FMT vs. Higher-Order Theories (HOT)¶

FMT and HOT share an emphasis on self-modeling, and HOT's higher-order representation corresponds to the extended, recursive rungs of FMT's ladder. HOT makes a higher-order representation of the particular state the condition of experience; FMT makes closure of the self-simulation the condition, and finds experience already at a minimal level without reflection.

Higher-Order Theories (Rosenthal, 2005; Lau & Rosenthal, 2011) propose that a mental state becomes conscious when it is the target of a higher-order representation -- roughly, when the system represents itself as being in that state. A perception of red becomes a conscious perception of red when there is a higher-order thought about that perception. The higher-order representation is normally itself unconscious; introspection is the case where it becomes conscious. The Four-Model Theory shares the emphasis on self-modeling and assigns higher-order representation a different place.

The Shared Insight¶

Both HOT and FMT place self-modeling at the center of consciousness. For HOT, a state without a higher-order representation is not conscious, even if it influences behavior. For FMT, consciousness requires the Explicit Self Model, the system's ongoing model of itself, closed with the explicit world model. At the minimal level the self-model is rudimentary: a phenomenal now, at its most primitive a primordial pain or a nothingness, with no reflection and no awareness that it experiences. No per-state higher-order representation is needed.

Conscious states vary on two axes, richness of content and depth of recursion. FMT's graduated levels climb the second: simply extended consciousness adds first-order self-observation (awareness that one experiences), doubly extended adds metacognition, triply extended adds reflection on the modeling itself. Reportability grows with recursion depth; minimal content, with no recursion above it, cannot be reported. HOT's higher-order representation, and the Lockean idea of an awareness of being aware that Rosenthal formalized, correspond to these extended rungs.

Where HOT Stops¶

HOT explains which states are conscious (those with higher-order representations) and partially addresses the Meta-Problem: consciousness seems mysterious because higher-order representations are imperfect models of first-order states. It does not explain why higher-order representation produces phenomenality.

A thermostat with a second sensor monitoring the first has a "higher-order representation" of room temperature, and nobody concludes it is conscious. HOT theorists reply that the higher-order representation must be of the right kind, a conceptual, assertoric representation in the right functional role. Specifying the "right kind" without circularity is the open task.

HOT also leaves several of the eight requirements unaddressed. It has no account of binding, its boundary-setting is imprecise, and the Hard Problem remains: why does a higher-order representation feel like anything?

What FMT Adds¶

FMT adds three elements that HOT lacks:

The four-model architecture. HOT posits higher-order representations without specifying the minimal architecture required. FMT specifies it: four model kinds along two axes (scope and mode), with the ESM as the locus of self-modeling, higher-order on the extended rungs.
Virtual qualia. FMT explains why self-representation closed on itself is phenomenality through the virtual qualia framework: qualia are constitutive properties of the computational level, and its third principle is an identity, "not a law connecting two things: phenomenality is closure" (self-referential closure). On FMT's account, a self-model is necessary, in rudimentary form at the minimal level; phenomenality is that self-model and the world model closed on each other in the open-ended Class 4 regime, an identity the theory argues for rather than derives.
The free-compute requirement. HOT does not specify what kind of substrate supports consciousness-producing higher-order representation. FMT adds it: the substrate must have Class-4 (universal-computation) capability and actually deploy it, autonomously and open-endedly, for self-modeling. Near-criticality -- the edge of chaos -- is the signature this leaves in the human brain and the quantity measured there. This explains why not every system with higher-order representations is conscious -- the substrate must also meet the computational requirement.

A Concrete Difference¶

Abolish metacognitive read-out (meta-d′/d′ near zero) while sparing first-order sensitivity, and measure experience with a no-report or criterion-fixed design. HOT predicts that experience is lost or altered. FMT predicts that it survives if the loop between self-model and world-model is intact, because metacognitive report is one read-out of the self-model. Report may still fail, since memory transfer can fail with it. HOT proponents need not accept meta-d′ as a measure of the higher-order representation they require, so the experiment must show what a manipulation removes.

Metacognitive disruption alone therefore does not test closure. FMT's closure claim would fail if disrupting the self-modeling loop itself left no-report markers of experience intact. Conversely, a higher-order representation in a system without closure, such as a monitor attached to a simulation it takes no part in, is not conscious on FMT.

Figure¶

graph TB
    subgraph HOT_THEORY["HOT"]
        direction TB
        H1["First-order state<br/><i>e.g., perception of red</i>"]
        H2["Higher-order<br/>representation<br/><i>awareness of perceiving</i>"]
        H3["Consciousness<br/><i>= having the HOT</i>"]
        H1 --> H2 --> H3
        H4["❓ Why does this<br/>feel like anything?"]
        H3 -.-> H4
    end

    subgraph FMT_THEORY["FMT"]
        direction TB
        F1["IWM/ISM<br/><i>substrate-level models</i>"]
        F2["EWM/ESM<br/><i>self-referential closure</i>"]
        F3["Virtual Qualia<br/><i>constitutive of<br/>computational level</i>"]
        F4["Free compute<br/><i>Class-4 capability deployed</i>"]
        F5["Extended rungs<br/><i>recursive self-modeling,<br/>reportability grows</i>"]
        F1 --> F2
        F4 --> F2
        F2 --> F3
        F2 --> F5
    end

    SHARED["Shared: self-modeling<br/>HOT's higher-order state ≈<br/>FMT's extended rungs"]
    SHARED --- H2
    SHARED --- F5

    style HOT_THEORY fill:#1a1a2e,stroke:#e76f51,color:#fff
    style FMT_THEORY fill:#1a1a2e,stroke:#2d6a4f,color:#fff
    style SHARED fill:#2d1b69,stroke:#9b59b6,color:#fff
    style H4 fill:#6a1b2a,stroke:#a4243b,color:#fff
    style F3 fill:#2d6a4f,stroke:#40916c,color:#fff

HOT's higher-order representation corresponds to FMT's extended rungs. On FMT, phenomenality is present earlier, at the minimal level, because it is the closure of a rudimentary self-model with the world model.

Key Takeaway¶

HOT places self-modeling at the center of consciousness. FMT puts HOT's higher-order representation on the extended rungs and holds that minimal consciousness is phenomenal without it, adding the architecture, virtual qualia, and free compute (whose neural signature is near-criticality). The testable split: does experience survive the loss of higher-order representation while closure is intact?