A Candidate Architecture of Consciousness

The Modeler-Schema Theory

Your eyes make several saccades a second as you read, each displacing the retinal image, yet the world appears stable. Perceptual systems integrate information across these movements using multiple mechanisms. Ask whether one of those mechanisms also contributes to experience, and one reaches a tractable part of the consciousness problem — not the whole problem by definition, but a place where an architectural theory can make contact with evidence.

Consciousness research contains several mechanistic programs, including global-workspace, recurrent-processing, predictive, and higher-order approaches. Each must answer a demanding question: what organization supports report, control, and experience, and how could rival accounts be distinguished? A functional architecture can address the first two directly. Identifying them with phenomenality requires a further argument.

Heile’s Modeler-Schema Theory¹ proposes a cybernetic, agent-structured architecture that locates experience in a single hypothesized subsystem — the Modeler-schema — and offers a visual experiment intended to expose part of its mechanism. MST is the principal consciousness proposal examined in this volume; my Agency-Model formulation generalizes it in Mirrors of the Mind. The theory fits Axio’s agency-first stance by treating experience as functionally embedded in a control architecture. That fit is a reason to investigate it, not evidence that it is true. The physics volume deliberately left the phenomenal question open (Life, Consciousness, and Time); this chapter specifies a candidate subsystem, computation, and test.

Three Agents, Three Schemas, One Experiencer

MST decomposes the mind into three functional agents:

• The Modeler constructs and updates the World Model.
• The Controller selects actions, uses language, forms narratives.
• The Targeter integrates bottom-up and top-down attention requests.

Each has a proposed regulatory partner — a schema-agent that monitors and tunes its performance. The design is motivated by the broad cybernetic reading of the Good Regulator Theorem discussed in Control Requires Models; the theorem does not derive this particular neural architecture. MST proposes that evaluating and refining the Modeler requires a process that models the Modeler’s modeling. That is the assigned role of a schema-agent.

Six functional components, then. MST assigns the relevant properties to one.

MST hypothesis: the Modeler-schema is the sole generator of qualia.

It receives the full state of the Concrete World Model — the system’s moment-to-moment sensory representation of the environment; the Focal Target Information stream — the data the Controller uses to track what it is currently attending to; and the sensory-level data needed for pre- and post-saccadic consistency checking. From these it alone constructs the Quale World Model: an internal, non-symbolic representation of what experience “feels like” to it — a representation that never reaches the Controller.

If that asymmetry is correct, it offers an explanation of a puzzle about introspection: the processes that generate a report need not be identical to those that generate its represented content. MST calls this the narrator–experiencer distinction. The Controller receives selected information and evaluative signals rather than the complete state assigned to the Modeler-schema. This predicts partial and reconstructive introspective report; it does not independently prove which subsystem experiences.

The Controller’s Blind Spot

A critical observational constraint separates MST from its rivals. Conscious content comes in two distinct forms:

• Focal consciousness: reportable, target-bound, Controller-accessible.
• Diffuse awareness: panoramic, continuous, Controller-inaccessible.

You may have a sense of diffuse awareness across the visual field while being unable to enumerate its detail. Report is selective, peripheral vision is lower-resolution than introspection often suggests, and attention changes what can be discriminated. The phenomenological distinction is therefore useful, but less decisive than the phrase “unmistakably present” would imply.

MST interprets this as evidence that the machinery supporting diffuse content is not identical to the machinery producing focal report. Introspection motivates that distinction but does not uniquely locate the mechanism in a Modeler-schema; rival architectures can also separate access, attention, and report.

Why Qualia Exist

Return to the saccades. Each one creates an enormous shift in retinal input, yet the world appears stable — the clearest and most experimentally tractable case where a consistency mechanism is required. But the role of qualia is not confined to vision. The Modeler-schema must monitor and refine the entire World Model: sensory content, recalled content, abstract content, appraisable emotions, multisensory integration. The qualia it generates are the internal representational format for detecting mismatches, anomalies, and modeling errors across all of these domains. Saccades merely provide a clean natural experiment for isolating the mechanism.

This calibration role explains empirical patterns that other theories leave as brute facts:

• Perceptual qualia are stable and largely universal — because evolutionary pressure is strongest where immediate sensory accuracy matters.
• Recalled imagery varies widely between individuals — because long-term memory reconstruction faces weaker pressure.
• Abstract inner experience diverges most of all — because abstraction lacks a dedicated comparator and recruits heterogeneous resources.

Qualia, on this view, are identified with the Modeler-schema’s general-purpose comparison medium: abstracted, compressed representations used to detect inconsistencies and guide representational refinement across the cognitive system. The functional comparison states are the proposed instruments; their identity with felt qualities remains the theory’s further claim.

A Rare Thing: A Falsifiable Experiment

Theories of consciousness need risky predictions that distinguish them from rivals. MST proposes an experimental hinge.

During a saccade, change a peripheral object. The change can be permanent — it persists after the saccade lands — or temporary — it occurs only mid-saccade and reverts before landing. The theory predicts: permanent changes should generate bottom-up attention targets; temporary ones should not.

Why? A retained representation can help compare post-saccadic input with what preceded it. If, under isoluminant conditions, observers detect small permanent changes but fail to notice temporary ones, the result would support some retained-comparison mechanism. Establishing that the mechanism is specifically the Modeler-schema, and that its comparison format is phenomenal, would require further discriminating evidence. Differences between perception and mental imagery may offer another test bed, but they have multiple possible explanations.

If temporary changes turn out to be more detectable than permanent ones under the stated conditions, a central prediction of this version of MST fails. That would count against the proposed mechanism, although a broad self-model theory could be revised rather than logically falsified by one result. The value of the proposal is that it exposes a specific claim to evidence.

The Hard Problem, Redesigned

Traditional formulations ask why functional cognition is accompanied by experience. MST proposes a functional identity claim: experience is the Modeler-schema’s act of generating a comparison format and using it over time. This would make phenomenality causally engaged rather than idle. But the identity is the theory’s central philosophical premise, not an empirical result already delivered by naming the computation.

This reframes the Hard Problem as a design question: why does the system need this operation? A self-updating world-model benefits from comparison across discontinuous states, and comparison requires some representational substrate. MST calls its proposed substrate the Quale World Model and identifies its operation with experience. Engineering can motivate the comparator and test aspects of it. It cannot establish merely from functional need that the substrate has phenomenal character or that no rival architecture could perform the work.

The architecture also proposes an explanation for the intuition that experience is ineffable. The Controller, as MST defines it, has no direct access to the Quale World Model; it receives selected model-data and appraisable emotion signals about the Modeler-schema’s evaluations. On that decomposition, the reporting role inherits effects of the proposed experiencing role without accessing its full content. The narrator would then find experience partly indescribable because its reports are built from a lossy channel. This is a prediction of the partition, not independent evidence that the partition exists.

This chapter has stolen its title from Dennett, and the theft is deliberate. Dennett demolished the Cartesian Theater and concluded that once the narrating self’s confabulations are explained, there is nothing left to explain — consciousness explained by being explained away. MST says he stopped one subsystem short: his multiple-drafts narrator is the Controller, and the Controller’s introspective blindness shows nothing about whether some other subsystem experiences. Completing that engagement — keeping what Dennett got right while locating what he could not find — is the work of the next chapter, Beyond Dennett.

What a Theory of Consciousness Owes Us

Set out plainly, MST aims to integrate several virtues: a concrete functional architecture, cybernetic coherence, differentiated representational roles, a separation between content and report, and a prediction that can fail. Those virtues make it a productive research proposal. They do not establish that its assigned comparator is the unique locus of experience, that the proposed representational differences explain all phenomenology, or that robust agency requires phenomenality rather than nonconscious coherence-checking.

The functional organization becomes less mysterious once machinery is specified. The phenomenal identity remains the controversial premise: a system may require comparison states without it following deductively that those states are qualia. MST identifies a candidate subsystem and an experiment that can test one of its predicted functions. It therefore specifies functional correlates and a research program. Whether those correlates exhaust phenomenal consciousness is the question Chapters 8–10 must argue rather than assume.