Free Energy Principle and Consciousness, World and Self Theories of Consciousness

World and Self Theories of Consciousness

As we discussed in the prior post in this series, consciousness, a complex and multifaceted phenomenon, has been the subject of various theoretical frameworks. Among these, the Integrated World Modeling Theory (IWMT) and Self-Modeling Theory (SMT) offer somehow related perspectives on how consciousness arises and its functions. IWMT suggests that consciousness is the result of generating predictions about sensory states based on learned causal world models, emphasizing the integration of spatial, temporal, and causal experiences. SMT, on the other hand, focuses on the self as an adaptive hypothesis, integrating sensory information into a generative model that situates the self as the center of control. The Inner Screen Model highlights the role of attentional focus and interoceptive self-modeling in constructing first-person experiences of selfhood and consciousness. Together, these theories contribute to a comprehensive understanding of consciousness, exploring how the brain synthesizes information to create a coherent, subjective experience of the world and self.

Integrated world modeling theory (IWMT) is an attempt to identify the underlying principles of theories of consciousness, proposing that consciousness corresponds to the ongoing generation of predictions about an embodied agent’s sensory states which are conditioned on causal world models learned through goal-pursuit. IWMT claimed that consciousness depends on organizing experience according to the Kantian categories of space, time, and cause, which afford compositionality, prediction, and modeling respectively. These properties require forms of agency that allow learning egocentric perspectives and evaluating counterfactuals, enabling subjectivity as a “point of view” on a “lived world.” According to IWMT, consciousness is “what it feels like” to generate an integrated representation of the body and world by synergistically combining relevant information sources. This integrated world model enables adaptive action selection and iterative refinement of policies and models. IWMT theory proposed that informational integration is facilitated by “self-organizing harmonic modes” (SOHMs) that are hierarchically structured, with smaller modes modeling rapidly evolving events and larger modes enabling more complex self-and-world-modeling. If sufficiently coherent multi-modal integration with egocentric frames is achieved, an agent may iteratively estimate the likely causes of its sensory states, realizing a stream of experience from a point of view on a “lived world.”

The self-modeling theory (SMT) focuses on how the concept of self and experience of selfhood, identifing the self as the center of control that connects bodily sensations with movement in external space. The self emerges from a process that integrates sensory information into a generative model representing the self as the cause of those sensations. The self-model is egocentric, centered on the individual’s perspective, encompassing both the subjective experience and objective experiences of the self as an entity conserved over space and time. According to SMT, the self is not static, but results from dynamically inferring the probable causes of an agent’s sensory states. This active self-model operates beneath conscious awareness, generating the experience directed towards objects and enabling world interaction. SMT also distinguishes between “transparent” mental states, where the process enabling the experience is inaccessible, and “opaque” states where the process itself enters into the content of experience. The self is ultimately an adaptive hypothesis about what causes one’s sensory experience. In this framework, conscious experience is constructed from multimodal bodily inputs, especially interoceptive and proprioceptive signals. The brain leverages a generative model with priors about the causes of these interoceptive signals. The conscious self emerges from generating predictions about actions and managing related prediction errors from interoceptive and exteroceptive brain centers. This approach links the stability of embodied selfhood to the differing timescales of motor actions and physiological homeostasis. Within this context, and according to felt uncertainty theory, conscious feeling is the experience of a system that minimizes free energy given homeostatic setpoints or “drives.” The locus of basic consciousness in the brain is proposed to be a set of upper brainstem structures, with feeling itself deriving from brainstem activity. This theory suggests that the contributions of the cortex to consciousness have been overemphasized, and that the perceptual contents of conscious experience arise from the top-down constraining effect of cortical structures on brainstem activity.

The brain contains a representation of the internal body (homeostatic needs or drives and error signals) and a representation of the external body (as an object of action and perception). Individuals form a map of their own bodies as “external objects” through spatially distributed sensory receptors. The affective representation of the internal body corresponds to need-detector systems (the Freudian “id”), regulating the body through homeostasis. The external, third-person representation of the body is a part of the generative model representing external objects. The maintenance of the external milieu depends on that of the internal milieu, as homeostatic perception and control is a prerequisite for allostatic perception and control. Individuals leverage memories of past events to build up the external body representation and cognition, repurposing them for counterfactual and imaginary projection into the future. The same brain regions (thalamo-cortico-thalamic circuits) are involved in these processes, implementing the same functions for perceiving one’s body and other bodies as objects in the world, while the “felt uncertainty” refers to an agent’s fluctuating confidence in its currently prioritized effort to maintain homeostasis.

The inner screen model suggests that nested (e.g. Russian dolls) Markov blankets allow agents to selectively focus on internal and external evidence while enabling a form of interoceptive self-modelling, while constructing first-person experiences of selfhood through internal representations of mental states. Self-modelling involves both a first-person perspective, where the self is a constant locus of sensory states, and a third-person perspective, where the self is an object in the world. Conscious systems model their physical embodiment in the world, integrating experiences with a temporal structure that reflects causal dependencies between internal and external states. This process allows agents to plan and act over extended timescales.

The inner screen model explains the stability of self-perception despite fluctuating sensorimotor processes, it remains stable while continuously updating its statistical structure. The model aligns with the holographic principle, suggesting that the projective structure of consciousness facilitates effective information gathering and action. The inner screen as the convergence point of dynamic mirrors where the brain uses prior beliefs to predict interoceptive signals, leading to conscious perception. This perspective allows agents to selectively choose and focus on specific aspects of their environment, guided by the expected value of the information. The Integrated World Modeling and Self-Modeling Theory each provide valuable insights into the nature of consciousness with the inner screen being their convergence point. IWMT highlights the role of predictive modeling and sensory integration, framing consciousness as the emergent property of complex world modeling. SMT emphasizes the self as an adaptive generative model, underpinning the continuous inference of sensory experiences and bodily states. The Inner Screen Model integrates these views by focusing on attentional mechanisms and the stability of self-perception through dynamic prediction error management. Together, these theories highlight the processes by which the brain synthesizes sensory information and cognitive functions, constructing a coherent and adaptive experience of self and world.

The inner screen model aligns closely with the IWMT, which proposes that consciousness arises from the brain’s ongoing generation of predictions about sensory states based on learned causal world models. Both theories emphasize the importance of integrating spatial, temporal, and causal experiences to create a coherent conscious experience. The inner screen can be viewed as the canvas upon which these integrated world models are painted, providing a unified space for the synthesis of various sensory and cognitive inputs. In the context of IWMT, the inner screen serves as the interface between the agent and its environment, facilitating the organization of experience according to the Kantian categories of space, time, and cause. The Self-Modeling Theory SMT provides another complementary perspective to the inner screen model. SMT focuses on how the concept of self and the experience of selfhood emerge from integrating sensory information into a generative model that represents the self as the cause of those sensations. The inner screen theory aligns with this view by suggesting that consciousness involves both a first-person perspective, where the self is a constant locus of sensory states, and a third-person perspective, where the self is an object in the world. This dual perspective facilitated by the inner screen allows for the dynamic inference of the probable causes of an agent’s sensory states, as proposed by SMT. The inner screen becomes the platform where the active self-model operates, generating experiences directed towards objects and enabling interaction with the world. It provides the space where “transparent” mental states, where the process enabling the experience is inaccessible, and “opaque” states, where the process itself enters into the content of experience, can coexist and interplay.

The integrative aspect of the inner screen model is particularly relevant when considering complex cognitive phenomena such as imagination, memory recall, and future planning. The same neural mechanisms that project sensory information onto the inner screen can be repurposed for these higher-order cognitive functions. This explains how we can vividly imagine future scenarios or relive past experiences in our mind’s eye. The inner screen theory also attempts to explain the relationship between consciousness and attention. The selective focus enabled by the nested Markov blankets allows agents to choose and concentrate on specific aspects of their environment, guided by the expected value of the information. This selective attention mechanism is crucial for efficient information processing and decision-making, aligning with the free energy principle’s emphasis on efficient inference and action selection. By framing consciousness as a projection onto an inner screen, it provides a compelling explanation for the integrated nature of our subjective experiences, the stability of our self-perception, and the flexibility of our cognitive processes. It aligns with and extends other prominent theories of consciousness, offering a holistic view of how our brains create our lived experience of the world. As research in neuroscience and cognitive science continues to advance, the inner screen theory stands as a valuable conceptual tool for understanding the complex and fascinating phenomenon of consciousness.

References

#Free Energy Principle    #Self    #Consciousness   

Octobre 2024