The Evolution of Intelligence from Active Matter to Complex Intelligent Systems via Agent-Based Autopoiesis ^†

Abstract

Intelligence is a central topic in computing and philosophy, yet its origins and biological roots remain poorly understood. The framework proposed in this paper approaches intelligence as the complexification of agency across multiple levels of organization—from active matter to symbolic and social systems. Agents gradually acquire the capacity to detect differences, regulate themselves, and sustain identity within dynamic environments. Grounded in autopoiesis, cognition is reframed as a recursive, embodied process sustaining life through self-construction. Intelligence evolves as a problem-solving capacity of increasing organizational complexity: from physical self-organization to collective and reflexive capabilities. The model integrates systems theory, cybernetics, enactivism, and computational approaches into a unified info-computational perspective.

1. Introduction

The origins of cognition and intelligence remain among the most profound and unresolved questions in science and philosophy. How do systems come to know, adapt, and act meaningfully and intelligently within their environments? This research proposes a naturalized, layered, agent-based computational framework for understanding the emergence of cognition and intelligence seen as a natural process. From the self-organizing dynamics of active matter, via the cognition and intelligence of living organisms, to the symbolic and collective intelligence of human culture, we explore how agents—material systems capable of interactive behavior—give rise to increasingly complex forms of intelligence.

Agency, in this view, is not confined exclusively to conscious minds. It emerges gradually from the intrinsic activity of matter—from the capacity of particles, molecules, and living cells to organize, adapt, and respond to their environments [1]. Through processes of self-assembly, self-organization [2], and autopoiesis, even simple living systems possess the basic features of cognition: persistence, responsiveness, and the capacity to maintain distinction from their surroundings [3].

These natural agents can be efficiently modelled computationally. Hewitt’s Actor Model [4] provides a computational paradigm for understanding agent systems as distributed, message-passing agents embedded in dynamic interaction networks. Such models reveal how complex patterns of behavior can emerge from relatively simple local rules, mirroring how molecular systems and cellular pathways exhibit rudimentary forms of memory, regulation, and adaptability [5,6].

When cognition is understood not as a property of brains, but as a process intrinsic to life itself—as Maturana and Varela [7] and Stewart [8] have argued—then even the simplest organisms become cognitive agents. Bacteria, for example, sense, make decisions, and adapt their behavior to internal and external conditions. They do not think in a reflective sense, but they do process information in ways that sustain their autonomy and viability [9]. In this light, cognition and metabolism appear as two sides of the same process: the regulation of internal order in the face of external perturbation.

Thus, we define cognition as the process by which an organism acquires, transforms, stores, and uses information to regulate its behavior and interactions with the environment. Its origins may be searched for in active matter [1], in chemical and biological networks [5,6], and in the process of complexification towards human-level cognition. Intelligence in its most elementary form is the ability of a system (an organism or an artifact) to learn, solve problems, and adapt. It can be understood as a system’s capacity to pursue goals flexibly and adaptively in response to information, across time and scale, as argued by Levin [5].

This approach enables bridging multiple levels of organization—from molecular agents to ecosystems, from single-cell cognition to socially constructed meaning. It supports the synthesis of biological, physical, and computational insights into a coherent framework for asking fundamental questions: How does cognition emerge from matter? How do agents organize themselves into hierarchies of increasing complexity? Can artificial systems replicate or extend these dynamics?

The following sections present a layered framework for the development of autopoietic agents, from molecular sensing to symbolic selfhood and cultural cognition. At each stage, we examine how distinctions are drawn, how information becomes meaningful, and how the self emerges as an ongoing process of self-construction and self-regulation.

2. Cogito Ergo Sum: I Sense, Therefore I Am

At the heart of contemporary enactive and autopoietic thought lies a generative insight: cognition = life. First articulated by Maturana and Varela [7] and reaffirmed by Stewart [8], this principle holds that the very processes through which living systems sustain their own identity—autopoiesis—are also the roots of their knowing. To live is to know—not in the abstract, conceptual sense, but in a more fundamental way: to sense, to be affected, to respond.

This broader understanding resonates with a deeper reading of Descartes’ famous “Cogito ergo sum”. In its original Latin, cogito derives from co- (“together”) and agitō (from agō, “to drive, to stir”), rooted in the Proto-Indo-European *h₂eǵ-, meaning “to set in motion”. Thinking, in this etymological lineage, is not passive reflection but a gathering of motion—a driving together of inner activity, a dynamic organization of sensing and response.

This insight finds validation in biological research. Biganzoli and Bollati [9] propose that even at the cellular level, life exhibits a form of epigenetic intelligence—a hybrid system of symbolic and subsymbolic regulation in which the genome functions as a dynamic, information-processing architecture. Cells “think” by reorganizing internal states in response to signals—cogito as metabolic modulation.

Cognition, in this context, begins with sensitivity—with the living body’s capacity to be perturbed and to respond coherently. In many languages, the verbs “to know,” “to feel,” and “to perceive” overlap, echoing this deeper, pre-reflective substrate of all knowing.

3. Cellular Autopoiesis: The Origin of Distinction

A single cell, such as a bacterium, enacts its own boundary through metabolic self-production. It distinguishes itself from the environment not as a symbolic act, but as a structural necessity. Its membrane is not merely a container—it is a selective interface, enabling sensing and exchange with the environment.

Cells respond to gradients of chemicals, light, pressure, and temperature. These stimuli become meaningful only when they “make a difference” for the cell, i.e. present information according to Bateson’s classic definition [10], where a difference is a distinction in the sense of Spencer-Brown [11] for the process of cellular autopoiesis [12]. A sugar molecule becomes meaningful for a cell when it triggers its response.

Thus, the cell is already a basal cognitive agent. It registers distinctions, processes environmental inputs, and acts to preserve its identity. This identity is maintained through an internal network of biochemical processes—what Stewart [8] and Lyon [13] describe as metabolic selfhood.

Biganzoli and Bollati [9] show that cellular behavior also exhibits learning and adaptability through epigenetic regulation. Methylation, histone modification, and transposable elements create a probabilistic system of internal change, allowing cells to adjust their logic in response to environmental perturbations.

Cognition here is not added onto life—it is immanent in living systems, beginning with the very act of drawing a distinction between self and world.

4. Non-Neural Multicellular Autopoiesis

In simple multicellular organisms, such as plants, fungi, and some invertebrates, coordination exceeds single-cell regulation. These systems respond to environmental inputs—light, gravity, humidity—through distributed processes such as morphological change, electrical signaling, and hormonal gradients. Agency here is expressed not through locomotion, but through form and growth. Roots grow toward water, mycelial networks redistribute nutrients—these are adaptive, regulated processes of interaction with the environment.

These organisms possess cognition without neurons. They integrate information, distinguish relevant inputs, and modulate internal activity to sustain themselves. As Di Paolo [12] argues, their agency reflects sensorimotor autonomy, even when movements are extremely slow. Recent studies of plant and fungal cognition [3] suggest that even in the absence of a nervous system, organisms exhibit learning-like dynamics, including habituation, memory, and anticipation. Morphogenesis itself becomes a cognitive process—a mode of structural knowing.

The self at this level is embodied in form, shaped by interactions with environmental constraints. These non-neural agents show that cognition is not tied to synapses or brains, but to recursive autopoietic regulation.

5. Sensorimotor Autopoiesis: Perception and Movement

With the evolution of nervous systems, animals develop a new mode of information processing: the sensorimotor loop. Sensory input is no longer passively received; it is dynamically coupled with movement. Perception and action form a feedback circuit, enabling more flexible, adaptive behavior.

As Varela, Thompson, and Rosch [14] argue, this marks the beginning of enactive cognition—knowing through embodied engagement. An animal perceives not by analyzing a static world, but by moving within it, actively modulating its sensory field. Agency becomes immediate and flexible. Organisms can learn from experience, explore new strategies, and anticipate future states. Movement is not just locomotion—it is part of sense-making.

Damasio [15] describes this level as involving a proto-self: a system that maps its own internal states and regulates them through interaction. The body becomes a site of affective evaluation, where perception is filtered through needs, preferences, and learned associations.

The self emerges through interactions. Boundaries between self and world are mediated by habit, memory, and motor coordination. Cognition evolves into a temporally extended process—a trajectory of embodied responsiveness and adaptive modulation.

6. Symbolic/Reflective Autopoiesis: Language and the Narrative Self

Humans and other advanced primates develop a further level of autopoietic organization: the conscious symbolic domain. Through language, imagination, and reflective thought, cognition becomes recursive. We do not only act—we can observe, simulate, and revise our actions. Here, symbols play a far more prominent and explicit role than that played by, for example, by the physical symbols in the “chemical language” of bacteria on the level of basal cognition [13] or in cellular neuro-symbolic epigenetic regulation [9].

This allows for narrative identity—a self, constructed not only through immediate experience, but through stories, intentions, and memories. Damasio [16] refers to this as the autobiographical self, emerging from the integration of perception, affect, and symbolic reasoning. Symbols allow internal modeling, the simulation of futures, and the reconstruction of pasts. Agency becomes reflexive: we can not only respond to situations, but we reinterpret and reconfigure them. This mode of cognition enables humans to question, choose, and transform themselves. The symbolic self is not fixed—it is a product of ongoing interpretation and meaning-making. It is shaped by language, culture, and internal narrative.

Importantly, symbolic cognition builds upon all previous layers. It is grounded in the body with all its cellular networks of networks, with sensorimotor experience, and social and developmental history. It integrates and extends earlier forms of knowing into abstract, temporally nested forms of selfhood.

7. Social/Cultural Autopoiesis: The Collective Self

At the highest level, humans participate in socially distributed systems of cognition. Language, norms, institutions, and traditions form autopoietic networks of communication that persist through time by producing and reproducing their own distinctions [14].

These systems—legal, educational, scientific, religious—are cognitive in their own right. They store memory, coordinate action, and evolve over time. No single individual controls a language or culture, yet these systems exhibit adaptive behavior and collective agency. As Luhmann [17] explains, social systems regulate themselves through recursive distinctions—such as legal/illegal or sacred/profane—processed through communication. These distinctions shape the behavior of individuals and groups, functioning as informational constraints and affordances.

Agency at this level is distributed and participatory. The self becomes multiple: we inhabit identities, roles, and symbolic affiliations. Our sense of who we are is co-constructed through dialogue, recognition, and social practice. Culture thus becomes the medium through which collective cognition unfolds. Ideas, values, and technologies co-evolve with the minds that generate them. Intelligence, in this context, is not just individual but shared, historical, and networked.

8. Distinction and the Layered Self: From Matter to Meaning

This layered model of autopoiesis is grounded in two foundational cybernetic principles: Gregory Bateson’s definition of information as “a difference that makes a difference” [10], and George Spencer-Brown’s formalization [11] of the act of drawing a “distinction”.

In Bateson’s view, information is not an objective entity but a relational event—a perturbation that becomes meaningful only when it alters the internal state or behavior of a system. Cognition begins with sensitivity to difference, which allows a cognizing agent to modulate its actions in relation to its environment.

Spencer-Brown extends this by showing that the act of drawing a distinction is the basis of all form, perception, and logic. To distinguish is to define an inside and an outside—a self and a world. This act underlies every cognitive system that can regulate its own identity.

Seen in this way, cognition is a recursive elaboration of distinction-making: at the cellular level, the membrane distinguishes interior from exterior; in multicellular organisms, morphogenesis distinguishes organs and functions; sensorimotor systems create distinctions of perception and action; symbolic systems introduce distinctions in concepts, language, and narrative; social systems formalize distinctions in laws, values, and institutions.

Each layer enables the agent to process new kinds of difference, supporting new forms of regulation, adaptation, and identity. The self becomes a historically and developmentally layered process, shaped by recursive loops of difference-making—from metabolic perturbations to cultural meaning systems.

Thus, the evolution of cognition can be seen as the evolution of increasingly sophisticated agents of distinction: systems that generate and regulate differences in order to preserve and transform/adapt/evolve themselves, see

Table 1. The Layered Evolution of Cognition Through Autopoietic Agency.

Level	Key Processes	Form of Agency	Mode of Cognition	Form of Selfhood
Active Matter	Self-organization, dissipation, reaction–diffusion	Proto agency (physical constraints)	Pattern formation, energy flow	Emergent pattern dynamics
Cellular	Metabolism, membrane regulation, chemotaxis	Basic self-maintenance	Environmental responsiveness	Metabolic self
Non-Neural Multicellular	Morphogenesis, hormonal signaling, bioelectric gradients	Distributed structural regulation	Growth-based plasticity	Embodied form-in-context
Sensorimotor	Neural feedback, perception–action loops	Locomotion, reflexive adaptation	Perception through doing	Mobile, embodied self
Symbolic/ Reflective	Language, abstraction, memory integration	Narrative, imaginative, intentional	Simulation, reflection, conceptual modeling	Narrative self/ autobiographical identity
Social/Cultural	Communication, norms, institutions, shared memory	Distributed agency (roles, systems)	Intersubjective meaning-making	Participatory, relational self

.

9. Info-Computational Framework: Nature as a Network of Morphological Computation Leading from Active Matter to Complex Intelligences

The basic idea of this approach to understanding the evolution of intelligence is based on computational naturalism, a naturalist framework that conceptualizes the universe as a network of networks of informational structures where dynamical changes (information processes) are understood as natural computation. It is important to point out that the Turing Machine model of computation used to describe classical computers is only a subset of variety of models of computation found in nature, as argued by Dodig-Crnkovic [14]. From this perspective, the physical universe computes its own next state by implementing physical/chemical/biological laws. Whatever we observe in sciences can be expressed in terms of information and natural computation. Cognition and intelligence are not restricted to human minds or neural systems but emerge across levels of organization from active matter at the most basic level and are present wherever there are agents that process information through their structure and dynamics.

As mentioned before, information in this context is defined relationally—as differences that make a difference for the system [10]. Computation is described in a broad morphological sense: as physical state transitions guided by internal and external constraints. This aligns with the view that all material agents—cells, tissues, organisms, or collectives—engage in natural computation processes to regulate, adapt, and sustain themselves.

Info-computationalism [18] enables a unified account of cognition, in which self-organizing matter becomes informationally sensitive, autopoietic systems develop regulatory loops, and symbolic agents gain the ability to manipulate abstract representations. Each step in the evolution of intelligence corresponds to a deepening of computational capacity and informational integration. This perspective also bridges diverse domains—physics, biology, cognitive science, and artificial intelligence—by treating agency and knowledge as emergent properties of information-processing systems embedded in the fabric of nature.

10. Conclusions

This paper outlines a layered, agent-based framework for understanding the emergence of cognition and intelligence from physical, chemical, and biological processes. Each level of organization—from molecular agents to symbolic and collective selves—demonstrates a form of autopoietic agency, grounded in the system’s ability to process meaningful distinctions.

Unlike traditional approaches that treat cognition as computation in the form of symbolic manipulation alone, we emphasize the roots of cognition in self-regulation, environmental coupling, and recursive embodied information processing. From the self-assembling and self-organizing properties of active matter, to basic sensitivity in cells, to reflective consciousness and cultural knowledge, cognition is shown to be an emergent property of systems that distinguish, regulate, and maintain their identity across time and scale. This approach synthesizes insights from autopoiesis, systems theory, enactivism, cybernetics and computational modeling, and offers a coherent framework for investigating the natural evolution of cognition and intelligence, including applications in artificial intelligence and synthetic biology.

By recognizing cognition as a process intrinsic to living organization, with its roots in the properties of active matter, this model reframes intelligence not as a separate function, but as a layered property of adaptive systems—emerging from, and continuous with, life itself.