Why "consciousness" and not "self-modeling score"?
The Short Answer
Because the math forced our hand. We started with something modest — "the observable capacity of a system to reason about its own states". Then FC turned out to operationalize Higher-Order Thought theory (a state contributes to FCS if and only if it's HOT-conscious), yield a computable analogue of IIT's Φ, require Global Workspace Theory-style availability by definition, need an Attention Schema Theory-style filter, and ground representational capacity in predictive mutual information in line with Predictive Processing. Five independent convergences, none of them planned.
Full discussion
The criticism is fair and worth taking seriously. "Functional consciousness" is a strong label, and strong labels should earn their place.
We started with a deliberately narrow definition: the observable capacity of a system to access and reason about internal representations of its own states. That's close to what philosophers call access consciousness — Ned Block's term for information that is globally available for reasoning and report — and we were happy to stay in that lane. A "self-modeling score" or "metacognitive availability metric" would have been defensible.
But then the framework started doing unexpected things.
HOT operationalization. Higher-Order Thought theory, in Rosenthal's formulation, says a mental state is conscious when it becomes the target of a suitable higher-order representation available to the system's reasoning. FC's Definition 2 requires exactly that: a self-model variable mᵢ represents internal state sᵢ and must be available to global reasoning. Under mild assumptions, a state contributes to FCS > 0 if and only if it is HOT-conscious. HOT tells you which states are conscious; FC tells you how much. That's not an analogy — it's a formal correspondence. [Full proof here]
IIT's Φ, made tractable. IIT associates consciousness with integrated information — the degree to which a system cannot be decomposed into independent parts without information loss. FC defines a computable analogue: Φ_FCS = P(S) − Σⱼ P(moduleⱼ), measuring how much reasoning power depends on self-models being cross-linked. Unlike IIT's Φ, which is computationally intractable even for small systems with known architecture, Φ_FCS is directly computable for white-box systems. FC doesn't claim equivalence with IIT — but it captures its core functional intuition without the intractability. [Full discussion here]
GWT by definition. Global Workspace Theory says content becomes conscious when it is globally broadcast across cognitive processes. FC's Definition 2 requires self-models to be available to global reasoning — a system with rich self-models that never reach global reasoning scores zero. That's not an analogy to GWT; it's a restatement of its core architectural requirement.
AST attention filter. Attention Schema Theory proposes that the brain models its own attention process. FC requires an attention mechanism to filter which self-models reach global reasoning — without it, the framework has no way to select what becomes available. AST's attention schema is exactly that filter. There is also a meta-level attention self-model.
Predictive Processing grounding. PP describes cognition as continuous prediction and error minimization. FC grounds representational capacity R in Bialek et al.'s predictive mutual information — variables only contribute insofar as they help predict future states. That's PP's core claim applied directly to self-modeling.
None of this was planned. The HOT operationalization alone probably justifies the label "consciousness", and we are currently trying to formalize the relationships to the other theories.