A Holistic and Ecological Understanding of Computation

Carlos E. Perez
Intuition Machine
Published in
4 min readAug 16, 2023


Midjourney Generated


  • Computation is not an inherent category or natural kind, but arises from complex processes and interactions at multiple levels. It should be viewed as dynamically emergent rather than in static, formal terms.
  • Computational systems are fundamentally embedded in and sustained by continual engagement with external environments, not isolated.
  • The computational is inseparable from the non-computational
  • cognition involves complex interactions between internal and external processes.


  • Computation must be understood empirically and situated, by studying “computation in the wild” rather than formal abstractions.
  • The theory should focus on metaphysics and ontology first, not semantics. Meaning emerges from underlying processes.
  • Computationalism requires grounding in an empirically adequate, dynamic theory of computation, not formal assumptions.


  • Internal representations acquire meaning through interaction and participation with external processes. Meaning is relational.
  • Complex blends of implicit and explicit representation defy classical symbolic/semantic models.


  • Computation unfolds dynamically over time through complex sequences of events, not as static symbolic manipulation.
  • Behavior emerges from underlying processes in context, not algorithms alone.

Overall, a process philosophy view sees computation as contingently emerging from open-ended temporal evolution of systemic interactions, rather than as formal rule-based manipulation of symbols. Meaning and cognition arise from precisely coordinated processes crossing brain, body and world. Computationalism needs reconceiving accordingly.

This computational framing can be analyzed using Tinbergen’s 4 questions.

Mechanism (How does it work?)

  • The brain does not inherently compute via formal rule-based manipulation of abstract symbols. Rather, cognition arises from dynamic neural processes that cannot be separated from broader embodied interactions situating brain, body, and environment as an integrated system.
  • Cognition relies on complex coordination of multiple implicit and explicit representations across internal and external processes. Internal representations do not have inherent meaning, but acquire intentionality relationally by interacting with external environments.
  • Computation unfolds as temporal sequences of events, not timeless transformations of static symbols. Behavior emerges dynamically from context-sensitive processes, not algorithms alone.

Ontogeny (How does it develop over a lifetime?)

  • Infants are not born with innate computational capacities for symbol manipulation. Complex cognition gradually emerges through years of embodied sensorimotor engagement situated in social and physical environments.
  • Regularities in computational practice reflect skills and efficiencies gained through repeated interactions and efforts to achieve goals, not pre-specified procedural algorithms. Development involves active exploration and enculturation.
  • Computational artifacts such as language alter trajectories of cognitive development by scaffolding new modes of representation and interaction that transform capacities.

Function (How does it contribute to survival/reproduction?)

  • Computation provides adaptive leverage by augmenting limited biological capacities with environmental and cultural resources in context-dependent, pragmatic ways.
  • Statistical learning across embodied interactions allows computational systems to pick up on useful regularities in perceptions and behaviors, enabling flexible navigation of environments.
  • Computational artifacts enhance evolutionary fitness by linking brains to external media that increase cognitive power, improve social coordination, and accumulate knowledge cross-generationally.

Phylogeny (How did it evolve?)

  • Computational cognition did not evolve in a single step, but emerged gradually over evolutionary time through a long series of adaptations for flexible, context-sensitive interaction with environments.
  • Language and culture enabled new evolutionary modes of computational augmentation by allowing brains to interact with external symbols and artifacts that amplified cognitive powers.
  • The computational artifacts and practices persisting today reflect bio-cultural heritage of retained efficiencies accumulated across generations. They are dynamic evolutionary processes, not fixed algorithms. This expanded Tinbergian synthesis provides a process perspective on how computational cognition arises from embodied experience situated in rich biological, cultural, and environmental contexts over multiple timescales, rather than static formal symbol manipulation.

We can further understand this using Marr’s 3 levels.

Computational level

  • The goal of cognition is adaptive success through flexible, context-sensitive interaction with complex environments.
  • Computational cognitive capacities emerge from the pragmatic demands of embodied action, not innate symbol manipulation algorithms.
  • Cognition involves picking up on statistical regularities in the flow of sensorimotor engagement guided by bio-cultural values.

Representational level

  • Representations are implicit patterns of coordination across brain, body and world, not explicit formal symbols.
  • Internal representations acquire meaning relationally through dynamic participation with external processes.
  • Computation unfolds as temporal sequences of events, not static transformations on discrete symbols.

Implementation level

  • Neural processes cannot be separated from embodied interaction — the brain dynamically couples internal and external resources.
  • Regularities reflect skills developed through repeated interactions situating brain, body and environment as an integrated system.
  • Computational artifacts transform trajectories of development by scaffolding new modes of representing and interacting.

In summary, a Marr-style analysis suggests computation should be viewed as an adaptive bio-cultural process emerging from embodied experience rather than rule-governed symbol manipulation. Cognition is the skillful picking up of pragmatic regularities through dynamic participation of brain, body and world. Computation is not an inherent natural kind but a cultivation of representational skills and interactive competencies transforming across multiple timescales.

We can contrast this new synthesis of computation with existing ones: The synthesis views computation as emerging from embodied, situated experience rather than as formal symbol manipulation.

Newell & Simon’s physical symbol system hypothesis claims cognition is the manipulation of discrete symbolic representations. But the synthesis sees representations as implicit patterns of temporal coordination, with meaning arising relationally, not from predefined symbols.

Fodor’s view of thinking as formal symbol manipulation assumes rule-governed algorithms operating on explicit discrete symbols. But the synthesis argues computation unfolds as flexible event sequences, cultivated skills, not formal transformations.

Dreyfus’ characterization of requiring explicit symbol manipulation overlooks the importance of implicit representations coordinating brain, body, and world. The synthesis emphasizes these implicit, embodied representations.

Van Gelder critiques computationalism as needing replacement by dynamical systems theory. But the synthesis incorporates key dynamical notions like temporal unfolding, embeddedness, and emergence within a computational framework.

Overall, the synthesized view sees computation as cultivated, embodied skills situated in pragmatic contexts, not inherent formal symbol manipulation. It stresses temporal unfolding, implicit representation, and interactive dynamics missing from other formalized conceptions. Computation emerges from, rather than governs, real-time embodied experience.