The phenomenon commonly referred to as consciousness has been interpreted through numerous theoretical frameworks, including global workspace architectures, high-dimensional causal structures, hierarchical generative models, and narrative or representational accounts of the self. Each framework emphasizes distinct functional properties of cognitive systems, yet none individually captures the full range of mechanisms contributing to conscious experience. A more comprehensive account can be generated by integrating principles from predictive processing, global broadcasting, and structural integration.
The position developed here presents consciousness as emerging when a hierarchical predictive system generates temporally deep expectations concerning its own state trajectories and when these predictions are integrated and broadcast across the system in a manner that stabilizes a functional representation of a self. This “self” is not treated as a discrete entity or metaphysical subject but as a set of the system’s most temporally stable and globally influential high-level priors. These priors act as organizational anchors within the generative model, guiding inference and action across multiple timescales.
2. Predictive Processing as the Base Mechanism
At the lowest theoretical level, the brain can be modeled as a hierarchical generative system that seeks to minimize variational free energy. Each layer of the hierarchy generates predictions concerning expected sensory or internal states and updates these predictions based on incoming prediction-error signals. This principle applies across several modalities:
- Exteroception: predictions about external sensory inputs and environmental causes.
- Proprioception: predictions concerning bodily position and motor configuration.
- Interoception: predictions regarding internal physiological states.
- Cognitive and attentional dynamics: predictions about future representational, attentional, motivational, or decision-related states.
Crucially, these predictions do not represent objects, features, or qualia directly. Rather, they encode conditional probability distributions over the system’s own future sensory and internal states, constrained by environmental regularities and the organism’s embodied structure. This framing treats the organism as continuously predicting its evolving state under the assumption of ongoing causal interactions with its environment.
As numerous predictive hierarchies interact, precision-weighted connections determine the relative influence of prediction errors at each level. When the system stabilizes high-level patterns of prediction that constrain activity across multiple subordinate layers, the aggregate effect is the formation of a relatively stable and coherent self-model—one that reflects long-term regularities in the organism’s dynamics.
3. Emergence of the Self-Model
A functional notion of selfhood emerges not through an explicit internal representation of a subject but through the convergence of high-level priors that remain stable across time, tasks, and perturbations. These priors include:
- persistent expectations concerning relationships among internal states.
- generalizable predictions about behavioral tendencies across contexts.
- stable affective and motivational set points.
- long-horizon predictions concerning bodily and cognitive trajectories.
Over time, the hierarchical generative model learns that maintaining coherence among these high-level priors yields consistent reductions in prediction error and environmental uncertainty. Consequently, these priors become entrenched and exert top-down influence on lower-level predictions across modalities. Because they modulate or constrain inference globally, they appear functionally central to the system’s organization.
This centrality forms the basis of the emergent self-model. The self is not treated as an ontologically distinct entity but as a structural feature of the generative hierarchy—specifically, the set of priors whose relative stability provides the system with long-term predictive continuity. The subjective impression of a persistent “I” corresponds to the stability of these high-level organizing principles.
Thus, the self-model is understood as a fixed point arising from the need to coordinate predictions across a temporally deep and spatially distributed system. The persistence of these priors across time yields the impression of personal continuity.
4. Consciousness as the Globally Integrated Component of Predictive Dynamics
Not all predictive or inferential processes in the system correspond to conscious experience. Many operate implicitly, without global influence or reportability. Consciousness, within this framework, arises when predictive dynamics satisfy several conditions:
Sufficiently high level of abstraction: The processes involved must operate at a level of generality that influences long-term priors or global organizational variables.
Global accessibility or broadcasting: Predictive contents must become available to a large subset of specialized systems, enabling modification of ongoing inference, decision-making, memory encoding, and behavior.
Integration within a unified causal structure: The neural assemblies supporting these predictions must form a sufficiently integrated informational system capable of maintaining coherent states over time.
When prediction errors or high-level inferential updates meet these conditions, they are integrated across the system and become candidates for conscious experience. Within this model, consciousness is associated with the subset of predictive dynamics that have global influence, update high-level priors, and participate in the ongoing maintenance of the self-model.
These globalized contents are those available for report, introspection, voluntary action, and longer-term consolidation. Consciousness, therefore, reflects not the entirety of predictive processing, but the portion that participates in and contributes to the global modification of the system’s generative structure.
5. Temporal Depth and Self-Prediction
From a free-energy minimization standpoint, systems must predict not only external causes but also their own future internal states, actions, and trajectories. Anticipating internal consequences is essential for maintaining physiological and cognitive stability. The generative model thus incorporates temporally extended predictions concerning:
- expected trajectories of sensory states.
- propagation of motor commands and resulting bodily changes.
- anticipated shifts in affective or motivational state.
- transitions of attentional or cognitive focus.
- internal configuration changes at multiple hierarchical levels.
These temporally deep predictions enable the system to maintain stability, avoid large unexpected deviations, and coordinate behavior across time. Because the organism continuously predicts the future evolution of its own states, high-level self-relevant priors become central organizational variables. When these predictions are coordinated and integrated across the global workspace or analogous mechanisms, the system experiences them as belonging to a unified subject.
This provides a functional basis for selfhood: the system models its own future in ways that require coordinated, system-wide participation. The resulting representational center emerges from the dynamics of prediction and integration, not from any inherent subject or entity within the architecture.
6. Compatibility with Computational and Hybrid Accounts
The hybrid model described here does not require anti-computational commitments. It remains compatible with computational, representational, and mechanistic approaches while integrating insights from multiple theoretical frameworks:
From the Global Workspace Theory (GWT) tradition, it incorporates the notion that certain contents become globally accessible and thereby capable of influencing widespread cognitive operations.
From Predictive Processing (PP), it adopts the core principles of hierarchical generative modeling, prediction-error minimization, and temporally deep inference as the base mechanism for cognition.
These components can be synthesized without internal contradiction when interpreted functionally rather than metaphysically. Integration provides a model in which consciousness corresponds to globally integrated, temporally deep predictive activity that contributes to the formation and maintenance of high-level priors constituting the self-model.
My apologies if I have made any mistakes or anything, this is my first time posting my thoughts seriously. I hope I have given you something to think about!