An Existential Reconstruction of the Self Based on the Free Energy Principle

This treatise forms the intellectual basis of the Artist Statement for the paintings.

Introduction: The Problem and the Structure of This Treatise

Section 1. Setting the Problem

The various systems of contemporary society possess an unprecedented complexity and autonomy. Diverse domains such as politics, economics, the military and communications are connected to one another under the demand for rationalization, and have reached a stage of overall self-organization that exceeds partial control. There, human actions and emotions are often treated as disturbances.

Meanwhile, the development of neuroscience and information theory has pushed human self-consciousness to the point where it can be described as a physical process of information processing. The naturalization of self-consciousness — the position of “understanding the human as part of natural phenomena” — means, philosophically, the end of anthropocentrism (Dennett, 1991). Yet the further that naturalization proceeds, the more we confront the fundamental problem: “What is the self, and how can it be redefined?”

At the same time, this problem does not remain a purely theoretical question. Within a social structure built on particular neurological traits as the norm, individuals who deviate from that norm are forced to align their own predictive structures with social norms. Within the framework of prediction-error minimization, forced alignment and voluntary alignment are described equivalently, as reductions of error. Yet for the person concerned, “I aligned myself” and “I was forced to align” are fundamentally different experiences. This gap between theoretical equivalence and existential asymmetry is the core of the question raised by the process of forced alignment. In other words, if the naturalization of the self is a theoretical problem, forced adaptation to social structure is an existential one. This treatise handles these two questions within a single theoretical framework.

Here it is necessary to state at the outset that the word “naturalization” carries two different problematics. The first meaning is naturalization in the sense of cognitive science and neuroscience: the methodological position of describing self-consciousness and subjective experience as physical and information-theoretic processes. The second meaning is naturalization in the sense of social systems theory: the process by which artificially designed institutions and norms leave explicit external control and are transformed into autonomous, self-maintaining systems.

On the surface these two naturalizations appear to deal with problems belonging to different domains. This treatise, however, takes the position that the two form one continuous problem field through a common optimization principle: “prediction-error minimization.” And its fundamental question is: under conditions in which this double naturalization proceeds simultaneously, how can the “self” be reconstructed?

The central thesis running through this treatise can be summarized as follows.

Self, society and value can be understood continuously as the self-organization of an informational order constrained by prediction-error minimization, and within that continuity the subjective experience of “existential self-renewal” arises.

The aim of this treatise is to examine this central thesis under the contemporary conditions of the autonomization of social systems and the naturalization of self-consciousness, while connecting it to the existential problem of forced alignment.

Section 2. The Structure of This Treatise and the Flow of the Argument

This treatise argues for the central thesis in two stages.

Chapter 1, based on predictive processing theory and the free energy principle, clarifies how self, consciousness and spacetime are constituted. The aim here is not mere conceptual organization but establishing the theoretical foundation of the central thesis. In particular, it shows that the self functions not as a closed entity but as an open system dynamically coupled with its environment through prediction-error minimization. This understanding is an indispensable premise for the social and existential discussions that follow.

Chapter 2, presupposing the theory of Chapter 1, discusses the three-layered structure of the human (Section 1) and examines the tension of the asymmetry between human and machine (Section 2). It then proceeds to the mechanisms by which social order is formed, becomes autonomous and is naturalized (Sections 3 and 4), and, by way of the problem of forced alignment in a naturalized society (Section 5), responds to the existential question of how the self can be reconstructed (Sections 6 and 7).

Section 3. Definitions of Key Concepts

This section defines the key concepts used throughout the treatise.

Information: not a mere quantity of data, but ordered difference that makes prediction and action possible. It is used in the following three layers. (1) Physical/statistical information: information flow captured indirectly as communication volume or rates of change. (2) Semantic/symbolic information: information observed as the stability of socially shared codes such as language, institutions and norms. (3) Information as prediction error: information experienced as the gap between expectation and outcome that arises when a subject predicts the world.

Order: a dynamically stable state in which such information is arranged in a form that persists over time and has predictability. It is understood not as a fixed structure but as a process maintained and updated through interaction with the environment.

Naturalization (two senses): (1) Cognitive-scientific naturalization: the position that describes self-consciousness as a neuroscientific and information-theoretic process. (2) Social-systemic naturalization: the transition to a stage in which institutions and norms leave external control and begin to behave as autonomous, self-maintaining orders. This treatise claims that these two senses can be understood continuously under the principle of prediction-error minimization.

Structural correspondence: the correspondence in which neural processes within an individual and social institutions behave as systems at different levels bound by the same optimization principle (prediction-error minimization). It does not mean isomorphism at the level of elements; it is set up as a theoretical working hypothesis. This correspondence is constructed as a relationship that can be compared indirectly through observable social-scientific indicators such as the frequency of institutional renewal, the distribution of behavioral deviation and the cycle of normative redefinition. It is not, however, a direct equation.

Prediction error: a quantity of discrepancy experienced within the individual as cognitive dissonance, and observed indirectly at the social level as deviations of action or institutional instability.

Forced alignment: a state in which a subject aligns not as a voluntary choice but because the constraints of social structure leave no alternative. It is asymmetrical with voluntary alignment, and this treatise treats this asymmetry as a distinct problem that can be distinguished from a homogeneous description in terms of “prediction-error minimization.”

This treatise partly overlaps with Luhmann’s social systems theory (Luhmann, 1984) and with the collective active inference model of Friston and colleagues (Friston et al., 2016), but differs in position on the following points. Luhmann describes society as a self-referentially closed system with communication as its basic unit, whereas this treatise places the neuroscientific principle of individual prediction-error minimization at the base and understands social order constructively as the aggregate interaction of individuals. The collective active inference of Friston and colleagues provides a framework in which multiple agents coordinate their behavior through shared generative models; this treatise draws on that framework, but its main purpose is to examine the existential implications specific to the “naturalization” of social systems and the phenomenon of forced alignment.

Chapter 1: The Constitutive Principles of the Self and Spacetime

Section 1. The Free Energy Principle and Predictive Processing

The purpose of this chapter is to clarify, as the theoretical foundation of the central thesis, how self, consciousness and spacetime are constituted.

The free energy principle, proposed by Friston and colleagues, is a framework that explains the perception, action and learning of biological systems in a unified way (Friston, 2010). Its core is the claim that organisms are organized so as to minimize “variational free energy” — a quantity formulated as an upper bound on the discrepancy between sensory input and an internal model.

Under this principle, the brain does not perceive the outside world directly; it builds a probabilistic predictive model (a generative model) of the “causes” of sensory input, and is constantly updated so as to minimize the error between its predictions and the actual input (Clark, 2016; Hohwy, 2013). This process is realized through two routes. The first is “perceptual inference,” which brings predictions closer to sensory input by updating the model itself. The second is “active inference,” which brings sensory input closer to predictions through one’s own actions.

Perception and action play symmetrical roles in this framework, and both are derived from a single objective function: the minimization of free energy. This is the substantive content of predictive processing theory (Rao & Ballard, 1999), and the theoretical ground on which this treatise uses “prediction-error minimization” as its basic principle.

It should be noted, however, that the free energy principle remains a theoretically and empirically contested framework. Critics point out that its excessive generalizability undermines the theory’s falsifiability (Klein, 2018). While recognizing this criticism, this treatise adopts predictive processing theory as a working framework for describing individual cognition and extends it to society and existence.

Section 2. Self and Consciousness

Under this framework, the “self” is defined as a collection of hierarchical generative models built by the organism as a whole, with the brain at its core (a system dynamically coupled with its environment), functioning as a free-energy-minimizing system that seeks to maintain homeostasis (Seth, 2021).

These generative models, while internalizing boundary conditions that distinguish self from non-self, try to minimize prediction error recursively and actively by predicting state transitions in the outside world and inside the body. The important point is that the self exists not as a complete entity but as a continuous process of minimizing prediction error — that is, as a verb-like process.

Consciousness is regarded as a phenomenon in which higher-order, integrated predictive representations within the hierarchical generative model become manifest while maintaining a certain stability (Hobson & Friston, 2014). Of the errors arising between bottom-up sensory input and top-down prediction, the brain must decide which errors to adopt for updating the model. This is called precision weighting: a control process that increases the gain of neural activity for particular error signals.

Consciousness is this very control process of precision weighting; by weighting the reliability of signals, it allows the self-system to selectively carry out adaptive learning within a flood of information. For example, the process of understanding an unexpected event and feeling that it “makes sense” is understood as the continuous result of this weighting and of model updating.

If the self is defined as a “process of prediction-error minimization,” then the identity of the self is described not as belonging to a fixed entity but as the continued stability of a particular predictive structure. This means that the self is, in principle, reconstructable. For as long as the generative model is updated, the self is transformed as well.

Section 3. Spacetime

In the framework of the free energy principle, time is an action-oriented predictive structure by which the self-model sequences events in the context of action and prediction. Space is understood as a representation of the arrangement of objects based on the possibilities of one’s own body (Friston et al., 2017). “Spacetime” is a conceptual model specific to the cognizing subject that integrates these two, a framework that functionally maps physical causal relations under prediction-error minimization.

This spacetime model represents what state transitions the self’s actions may bring about in the future, by organizing the contingencies between sensory input and motor output temporally and spatially. Its stability is guaranteed as long as consistency is maintained between the generative model internalized by the self-model and the statistical regularities of the environment.

This subjective, functional spacetime, however, is not entirely arbitrary in its correspondence with physical spacetime; it holds insofar as it indirectly reflects the causal structure of physical spacetime under the constraint of prediction-error minimization. Subjective spacetime may not coincide with physical absolute order, but the causal relations the subject needs in order to predict and control action are maintained, and it functions as an approximate map of physical spacetime.

Section 4. Summary

This chapter set out the theoretical substance of the free energy principle and predictive processing theory, and then described how self, consciousness and spacetime are constituted. The three points established here form the foundation of the next chapter. First, the self is described as a process of prediction-error minimization and is an essentially dynamic, updatable structure. Second, consciousness, as the control process of precision weighting, has the function of adjusting the self’s choices of learning. Third, spacetime too is a product of subjective generative models, understood as a functional mapping under the constraint of prediction-error minimization. On the basis of this understanding, the next chapter extends the same logic to the dimensions of society, value and existence.

Chapter 2: The Naturalization of Systems and the Reconstruction of the Self

This chapter, presupposing the framework of predictive processing and free energy established in Chapter 1, discusses the three-layered structure of the human (Section 1) and examines the tension of the asymmetry between human and machine (Section 2). It then proceeds to the mechanisms by which social order is formed, becomes autonomous and is naturalized (Sections 3 and 4), and, by way of the problem of forced alignment in a naturalized society (Section 5), responds to the existential question of how the self can be reconstructed (Sections 6 and 7).

Section 1. The Three-Layered Structure of the Human

As a premise for the social and existential theory this treatise develops, it is first necessary to clarify the structure of the human as an individual.

This treatise understands the “human” not as a being defined by a single essence but as a three-layered generative structure operating under different time scales and constraints. This three-layer model overlaps with Ricoeur’s theory of narrative identity (Ricoeur, 1990) and Damasio’s theory of the bodily self (Damasio, 1994), but here they are redescribed in a unified way within the framework of the free energy principle. The three layers are not mutually independent entities; they represent different functional levels, hierarchically coupled under the common principle of prediction-error minimization.

The first layer: the human as a biological being. What plays the central role in this layer is the regulatory process that maintains the body’s homeostasis and keeps it within a viable state space. Phenomena such as pleasure and displeasure, fear and desire function as indicators for maintaining the order of internal states within interaction with the environment. Human activity in this layer is described not as purpose-rational but as a constraint-driven process of adaptation for securing viability.

The second layer: the human as a social being. For multiple subjects to coordinate their actions with one another, their behavior must be predictable to a certain degree. Shared frameworks such as language, institutions, norms and roles are formed in response to this demand. The desire for recognition and sensitivity to evaluation can be understood as regulatory mechanisms for maintaining this mutual predictability. In this layer the human acts not as a pure organism but as a member participating in shared generative models (Friston et al., 2016).

The third layer: the human as a narrative being. In this layer the individual links past experience, present choices and future possibilities, forming a long-term generative model in which it understands itself as a coherent being. Phenomena such as value, ethics, life goals and identity appear as functions that maintain the coherence of this long-term self-model. “Truth” here appears not as an external absolute standard but as the stability of a predictive structure that can integrate one’s experience without breaking down.

These three layers are not separated in a relation of above and below; they operate in layers, such that on top of the range of behavior made possible by the biological layer, the social layer forms frameworks of mutual coordination, and the narrative layer integrates them over time. Conflict and wavering of meaning can be understood as phenomena that arise when the predictive structures demanded across these layers do not agree.

Section 2. Contingency and Planning

So far we have described the human as a prediction-error-minimizing system. But here we must face one tension squarely. This description seems to treat the human as a kind of optimizing machine. When we contrast humans as “contingent beings” with machines as “planned systems,” is this description not contradictory?

This question contains an essential tension that should be examined head-on. As the pressure toward rationalization, regulation and calculability of social structure advances, humans accumulate survival-strategic learning as adaptation to their environment, and their patterns of behavior gradually take on a mechanical planning. Meanwhile, the progress of science and technology gives machines probabilistic processing, context-dependent response and self-correcting capacities, and machines try to approach human characteristics. This mutual approach of “mechanizing humans” and “humanizing machines” calls into question the boundary between the two.

From the standpoint of this treatise, this tension can be organized as follows. The principle of prediction-error minimization is a descriptive framework applicable to both humans and machines. But applying this common principle does not mean identifying the two. The human generative model is formed under the constraints of three layers — biological, social and narrative — at evolutionary, developmental and personal time scales. Its process essentially contains contingency that no designer could specify in advance. Human “contingency” is understood as unpredictable emergent behavior arising from the nonlinear interaction of these multilayered constraints.

The “planning” of machines, by contrast, is understood as optimization within the range of a design specification. Even if a machine learning system shows probabilistic behavior, the design space of its probability distributions is itself defined by humans. The “contingency” a machine can acquire is contingency within a designed probabilistic structure, not emergence from outside that structure.

This distinction, however, does not draw a stable boundary. As the rationalization and naturalization of social systems advance, the space of human choices of action is also structurally limited by social generative models. That is, the room in which humans can exercise emergence as “contingent beings” shrinks as the naturalization of social systems proceeds. This process of shrinking is the phenomenal substance of “mechanizing humans,” and the ground on which the task of “reconstructing the self” in this treatise arises.

The contrast between contingency and planning is therefore understood not as a fixed essential distinction but as a dynamic asymmetry that varies depending on the state of social systems. The reconstruction of the self is positioned as a practice of trying to preserve room for human emergence under conditions in which this asymmetry is being compressed.

This section discusses how social order self-organizes when several of the individuals defined so far come together.

To cope with the uncertainty that is the other, each individual relies on common protocols (law, money, language) to predict one another’s behavior. This demand for “mutual predictability” is the driving force of the formation of social order. The aggregate of each individual’s behavior in trying to minimize prediction error precipitates a shared predictive structure outside the individual. This is a genetic explanation of institutions and norms.

These protocols are at first merely tools of individuals, but as the system grows larger, they become fixed as external constraints for maintaining connections between individuals. Individual agents oriented toward prediction-error minimization can no longer choose anything but conformity to this fixed external structure (the social system) as the optimal solution, and as a result the social system forms an autonomous feedback loop that incorporates its members as resources for its own maintenance.

In this way, social order is formed as the aggregate consequence of individuals’ prediction-error-minimizing behavior, but once established, social order itself comes to function as a higher-level context that directs individuals’ behavior. With this bidirectional feedback, this treatise regards the relation between individual and society not as an irreversible break but as a hierarchical continuity.

This correspondence should be understood not as an equation but as a structural analogy. Prediction error within the individual and social instability are set up as a relationship that can be compared indirectly through social-scientific indicators such as the frequency of institutional renewal, the distribution of behavioral deviation and the cycle of normative redefinition, but directly identifying the two is not the claim of this treatise.

When the self-organization of social order described in the previous section reaches a certain level of maturity, it moves into the stage this treatise calls “the naturalization of social systems (the second sense).”

The naturalization of social systems is the phase in which institutions go beyond being means of realizing human purposes, and the institutions themselves begin to behave as orders that give priority to their own survival. Laws and institutions are initially introduced as tools that assist individuals’ actions, but as their operation advances they become fixed as preconditions for the choice of action. Credit scores, algorithmic performance evaluation, automated screening systems and the like are likewise introduced at first as tools to assist human judgment, but eventually the systems themselves begin to have their own optimization metrics and start to function as autonomous environmental constraints that exclude human intervention. Individual subjects act on the rational choice of conforming to these standards, and as a result the institutions stabilize in a self-reproducing way beyond human purposes.

In this process, technology functions as a medium that accelerates the naturalization of social systems. The streamlining of social administration through artificial intelligence and automation technologies can also be seen as a form in which, beyond the initial intention of achieving human purposes, the system itself evolves with self-maintenance as its purpose.

Here the two naturalizations distinguished in the Introduction are connected. Cognitive-scientific naturalization (the first sense), by describing the individual-level self as a process of prediction-error minimization, brings about the “transparency” of the self. Social-systemic naturalization (the second sense), by fixing social order as an external constraint, structurally limits the individual’s space of choice. When this double naturalization proceeds simultaneously, the question “What is the self, and how should one live?” emerges not as a mere philosophical exchange but as an existential task: taking on one’s own position within the information environment.

Section 5. Forced Alignment

The naturalization of social systems discussed so far raises one concrete problem: under what conditions are individuals placed who deviate from the “standard neurological traits” that a naturalized social order presupposes?

Social generative models internalize, as a “predictable standard,” patterns of behavior, language and emotional expression built on the neurological traits of the majority. Individuals with neurological traits that deviate from this standard — the neurological minority — experience a constant discrepancy with the predictive structure of the social system. This discrepancy differs in nature from prediction error in the ordinary sense. Ordinary prediction error is reduced by the individual updating its generative model. But the discrepancy arising from neurodiversity cannot be fundamentally resolved by the individual updating its generative model. This is because the root cause of the discrepancy lies not in an “error” of the individual’s generative model, but in the fact that the social generative model is designed with particular neurological traits as the standard.

Within this asymmetry, many in the neurological minority try to reduce social prediction error by concealing their own traits and imitating the behavior, language and emotional expression of the majority. This mimetic behavior (masking) is a rational adaptive strategy that an individual, as a free-energy-minimizing system, can take within the environmental constraints given to it.

This adaptive strategy, however, comes at a high cost. Masking is the act of continuously maintaining an artificial gap between the individual’s actual predictive structure and the predictive structure presented to society. This continuous internal gap is experienced as a chronic increase in free energy. That is, masking has an asymmetrical structure of cost: while it reduces social prediction error externally, it constantly increases prediction error inside the individual.

What this treatise emphasizes here concerns the attribution of the cause of this problem. A position that locates the cause of “maladjustment” in the individual forced to mask makes the bias of the social generative model invisible, and thereby naturalizes and fixes that bias. In the framework of this treatise, a social generative model designed with particular neurological traits as the standard is not a product of universal optimization but a product of historically contingent design, established by privileging the predictive structure of the majority.

Forced alignment within a naturalized social system therefore raises a distinct problem that cannot be absorbed into the homogeneous description of “prediction-error minimization.” It is a political question: whose predictive structure is taken as the standard on which the social generative model is built? The task of reconstructing the self cannot be separated from this question. For to accept adaptation to a naturalized social system as a rational choice means also taking on the bias of its social generative model.

Section 6. The Reconstruction of the Self

On the basis of the considerations so far, this section discusses how the “self” can be reconstructed within a naturalized social system.

As confirmed in Chapter 1, the self is defined not as a static entity but as a continuous process of prediction-error minimization. This implies that the self is, in principle, reconstructable. However, reconstructability as a theoretical possibility must be distinguished from self-reconstruction as an existential task. The latter is the question of how to update one’s own generative model within the concrete constraints of a naturalized social system.

The “reconstruction of the self” in this treatise is not the negation of existing contexts but the act of rearranging them within a different order. “Rearranging” here does not mean erasing the various contexts that have constituted the self — family relationships, professional roles, moral beliefs, images of the future and so on — but recombining the hierarchical relations that determine to which level of order each is subordinate.

For example, when a subject who had placed particular social roles or evaluations at the core of their identity repositions them beneath a higher-order value system or exploratory goals, the individual roles remain while the predictive structure of the self as a whole moves to a different point of stability. Such hierarchical reorganization is what this treatise calls the “reconstruction of the self.” This shares its concern with the attempts of Heidegger’s “call to authenticity” (Heidegger, 1927) and Sartre’s “fundamental choice” (Sartre, 1943) to describe the renewal of meaning by the subject’s will. This treatise transplants that description into the framework of predictive processing.

Existential self-reconstruction, in other words, is understood as a process in which the subject actively shifts the minimum point of free energy by consciously recombining the hierarchical structure of its own generative model.

This process of rearrangement temporarily destabilizes the reliability of the existing generative model. The important behavior the self engages in at this point is “novelty-seeking behavior” arising from curiosity (Schmidhuber, 2010). This is a form of active inference that actively accepts a temporary increase in prediction error in order to prevent the fixation of the self-system and to explore a more optimized higher-order order.

What matters here is that this increase in prediction error is not passive confusion imposed from outside but uncertainty that the subject itself selectively takes on. Curiosity is an attitude of deliberately lowering the reliability of the existing generative model for a time, even in situations where it is functioning well enough, in order to explore a new space of hypotheses. The anxiety and unease that result are positioned not as mere dysfunction but as the exploration cost the self-system pays to acquire a broader predictability.

A reservation based on the previous section is needed here, however. For the reconstruction of the self to hold as an “active choice,” it is presupposed that the room for reconstruction is open to some degree. In situations where the pressure of forced alignment acts constantly, describing self-reconstruction purely as “active inference” risks reducing that coercion to a problem of individual choice. The reconstruction of the self must therefore be understood not as a task attributable only to individual activity, but as a practice linked to the social process in which the bias of social generative models is called into question.

Section 7. Differentiating Subjects

The previous section depicted the reconstruction of the self mediated by curiosity as an active rearrangement of information. But it must be made clear here that such self-reconstruction is not a universal attitude demanded of every subject.

In the position of this treatise, subjects are not beings sharing a single essence but prediction-error-minimizing systems that adopt different predictive strategies. It is therefore a theoretically natural consequence that, while there are “exploring subjects” who actively take on prediction error and explore higher-order order, there are also “stable conforming subjects” who conform to existing social generative models and conduct their lives within a stable order. This differentiation does not mean any superiority or inferiority of value between subjects; it is merely a difference in modes of adaptation to the information environment.

In a given society, exploring subjects attempt new informational orders while stable conforming subjects maintain the order of the whole. With both kinds of subjects present in suitable measure, prediction error in society as a whole is reduced and the stability of the system is kept. A balance is required: if exploring subjects become the majority, order wavers; with only stable conforming subjects, innovation stagnates.

The reservation of the previous section remains valid here too. That the existence of “stable conforming subjects” is a theoretically natural consequence must be distinguished from being forced to be a stable conforming subject through forced alignment. The former is affirmed as diversity of adaptive strategies, but the latter needs to be examined critically as a problem produced by the bias of social generative models.

Therefore neither a way of life that keeps actively reconstructing the self nor a way of life that entrusts itself to existing order and lives stably should be excluded, so long as each is taken as a voluntary choice. The former plays the informational role of carrying exploration and change; the latter contributes to the reproduction and stabilization of order. For the coexistence of these two kinds of subjects to hold as the basic structure of a post-anthropocentric society, it is a precondition that this coexistence function as diversity rather than coercion.

From this perspective, the end of anthropocentrism does not mean that the value of human life is denied. Rather, it means redescribing the human not as a privileged subject of purposes but as a being that can take diverse positions within an informational order. The reconstruction of the self depicted in this treatise is a concept that, presupposing such differentiation, still opens the possibility for subjects to take on their own positions, or to rearrange them.

Section 8. Summary

This chapter developed the central thesis at the existential level. The new points presented in this chapter are the following four.

First, it formulated social order as an information environment that self-organizes from the demand for mutual predictability.

Second, it positioned the contrast between contingency and planning not as a fixed essential distinction but as a dynamic asymmetry depending on the state of social systems.

Third, it raised forced alignment as a distinct problem that cannot be absorbed into a homogeneous description of prediction-error minimization.

Fourth, it positioned the reconstruction of the self not as a task attributable only to individual activity but as a practice linked to the re-examination of social generative models.

Conclusion: Findings and Remaining Tasks

Section 1. Summary of the Argument

This treatise has examined, step by step at the theoretical level (Chapter 1) and the existential level (Chapter 2), the central thesis that “self, society and value can be understood continuously as the self-organization of an informational order constrained by prediction-error minimization, and within that continuity existential self-renewal arises.”

Chapter 1 showed that, within the framework of the free energy principle and predictive processing theory, the self is understood as a collection of dynamic, updatable generative models. This not only implies that the reconstruction of the self is theoretically possible, but also means that such reconstruction can be described concretely as “the hierarchical reorganization of generative models.”

Chapter 2 extended this individual model into the dimensions of society and existence. The contrast between human contingency and mechanical planning was presented as a dynamic problem dependent on the state of social systems (Section 2); the aggregate of individuals’ behavior self-organizes social order (Section 3); as that order becomes naturalized it structurally limits the individual’s space of choice (Section 4); and the asymmetrical problem of imposing forced alignment on individuals with particular neurological traits was brought out (Section 5). Under these conditions the self can be actively reconstructed in the form of a hierarchical reorganization of generative models, but that practice is linked to calling into question the bias of social generative models (Section 6). Moreover, that reconstruction is positioned not as a universal duty but as an existential possibility premised on the differentiation of diverse adaptive strategies (Section 7).

Through this argument, the central thesis has been developed not as a mere claim of principle but as an existential theory under the concrete conditions of the naturalization of social systems and forced alignment. Existence in contemporary society lies not in recovering fixed meanings or purposes, but in the very possibility of taking on, within a naturalized information environment, at which level and as what predictive strategy one is positioned — and of rearranging it when necessary. What this treatise has presented is a framework for making this possibility theoretically visible and for keeping it open.

Section 2. Limitations of This Treatise and Future Tasks

This treatise has explicit limitations in the following respects.

First, the theoretical bridge from individual prediction-error minimization to the self-organization of social order is presented in this treatise as a working hypothesis; no design for empirical verification has been made.

Second, while this treatise has stated a position on the controversy over the falsifiability of the free energy principle itself, examining that controversy in detail is beyond its scope.

Third, discussing the problem of forced alignment within the framework of prediction-error minimization may not adequately address its political and ethical scope. Discussions of neurodiversity include dimensions of power and normativity that cannot be fully captured by information-theoretic description alone, and examining these separately is a task for the future.

Fourth, this treatise has only suggested the originality of the concept of “reconstruction of the self” in relation to prior work in existential philosophy. A systematic comparison with Heidegger, Sartre, Stoic philosophy and others is a task for the future.

References

Clark, A. (2016). Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford University Press.
Damasio, A. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain. Putnam.
Dennett, D. (1991). Consciousness Explained. Little, Brown.
Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.
Friston, K., et al. (2016). Active inference and learning. Neuroscience & Biobehavioral Reviews, 68, 862–879.
Friston, K., et al. (2017). Active inference, curiosity and insight. Neural Computation, 29(10), 2633–2683.
Heidegger, M. (1927). Sein und Zeit. Niemeyer. (English translation: Being and Time.)
Hobson, J. A., & Friston, K. J. (2014). Consciousness, dreams, and inference: The Cartesian theatre revisited. Journal of Consciousness Studies, 21(1–2), 6–32.
Hohwy, J. (2013). The Predictive Mind. Oxford University Press.
Klein, C. (2018). What do predictive coders want? Synthese, 195(6), 2541–2557.
Luhmann, N. (1984). Soziale Systeme. Suhrkamp. (English translation: Social Systems.)
Rao, R. P. N., & Ballard, D. H. (1999). Predictive coding in the visual cortex. Nature Neuroscience, 2(1), 79–87.
Ricoeur, P. (1990). Soi-même comme un autre. Seuil. (English translation: Oneself as Another.)
Sartre, J.-P. (1943). L’Être et le Néant. Gallimard. (English translation: Being and Nothingness.)
Schmidhuber, J. (2010). Formal theory of creativity, fun, and intrinsic motivation. IEEE Transactions on Autonomous Mental Development, 2(3), 230–247.
Seth, A. (2021). Being You: A New Science of Consciousness. Dutton.