Eyes evolved to transduce photons. Ears transduce pressure waves. The vestibular system tells you which way is down. Each sensory faculty connects an organism to a specific domain of physical reality that matters for survival.
Consciousness is a faculty in the same sense. It evolved to navigate the landscape of what could happen next [A].
Not "navigate" as a metaphor for thinking. Navigate the way vision navigates the electromagnetic spectrum: as a biological process that transduces a domain of physical reality into actionable information. Vision is the transduction of photons. Consciousness is the navigation of possibility space — the landscape of available futures.
This is distinct from Hoffman's interface theory, which treats consciousness as a "desktop," a simplified representation of reality that hides its true complexity. Hoffman's consciousness is passive and representational. The navigator is active and selective. It doesn't perceive a simplified version of what could be. It selects from it, steering the organism through branching points that determine which futures get realized.
"Consciousness navigates possibility space" is shorthand for four operationally distinct components:
Generation: the production of genuine physical indeterminacy from which selection can occur. This is what the quantum substrate provides at Level B. Without generation, there is nothing to navigate; the system is deterministic and follows a single predetermined path. At Level A, generation can be understood more broadly as the production of multiple candidate trajectories, whether through quantum indeterminacy or classical stochastic processes.
Selection: the collapse of possibilities into definite outcomes. This is the act of navigation itself, the steering. Each selection determines which branch of possibility space the organism follows.
Integration: the binding of the selected outcome into a unified, irreducible state across the organism's neural architecture. This is IIT's Φ requirement. Without integration, selections at different sites would be incoherent, producing fragmentation rather than navigation. A vehicle with two steering wheels pulling in different directions does not navigate. It decoheres. There is something satisfying about Phillips and Tsuchiya's result here: they showed that all of IIT's axioms derive from a single mathematical structure (the universal mapping property). The requirement for integration is not an arbitrary postulate. It follows from what it means to select a single path from many.
Propagation: the broadcasting of the integrated selection globally for coordinated action. This is GNWT's broadcasting mechanism. Without propagation, navigation would be private and functionally inert. The organism must act on its selections, and acting requires that all subsystems (motor planning, memory encoding, emotional evaluation, linguistic processing) receive the navigational output simultaneously.
The first two components, generation and selection, are NFT's distinctive contribution. The latter two are inherited from IIT and GNWT respectively. IIT describes what the vehicle must be. GNWT describes how it communicates internally. Generation and selection describe what it is for.
If you know Friston's active inference framework, you will recognize these operations wearing different clothes. Generation is stochastic sampling in a generative model. Selection is policy selection under expected free energy. Integration is IIT's Φ requirement. Propagation is GNWT's broadcasting. At Level A, the four-operations framework organizes these established components under a common architecture. At Level B, it transforms them: generation becomes physical indeterminacy via radical pair spin states, and selection becomes quantum measurement with structured back-action. Level A unifies. Level B transforms.
The four operations are abstract. Here is what they look like in practice.
You are standing at the edge of a busy street. Traffic is moving. You need to cross.
Generation. Your brain produces candidate futures. Cross now, in the gap between the taxi and the bus. Wait two seconds for a wider gap. Walk thirty meters to the crosswalk. Step back from the curb entirely. These are not sequential deliberations but a field of simultaneously available trajectories, a landscape of what could happen next. At Level A, this is stochastic sampling from a generative model. Level B adds something stranger: genuine quantum indeterminacy in the radical pair substrate, contributing physical alternatives that no classical sampling process can replicate.
Selection. You step off the curb into the gap. In that instant, one trajectory is selected and the others collapse. You are now committed to a path that forecloses the crosswalk option, the waiting option, the stepping-back option. This is what navigation is, the irreversible commitment to one future from among the genuine alternatives. The felt sense of deciding, the phenomenological weight of the moment when you go, is what selection feels like from the inside.
Integration. The decision to step is not fragmented. Your visual system's estimate of the taxi's speed. Your motor cortex's assessment of your stride length. Your amygdala's spike of urgency. Your memory of a near-miss last month. All of it, bound into a single, unified navigational state. If they were unbound, if your legs moved while your fear screamed stop, you would not navigate. You would fragment. This binding is IIT's integration, and without it the other three operations produce incoherence rather than agency.
Propagation. The integrated decision broadcasts everywhere, simultaneously. Motor cortex fires the stepping sequence. Hippocampus encodes the spatial context. Prefrontal cortex updates the plan for the next thirty seconds. The autonomic system elevates heart rate. Every subsystem receives the navigational output at once and acts on it. This is GNWT's broadcasting, the mechanism that turns a private selection into coordinated whole-organism action.
The whole thing takes a fraction of a second. It is so fast and so ordinary that you don't notice it happening. But every waking moment is composed of these sequences, nested and overlapping — micro-navigations through the landscape of what could be, each one an act of conscious steering.
Chen and Sanders' Consciousness as Entropy Reduction (CER) model provides an independent computational-level description that aligns with Level A. In the CER framework, "subconsciousness" is a probability distribution over scenarios and "consciousness" is a determinate, zero-entropy selection. It's a single scenario chosen from infinitely many. The transition from subconscious to conscious is modeled as entropy-gradient descent on a weighted general entropy, with noise to break symmetry.
CER cleanly operationalizes what Level A claims: conscious selection collapses a scenario distribution. The language is different (entropy reduction rather than navigation) but the operation is the same. The organism faces a landscape of possibilities, and consciousness reduces it to a committed path.
CER is explicitly classical. It does not require quantum mechanics. And that is exactly the point. Level A is the claim that navigation is what consciousness does, regardless of how it is implemented. The how is Level B's business.
A note on terminology, because the next two chapters use several phrases that all point at the same thing. "Navigation" is the process. "Selection among futures" is the act. "Trajectory entropy reduction" is the measure. "Active inference" is the same faculty described from the information-theoretic perspective. These are four descriptions of one operation, not four different operations. When the language shifts between them in what follows, the referent does not.