to an ordinary case where a person kisses the picture of a loved one (ibid.). It is clear to everyone that such an act does not arise from the belief that the kissing of the photo has an effect on the loved person. This act is simply what people usually do to attain satisfaction, “[o]r rather, it does not aim at anything; we act in this way and then feel satisfied” (Wittgenstein 2018, p. 36).

By putting together Wittgenstein’s concept of form of life, which encompasses both biological and cultural dimensions, and Clark’s account of predictive processing sensitive to the sociocultural aspects of human life, one can establish a bridge between the two paradigms through their almost aligned perspectives on human life. This opens the possibility of discussing self-evidencing within Wittgenstein’s philosophical framework.

In the case of the example of kissing the picture of a loved one, for instance, viewed through the lens of predictive processing, one could understand the same act as a concrete case of self-evidencing: The loving agent, despite knowing the fact that kissing the photo will not have the slightest effect on the person in the photo, predicts certain effects on oneself and, perhaps, also on the way that the agent perceives one’s relationship to the person in the photo. In performing that act, human agents simply verify their prediction of such an effect without feeling the need to correct or change anything about their generative model in this particular case. This view can be extended to a whole range of human cultural practices that do not necessarily involve theoretical reflection but, like self-evidencing, integrate both cognitive and non-cognitive aspects of human orientation in the world.

4. To Believe, or Not to Believe, Is That the Question?

This article has attempted thus far to show that Wittgenstein’s philosophy, in general, could be reformulated using concepts from the predictive processing paradigm, a paradigm that is considered to explain a wide range of tasks ascribed to brain function. In this section, I would like to narrow down the analysis to religious discourse and to show that anchoring Wittgenstein’s ideas about religion in the context of this paradigm could lead to a clearer understanding of these ideas, making them probably more persuasive.

As already mentioned in the first section of this article, Michael Scott attributes to Wittgenstein an understanding of religious beliefs that held these as evidence-resistant/repellent. He moves on by referring to the communities of believers in end-of-the-world prophecies (such as UFO cults), whose beliefs were as much evidence-repellent as Scott took religious beliefs to be, to conclude that Wittgenstein’s immunization of religious propositions from face-value evaluation based on their inherent evidence resistance/repellence was an untenable position leading to a weakness in his philosophy of religion. Let us review this criticism by taking into consideration the above analysis.

Speaking of world image as an inherited background against which one distinguishes between true and false (OC sct. 94) and of agreement in form of life (rather than in opinions) as decisive for what is true and what is false, Wittgenstein already makes a general claim about human orientation in the world regardless of the nature of this orientation—whether it be religious or not. “Evidence resistance” thus is more or less a common feature of all human convictions. As already discussed, within the predictive processing paradigm the generative model also functions as the background against which evidence for the existence of oneself is selected (bottom-up selection of data that fits the generative model). Nevertheless, it is also the nature of the self-evidencing agent to engage actively in the world and to explore the realms that would offer more such evidence. An inevitable aspect of this endeavor, in the case of human beings, is to engage in dialogue with those whose generative model differs from ours. Returning to the context of religious discourse, this dynamic may help explain why apologetic encounters with “the others”—particularly in Islam and Christianity—played a foundational role in the development of theology in history. As the history of theology demonstrates, such encounters did not merely lead to the reaffirmation of the believers’ religious convictions but also contributed to the vitality and enrichment of religious discourse. To borrow Wittgenstein’s terminology, they gradually gave rise to specific language games within a shared (dialogical) stage throughout our “natural” history. From this perspective, theology is no less an intellectual discipline than science or history, provided one acknowledges the decisive role of religious convictions in shaping, updating, and transforming the generative model of a self-evidencing agent. I would argue, in line with this reasoning, that the appeal of Wittgenstein’s philosophy of religion, thus, does not lie in its emphasis on the distinctiveness of religious language in contrast to scientific and historical discourses but (partly) in the subtle idea that there is a common feature to all human convictions from which no specific discourse is exempt.6 This is why the prioritization of a specific discourse in reading his philosophy remains a contradictory endeavor. It is for the same reason that one should suspend judgment concerning the meaning of the term “natural” in Wittgenstein’s conception when he speaks of the “natural history” of humankind.

But what about evidence repellence? What does it mean when Wittgenstein says that, “if there were evidence [for religious belief], this would in fact destroy the whole business”? Are religious beliefs, as suggested by Scott, necessarily evidence-repellent in Wittgenstein’s philosophy?

To unpack this, we can turn to Wittgenstein’s On Certainty and examine passages that discuss the nature of belief in contrast to knowledge. These insights provide a framework for understanding his claim, revealing the relationship between belief—in general—and evidentiary justification in his thought. As for the relevant passages, to start with one could refer to the following:

“… And ‘I knew he was in the room, but he wasn’t in the room’ is like “I saw him in the room, but he wasn’t there”. ’I know’ is supposed to express a relation, not between me and the sense of a proposition [Satzsinn] (like “I believe”) but between me and a fact. So that the fact is taken into my consciousness…”.

(OC sct. 90)

“’I know it’ I say to someone else; and here there is a justification. But there is none for my belief”.

(OC sct. 175)

“What I know, I believe”.

(OC sct. 177)

“It would be correct to say: ‘I believe…’ has subjective truth; but “I know…” not”.

(OC sct. 179)

From the above quotations, it becomes clear that Wittgenstein, referring to our common sense and use of the verbs “to know” and “to believe”, holds a general distinction between knowing something and believing something. The distinction shows itself, however, only in our application of such terms and not within a broader theoretical context. Wittgenstein’s point is simply that the functions of these two verbs are different. In one case, namely, in the case where I say, “I know”, I state something about my relation to a fact, which, if you would also follow the same steps of justification, would also become a fact for you. In such a case, using the terms “I know” or “I believe” would not make a huge difference as sct. 177 suggests. Moreover, as Wittgenstein himself suggests in other passages of the same work, a huge part of what we learn in life relies on us believing what others—parents, teachers, scientists, prophets (?), etc.—have taught us (see, for example, OC sct. 144, 170). Moreover, without initial belief—and this is usually a matter of (practical) decision—not just learning but also acting in accord with what we have learned in an intersubjective realm of knowledge and action becomes impossible.

“Sure evidence is what we accept as sure, it is evidence that we go by in acting surely, acting without any doubt. What we call “a mistake” plays a quite special part in our language games, and so too does what we regard as certain evidence”.

(OC sct. 196)

“Giving grounds, however, justifying the evidence, comes to an end;—but the end is not certain propositions’ striking us immediately as true, i.e., it is not a kind of seeing on our part; it is our acting, which lies at the bottom of the language-game”.

(OC sct. 204)

One could interpret the above lines as stating that belief is thus an assertion based on our “decision” to play as an active agent in a shared context. When Wittgenstein then regards lack of “evidence” as defining for religious beliefs, he means that a decision to act—or not to act—in accord with those beliefs is prior to any justification based on what we “normally call evidence”. This decision, however, as also suggested by the fact that the concept of self-evidencing in predictive processing is not always a reflected decision. It is rather best described as a practical one based on what form of agency, what form of life, one has grown into, and what one takes to correspond to the “self” in relation to the world and to other human beings.

Both Wittgenstein and predictive processing, in their talk of belief, are less concerned about the objective truth outside the realm of human consciousness than the subjective aspect of belief and the kind of agency one develops through acting according to those beliefs.7 Finally, one could say that in both paradigms it is as much about being as it is about believing.

5. Conclusions

This article has aimed to demonstrate how predictive processing theory could support and, in certain respects, shed light on Wittgenstein’s ideas about religion. Using Michael Scott’s critique as representative of the common criticisms that analytic philosophers of religion often direct toward Wittgenstein’s views on religious belief, I have explored how the conceptual framework of predictive processing might enable a more charitable reading of Wittgenstein for such readers. It has at the same time been suggested that certain insights from Wittgenstein’s philosophy, that is, those associated with the significance he gives to the intersubjective realm of efficacy and of exchange between human subjects, could help expand the scope of the areas that predictive processing, I believe, still has the potential to explore.

One central point of the present analysis has been to challenge the claim that Wittgenstein’s philosophy of religion isolates religious discourse as uniquely evidence-resistant or evidence-repellent. By reading Wittgenstein’s ideas on world image and form of life through the lens of predictive processing—with a special focus on the notion of self-evidencing—it has been shown that evidence resistance is not an exclusive feature of religious discourse but a characteristic of any deep human conviction, grounded in the world image or, to speak with the language of PP, in the generative model.

As far as evidence repellence is concerned, I have argued with Wittgenstein’s specific approach to evidence—as what we accept to be evidence—that his philosophy is less concerned about the objective truth than it is a subjective, practical decision concerning which language games or which intersubjective realm of exchange one is to participate in. Following this line of thought, it becomes evident why belief in UFO cults, flat Earth theories, and, to use Scott’s terminology, any other evidence-repellent phenomena often arises within the context of conspiracy theories, as such belief is usually based on a profound distrust in the shared, intersubjective realm of knowledge exchange. This distrust is rooted in the belief that this shared realm is somehow externally controlled, with individuals merely assuming that they are active agents. Once trust in this shared space is lost, the processes of learning and acting within that space cease. The self becomes trapped in a circular process of self-evidencing, confining its exploration to a space shared only with those who hold the same beliefs. It is in such cases that a belief system becomes evidence-repellent.

Religious beliefs, like all other forms of belief, are not immune to this risk, just as they are not, as claimed by some Wittgensteinians, immune to ordinary evidence. Yet, the more one broadens the scope of self-evidencing to encompass diverse discourses and contexts—albeit with various degrees of commitment to each—the more dynamic and vibrant religious language and religious thought become.