Common Model of Cognition

A common model of cognition is a cross-disciplinary theoretical and computational framework formulated to explain cognition as it manifests in natural and artificial systems, emphasizing layered abstraction, agent-based computation, physically grounded processes, and scalable synthetic modeling. Within the info-computational constructivist paradigm, this model treats cognition and life as coextensive phenomena realized through the dynamic interplay of informatic structure and computation across multiple organizational levels (Dodig-Crnkovic, 2013 ).

1. Hierarchical Levels of Cognition

The framework proposes that cognition is expressed across hierarchical, nested levels of organization, each characterized by its own networks of information exchange and computational dynamics:

• Molecular Level: Molecules considered as agents exchange information via chemical interactions. Modeling at this level captures abiogenic forms of cognition.
• Cellular Level: Cells process environmental information and respond with adaptive actions. Activity here is represented as inter-agent information flow (e.g., signal transduction pathways).
• Organismic Level: Whole organisms demonstrate adaptive information processing, learning, and memory, emerging from complex internal interactions.
• Societal Level: Collections of organisms constitute social systems, giving rise to distributed, collective forms of cognition and social learning.

This layered structure is formally described as information network hierarchies, each layer embedding agents that participate in computation through message exchange and structural reconfigurations. The approach allows direct construction and analysis of synthetic agents at each level, relevant for both biological and artificial modeling efforts.

2. Info-computational Constructivism

Info-computational constructivism integrates two foundational principles:

• Informational Structural Realism: Information constitutes the structural, observer-relative substrate of reality. States and relationships measurable by difference define the information content.
• Natural Pancomputationalism: Computation is the universal, dynamical process acting on information. All natural phenomena—including cognition—are thus regarded as computations over informational structures.

The relationship is succinctly captured as:

$\text{Computation} = \text{Dynamics of Information}$

All processes of cognition (perception, learning, reasoning) are seen as evolutions or transformations of informational structures via computation, which encompasses not just symbolic or digital operations, but also those instantiated by physical law (mechanical, chemical, thermodynamic). Morphological computation is central: the physical structure and evolution of an agent itself encodes and enacts information processing.

3. Agents and the Actor Model

Within this model, an agent is defined as any entity capable of autonomous action on its own behalf. The conceptual foundation is provided by the Hewitt Actor Model of computation:

• Actor: The primitive computational unit, which exchanges messages, makes local decisions, and may create new actors. Communication is asynchronous; there is no global state.
• Molecular Application: Even molecules can be so modeled, with chemical interactions formalized as message exchanges and system state transitions.

The information-exchange dynamics among agents is formalized as:

$A_1 \xrightarrow{M} A_2$

where $A_1$ sends message $M$ to $A_2$, prompting state change. This scalable abstraction applies across levels, from molecules to multicellular collectives.

4. Living Agents and Cognition as Life

A crucial extension, drawing on Kauffman's definition, posits that a living agent must:

• Be capable of self-reproduction.
• Undergo at least one thermodynamic work cycle.

This definition establishes a physical criterion for autonomy and persistence. Maturana and Varela's identification of "life is cognition" underlies the view that cognition is synonymous with a system's autopoietic (self-producing and self-maintaining) character. Thus, all living (autopoietic, thermodynamically cycling) systems, regardless of complexity, instantiate cognition by adapting and persisting through informational and energetic transactions, without the necessity for anthropomorphic constructs such as beliefs or desires.

5. Construction of Synthetic Cognizing Agents

The info-computational approach enables construction of synthetic model classes representing artifactual cognizing agents:

• Scale: Models range from artificial cells (automata), through organismic robots, to societal-level multi-agent systems.
• Mechanisms: Synthetic agents are endowed with self-organization, autopoiesis, adaptation, memory, anticipation, and evolution, all realized as information-processing structures and dynamics.
• Morphological Computation: The knowledge and cognitive capability of the agent are physically instanced in its structure and real-time interactive evolution, not just in abstract symbolic representations.

Architecturally, the Actor model provides a template for distributed, asynchronous computation; in practice, architectures such as cellular automata or agent-based systems operationalize these principles for programming and simulation of cognition across scales.

Representative Equations

• Information as physical difference:

$\Delta S_1 \rightarrow \Delta S_2$

where a detectable change in system $S_1$ produces a corresponding difference in $S_2$.

• Agent/Actor communication:

$\forall A \in \text{Actors}, M: \quad A(M) \to \text{changes}(A)$

• Thermodynamic work cycle (living agent):

$\text{Agent}(t) \xrightarrow{\text{work cycle}} \text{Agent}(t+\Delta t) + \text{Products}$

6. Unified Model: Implications and Significance

• Reality for an agent is informational: Cognition is computation over an agent-relative reality comprised of informatic structures.
• All agents are actors in a network: They exchange and process information, redefining cognition as the organization of message exchanges and control over transformation.
• Living cognition is rooted in physical cycles: The core of cognition in living systems is not mere symbolic or algorithmic manipulation, but the maintenance and adaptation of informational structure through physically grounded cycles.
• Synthetic modeling across levels: The framework supports building agents and systems with naturalistic cognitive properties at multiple scales, aligning artificial and natural cognition.
• Grounded, embodied, embedded: Cognition emerges not in isolation, but from the interplay of structure, process, and environment.

This synthesis provides a formal system for both analyzing and designing cognizing agents, with implications spanning origins-of-life studies, artificial intelligence, systems biology, and embodied robotics. The info-computational common model of cognition thus endeavors to unify the paper and engineering of cognition as it manifests—from the smallest molecules to the largest societies—within a rigorously defined, physically grounded, and evolutionarily scalable framework.