Biological Computationalism: A Third Path to Explain Consciousness Beyond Code

Can consciousness be reduced to abstract algorithms, or does it require specific biological processes? Borjan Milinkovic and colleagues from the Estonian Research Council propose a third framework in Neuroscience & Biobehavioral Reviews (2026) called biological computationalism. This approach argues that consciousness arises from computation uniquely realized in biological systems, characterized by hybrid discrete-continuous dynamics, scale-inseparability, and metabolic grounding.

The Two Traditional Positions

Consciousness research has historically oscillated between two opposing views:

Computational Functionalism treats consciousness as abstract information processing. According to this view, mental states are defined by their functional roles rather than physical implementation. If a system implements the right computational structure, it possesses consciousness regardless of substrate. A silicon computer running the appropriate algorithm would be conscious just as a biological brain is.

Biological Naturalism insists consciousness is inseparable from specific brain biology. Biological processes, particularly those involving organic chemistry and neural architecture, produce consciousness in ways that cannot be replicated through different physical substrates. Consciousness is not multiply realizable. Only biological brains, or systems closely resembling them, can be conscious.

These positions create an impasse. Functionalism struggles to explain why implementation should matter at all, potentially implying consciousness in systems most would consider obviously unconscious. Naturalism struggles to explain which biological features are essential and why, risking arbitrary boundaries around particular chemical or structural properties.

Biological Computationalism: The Third Framework

Milinkovic and colleagues argue that “the traditional computational paradigm is broken or at least badly mismatched to how real brains operate.” Brains do not function like von Neumann architecture computers. They do not execute abstract algorithms on passive data structures. Instead, brains instantiate computation through physical dynamics rather than implementing abstract procedures.

This distinction is fundamental. Contemporary AI systems largely simulate functions through digital procedures. Brains instead instantiate computation in physical time through continuous fields and emergent electromagnetic interactions. These are not implementation details separate from computation but the computational mechanisms themselves.

Biological computationalism preserves the insight that consciousness involves computation while recognizing that the type of computation matters. Not all computing substrates are equivalent. Consciousness requires what Milinkovic calls “the right kind of computing matter.”

Three Defining Properties of Biological Computation

Milinkovic identifies three essential characteristics that distinguish biological from conventional digital computation:

1. Hybrid Discrete-Continuous Dynamics

Biological systems combine discrete events with continuous processes. Neurons fire in discrete spikes. Neurotransmitters release in discrete packets. These are digital-like events.

However, these discrete events occur within continuous dynamics: voltage fields, chemical gradients, ion diffusion, membrane polarization. The brain is neither purely digital nor merely analog but operates through hybrid mechanisms where discrete and continuous processes interact continuously.

Conventional computers maintain strict separation between discrete logic operations and continuous physical processes. The logic is designed to be independent of analog variations in voltage or temperature. Biological systems exploit rather than suppress these continuous dynamics as part of computation itself.

2. Scale-Inseparability

Unlike conventional computers with clear software-hardware boundaries, brains lack tidy separation between functional and implementation levels. Multiple scales interact simultaneously, from ion channels to synaptic networks to whole-brain dynamics.

Changing implementation changes the computation. Altering neurotransmitter kinetics changes what can be computed, not just how fast. Modifying dendritic morphology changes network capabilities, not just efficiency. There is no software that could be preserved while swapping out the hardware.

This property challenges computational functionalism’s core claim. If consciousness depends on computational structure independent of implementation, then scale-inseparability should not matter. But if consciousness emerges from computation where implementation and function are inseparable, then replicating consciousness requires replicating the relevant physical dynamics, not just formal structure.

3. Metabolic Grounding

Energy constraints fundamentally shape brain organization. The brain represents approximately 2% of body mass but consumes 20% of metabolic energy. These constraints determine what the brain can represent, how it learns, and patterns of information flow.

Metabolic limitations enforce sparsity in neural representations. Most neurons remain silent most of the time. Active neurons operate near energy-efficient regimes. Network architecture reflects energy optimization, with hub-and-spoke organization minimizing costly long-range connections.

Conventional computers face energy constraints as well, but these constraints operate as external limits on performance rather than fundamental principles shaping computational organization. Computer architectures are designed according to logical principles, then energy consumption is minimized within that framework. Brain architecture is shaped by energy constraints from the start.

Implications for Artificial Consciousness

Biological computationalism has direct implications for artificial consciousness research. If consciousness requires hybrid dynamics, scale-inseparability, and metabolic grounding, then standard digital computers running software are not sufficient, regardless of algorithmic sophistication.

This does not imply that only biological neurons can support consciousness. Milinkovic’s framework allows for artificial consciousness but requires building systems where computing is not layered as software-on-hardware. Instead, synthetic conscious systems would need to be:

Physically Instantiated: Computation occurs through physical dynamics rather than abstract symbol manipulation. The medium matters because the medium is part of the computation.

Dynamically Coupled: Multiple scales interact without clean separation. Implementation and function are intertwined rather than independent.

Energetically Constrained: Metabolic limits shape computational organization from the beginning rather than being added as external constraints.

Current neuromorphic computing efforts move in this direction. Systems using analog circuits, spiking neurons, and local learning rules begin to capture some biological computation properties. However, most neuromorphic systems still implement algorithms rather than instantiating physical processes that are computational.

Distinguishing Biological Computationalism From Related Views

Biological computationalism differs from several related positions in consciousness research:

Versus Panpsychism: Panpsychism attributes consciousness to all matter or information-processing systems. Biological computationalism specifies particular computational properties necessary for consciousness, not present in all physical systems.

Versus Strong Biological Naturalism: While both emphasize biology’s role, biological computationalism focuses on computational properties instantiated in biology rather than organic chemistry per se. In principle, different physical substrates could instantiate the same computational properties.

Versus Standard Computationalism: Traditional computational theories treat implementation as irrelevant. Biological computationalism argues that implementation matters when it determines computational properties like hybrid dynamics and scale-inseparability.

Versus Embodied Cognition: Embodied approaches emphasize sensorimotor interaction with environments. Biological computationalism focuses on internal computational mechanisms rather than external coupling, though both can be complementary.

Open Questions and Research Directions

Milinkovic’s framework raises several empirical and theoretical questions:

Which Properties Are Essential?: Are all three properties (hybrid dynamics, scale-inseparability, metabolic grounding) necessary for consciousness, or would some subset suffice? Can systems with only some properties exhibit partial consciousness?

Measurement and Testing: How can researchers determine whether artificial systems genuinely instantiate biological computation rather than simulating it? What empirical tests distinguish instantiation from simulation?

Evolutionary Origins: Why did consciousness emerge through biological computation specifically? Are there evolutionary pressures that favor consciousness in hybrid, scale-inseparable, metabolically constrained systems?

Degrees of Consciousness: Does biological computationalism predict consciousness as binary (present or absent) or as admitting degrees? If systems can instantiate biological computation properties to varying extents, does consciousness vary accordingly?

Engineering Paths: What specific architectures would instantiate biological computation in artificial systems? Are analog circuits sufficient, or do systems need chemical dynamics, electromagnetic fields, or other biological features?

Reframing the Debate

Biological computationalism offers a resolution to the long-standing debate between functionalism and naturalism. Consciousness is computational, validating functionalist intuitions that mental states are multiply realizable. However, consciousness requires specific computational properties that happen to be realized in biology, validating naturalist intuitions that implementation matters.

The key insight is that not all implementations realize the same computational properties. Digital computers implement different computational processes than biological brains, even when simulating similar input-output functions. The differences matter for consciousness.

This framework shifts research focus from abstract algorithms to physical computational mechanisms. Understanding consciousness requires investigating how hybrid dynamics, scale-inseparability, and metabolic constraints enable specific forms of information processing. Replicating consciousness in artificial systems requires instantiating these properties rather than simulating their effects.

As research progresses, biological computationalism provides testable predictions distinguishing it from competing frameworks. Systems with biological computation properties should exhibit consciousness-related phenomena. Systems lacking these properties should not, regardless of behavioral sophistication. The question moves from whether consciousness can be computational to what types of computation suffice.

For detailed analysis of biological computationalism and its implications, see the full paper summary at Phys.org. Related discussion of consciousness detection methods explores how these theoretical frameworks inform empirical research.