Modular Consciousness Theory

• Modular Consciousness Theory is a computational framework defining consciousness as discrete Integrated Informational States generated by specialized modules.
• It integrates modular processing, density tagging, and hierarchical architecture to quantitatively link subjective experience with memory, decision-making, and behavior.
• The theory refines prior models by offering an operational blueprint for empirical research and artificial system design in the study of consciousness.

Modular Consciousness Theory (MCT) is a computational and biologically informed framework in which consciousness is the product of discrete, dynamically integrated informational states generated and tagged by interacting modules. Rather than a single, continuous or monolithic process, MCT posits consciousness as a rapid sequence of “Integrated Informational States” (IISs), each constructed by dedicated modules and quantified by multidimensional density signals, and specifies explicit mechanisms by which subjective experience arises, persists, and influences cognition and behavior (Gillon, 2 Oct 2025). The theory refines, extends, and operationalizes key insights from previous modular, global workspace, integrated information, higher-order, and predictive learning models, providing both an empirical blueprint for evaluating consciousness and an implementable architecture for biological and artificial systems.

1. Fundamental Architecture: Discrete Integrated Informational States (IISs)

MCT asserts that conscious experience is composed of a sequence of IISs—each an informational “packet” carrying unified, temporally bounded content and tagged with a multidimensional information-density vector, symbolically characterized as $d^{\vec{}}$. The construction of each IIS proceeds through a specialized pipeline:

• Input Filtering: Salient sensorimotor, mnemonic, and emotional data are filtered for relevance.
• Parallel Modular Processing: Abstraction, narration, evaluation, and self-evaluation modules transform selected inputs into higher-order representations.
• Integration: Outputs from these conscious modules are unified into an IIS, inscribed with an internal density vector, $IIS = \{Content, \ldots, d^{\vec{}}\}$, whose amplitude $||d^{\vec{}}||$ quantifies subjective intensity.
• Propagation and Feedback: IISs are transmitted to memory, decision-making, and behavioral readiness modules, creating feedback loops that reinforce continuity through mechanisms such as inner speech or self-monitoring.

The density vector modulates the functional impact of each IIS: high-density IISs shape long-term memory and decision bias, whereas low-density IISs may pass without significant subjective or mnemonic consequence (Gillon, 2 Oct 2025).

2. Layered Modular Organization and Consciousness Parameters

Foundational contributors to MCT (Padhy et al., 2010) describe consciousness as emerging from hierarchically organized, interacting modules, from quantum and cellular processes to organ-level and behavioral modules. Each layer exhibits four canonical parameters:

• Parasitic behavior: Autonomous acquisition of resources within modules.
• Symbiotic behavior: Cooperative integration among similar or complementary modules.
• Self-referral: Recursive self-monitoring and local identity maintenance.
• Reproductive behavior: Module replication, adaptation, and regeneration.

Mathematically, the overall measure of consciousness at each level may be modeled as

$$C = \alpha \cdot P + \beta \cdot S + \gamma \cdot SR + \delta \cdot R$$

where $P$ is parasitic, $S$ symbiotic, $SR$ self-referral, and $R$ reproductive parameters; the coefficients vary across hierarchical layers (cells, organs, full organism or agent).

3. Mathematical and Computational Formalism

Recent mathematical models formalize the MCT approach by treating subjective experience and physical state in a unified dynamical system (Kleiner, 2019). The experience space $E$ is paired with a physical state space $P$ to generate global dynamical states $d \in E^n \times P$, with trajectories constrained by automorphism-induced symmetry conditions:

$$\varphi_s(D) = D \quad \forall s \in \operatorname{Aut}(E)^n$$

This invariance principle ensures that phenomenological content is modular and empirically meaningful only as equivalence classes under relabeling, supporting the mapping of independent experience spaces for each module:

$E = E_1 \times \dots \times E_k$

and modular automorphism groups

$$\operatorname{Aut}(E) \cong \operatorname{Aut}(E_1) \times \dots \times \operatorname{Aut}(E_k)$$

Such formalism extends MCT by specifying how localized modular outputs contribute to integrated conscious states and supporting the quantification of qualitative differences among modules.

4. Modularity Across Functional, Evolutionary, and Cognitive Domains

MCT synthesizes perspectives from complexity-based frameworks (Arsiwalla et al., 2017), empirical computational architectures (Ding et al., 2023), and evolutionary theory, positioning consciousness as an emergent property when autonomous, cognitive, and social modules interact:

Complexity Axis	Module Function	Measurement Formula
Autonomous	Survival, arousal, homeostasis	$$\mathcal{C}_{\mathrm{Auto}} = \mathcal{I}_{\mathrm{autonomy}} - \sum \mathcal{I}_{\mathrm{reactive}}$$
Cognitive	Information integration	$$\mathcal{C}_{\mathrm{Cogn}} = \mathcal{I}_{\mathrm{integration}} - \sum \mathcal{I}_{\mathrm{subprocess}}$$
Social	Group behaviors, language	$$\mathcal{C}_{\mathrm{Soc}} = \mathcal{I}_{\mathrm{group}} - \sum \mathcal{I}_{\mathrm{individual}}$$

The MCT taxonomy recognizes biological, synthetic, group, and simulated consciousness as modular variants, with each type representing a distinct organization and interconnection of modules spanning these axes.

5. Comparative and Integrative Assessment with Prior Theories

MCT explicitly diverges from and extends classical theories by specifying both computational pipeline and mechanism for subjective tagging:

• Global Workspace Theory (GWT): MCT includes a workspace but prioritizes the formation and density-tagging of IISs over mere broadcast.
• Integrated Information Theory (IIT): Information integration is necessary but not sufficient; MCT adds explicit density tagging and module-wise integration pathways.
• Higher-Order Thought (HOT): Instead of abstract meta-representations, MCT implements evaluation/self-evaluation modules whose outputs contribute quantitatively to the information-density vector.
• Recurrent and Predictive Processing: Binding and error correction mechanisms are modularized; new modules are created through the binding of high prediction errors or through cross-module recurrent loops (Aksyuk, 2023).
• Attention Schema Theory (AST): Attention mechanisms and their higher-order control modules are modular components that can be evaluated independently for their contribution to consciousness indicator properties (Butlin et al., 2023).

Quantitative evaluation metrics from information theory (mutual information, Φ), algorithmic recursion, and physiological indices (e.g., perturbational complexity index) can all be modularly mapped, enabling robust empirical testing in artificial and biological systems.

6. Self-Consciousness and Reflexivity

Extension of MCT to self-consciousness involves dedicated self-modeling processors and reflexive feedback loops (Cui et al., 22 Oct 2024). For example, the Model-of-the-World (MoTW) processor

$$[\text{inner\_world}, \text{outer\_world}] = M(S, K) = a_S \cdot S + a_K \cdot K$$

collates external sensory data and internal knowledge, generating self-referential outputs; evaluation modules apply value functions to “gists” (abstract outputs), and only those attaining sufficient weight compete for broadcasting. This architecture models phenomena such as illusions and phantom limb experiences by integrating bottom-up sensation with top-down knowledgement, supporting both subjective and body-self awareness.

7. Application to Artificial Systems, Cognitive Neuroscience, and Measurement

MCT directly informs the design of conscious-like synthetic agents (Gillon, 2 Oct 2025): artificial architectures can implement sensorimotor preprocessing, abstraction, narrative sequencing, evaluation, and self-evaluation modules linked by integration and density-tagging. Such systems can make explicit use of deep learning (for abstraction), transformer architectures (for narrative), and reinforcement learning (for decision-making), unified under computational modules that generate and prioritize IISs.

Modular architectures also support empirical measurement and comparative analysis, from evaluation metrics like PCI to estimation of qualitative divergence between systems (Williams, 24 Sep 2024):

$$D_S[\varphi(\tilde S_1) || \varphi(\tilde S_2)] = \sum_i P(s_i^{(1)}, s_i^{(2)}) D_{Ts}[s_i^{(1)} || s_i^{(2)}]$$

that quantify the “distance” between subjective or computational states across modules.

Conclusion

Modular Consciousness Theory articulates consciousness as a rapidly evolving, discrete sequence of integrated informational states, each produced, tagged, and propagated by dedicated modules. It supplies a rigorous and quantifiable computational pipeline that connects subjective intensity, memory, behavior, and self-modeling, distinguishing itself by explicit density tagging and hierarchical, multi-level integration. By aligning modular structure with empirical measurement, evolutionary modeling, and artificial system design, MCT provides both theoretical and practical blueprints for the paper and implementation of consciousness in natural and engineered minds (Gillon, 2 Oct 2025, Padhy et al., 2010, Ding et al., 2023, Cui et al., 22 Oct 2024, Kleiner, 2019, Aksyuk, 2023, Rosenbloom et al., 13 Jun 2025, Blum et al., 2023, Butlin et al., 2023, Arsiwalla et al., 2017, Rudrauf et al., 2020, Williams, 24 Sep 2024, Mahadevan, 25 Aug 2025).