Can the theory be applied across different types of systems, including artificial systems?

IIT's central identity claim ties consciousness to Phi-structure — a maximally irreducible cause-effect structure — defined purely in terms of physical cause-effect power, not in terms of neurons, biochemistry, or specific neuroanatomy. The theory explicitly holds that any physical system, "silicon included," could be conscious if it forms a genuine local maximum of integrated information with the right recurrent, bidirectional cause-effect structure. This is substrate-NEUTRAL with respect to biology, which is precisely what the yes-bar requires: a substrate-independent (in the relevant biological sense) formulation that could in principle apply to artifacts. Tononi and Koch's "here, there and everywhere?" framing and the neuromorphic-hardware discussion show the proponents apply the theory to artifacts directly.

Key evidence: Tononi & Koch 2015 ('Consciousness: here, there and everywhere?') and IIT 4.0 (Albantakis et al. 2023): Phi is defined over any physical substrate's intrinsic cause-effect structure, and IIT explicitly allows that artifacts (e.g., neuromorphic hardware with strongly integrated, re-entrant connectivity) could be genuinely conscious.

Question 7 asks only whether the theory is substrate-independent and could in principle apply to artifacts; the yes-bar fails a theory only if it is framed solely in particular biological neural mechanisms with no substrate-independent formulation. PCT is the opposite case: it is a control-theoretic, functional framework whose core mechanisms (control of perception, the level hierarchy, the reorganizing system) are explicitly machine-realizable and are routinely implemented as digital simulations and in physical robots, with the thermostat/guided-missile artifact as its canonical illustration. It contains no essential dependence on brain matter or specific neuroanatomy. Even the consciousness extension (Mansell) ties consciousness to reorganization-type processes that are themselves implementable, implying no biological substrate is required in principle. The unresolved point is only whether an artificial PCT system would actually be phenomenally conscious, which is a separate hard-problem question, not the substrate-independence question Question 7 poses.

Key evidence: PCT is implemented in digital computer simulations (often >95% fit to tracking data) and in robots (rovers, balancing robots, robot arms), with the thermostat/guided missile as its canonical illustration — establishing that the theory's mechanism is explicitly artifact-realizable with no biological substrate requirement (Powers 1973; IAPCT; Mansell 2022).

HOT is formulated by its proponents as a reductive representational/functionalist theory: consciousness consists in a first-order state being targeted by a suitable higher-order representation, a metacognitive relation defined functionally rather than in terms of biological tissue. This makes consciousness multiply realizable and substrate-independent by construction — what matters is implementing the right higher-order representational architecture, not neurons. The theory therefore explicitly licenses machine/artifact consciousness for any system that genuinely forms suitable HOTs about its own first-order states. Lau and Michel's caveat that 'Swiss cheese cannot implement the relevant computations' is a constraint on the right functional/computational organization, not a reversion to biological essentialism, and so is fully consistent with substrate independence. Crucially, Q7's bar asks only for in-principle applicability to artifacts via a substrate-independent formulation, which HOT meets directly from its core commitments.

Key evidence: The theory's own statement that HOT is a representational/functionalist account making consciousness "multiply realizable and substrate-independent—what matters is implementing the right higher-order representational architecture, not biological tissue," explicitly licensing machine/artifact consciousness (Rosenthal 2005; Lau & Rosenthal 2011; Michel & Lau 2021).

AST defines awareness as the content of an information-processing model (the attention schema) — a self-model S, a modeled attention relation A, and a represented stimulus V — none of which is tied to biological neurons or specific neuroanatomy. Graziano explicitly frames AST as substrate-independent: his 2017 paper is titled "The Attention Schema Theory: A Foundation for Engineering Artificial Consciousness," arguing any system that builds a rich model of its own attention would, when reporting, claim awareness and "cannot tell the difference." The 2021 PNAS neural-network agent is a concrete artificial instantiation showing an attention schema implemented outside biology. The Q7 yes-bar requires only substrate-independence and in-principle applicability to artifacts, not a solution to the hard problem, so AST clears it.

Key evidence: Graziano, M. S. A. (2017). The Attention Schema Theory: A Foundation for Engineering Artificial Consciousness. Frontiers in Robotics and AI, 4:60 — explicitly presents AST as substrate-independent and applicable to machines, supported by the artificial attention-schema agent in Wilterson & Graziano (2021, PNAS).

The yes-bar requires substrate-independence and in-principle applicability to artifacts, with no essential dependence on biological brain matter or specific neuroanatomy. Irruption Theory satisfies this: its core formal dynamics (constraint-neutralization, self-optimization, entropy/underdetermination bursts) are developed and tested in artificial-life and ANN models, and the gating requirement is autonomy/precariousness/intrinsic motivation grounded in "autopoiesis OR EQUIVALENT" — an organizational, multiply-realizable condition, not a carbon-vs-silicon one. Froese's own stance is explicitly that an artifact would qualify if it were genuinely autonomous and self-maintaining, which is a functional criterion an engineered system could in principle meet. The autonomy requirement narrows WHICH artifacts qualify but does not reintroduce biological substrate-dependence, so the theory is substrate-flexible rather than substrate-bound.

Key evidence: Froese's irruption/self-optimization dynamics are replicated across multiple ANN architectures and agent-based simulations (Froese 2023, Entropy 25:748), and the qualifying condition is stated as autopoiesis "or equivalent" autonomy/precariousness rather than biological neural matter — implying a genuinely autonomous artifact could exhibit true motivated irruptions.

Orch OR identifies consciousness with a substrate-neutral physical event — orchestrated objective reduction crossing the gravitational self-collapse threshold τ ≈ ℏ/E_G — not with biological tissue per se. Penrose and Hameroff explicitly treat microtubules as the means by which brains happen to realize this process, "not a logical necessity," and explicitly state that a non-biological system physically realizing orchestrated OR (e.g. an appropriately engineered quantum device) could in principle be conscious. The criterion for consciousness is therefore stated in physics (quantum-state reduction under orchestration), which is medium-independent in principle, satisfying the substrate-independence yes-bar. The theory's rejection of CLASSICAL/algorithmic computers is not a failure of universality — it is a substantive prediction derived from the non-computability commitment, and it leaves a clear door open for genuinely quantum OR-implementing artifacts. The bar asks whether the theory could in principle apply to artifacts, not whether all artifact types qualify, and Orch OR's own proponents affirm the in-principle case.

Key evidence: Hameroff & Penrose (2014) and the theory's stated commitment that biology/microtubules is the substrate that achieves OR in brains "not a logical necessity," with consciousness identified with the substrate-neutral OR collapse meeting the ℏ/E_G threshold — implying an engineered quantum device realizing orchestrated OR could in principle be conscious.

Question 7 asks only whether the theory is substrate-independent and could in principle apply to artifacts — it does not require explaining phenomenal consciousness (that burden falls on other questions). PP/FEP is explicitly formulated in substrate-neutral terms: the Free Energy Principle applies to ANY self-organizing system with a Markov blanket separating internal from external states, making no essential appeal to biological neurons or specific neuroanatomy. Active inference is routinely and concretely implemented in software agents and robots, demonstrating that the functional organization transfers to non-biological substrates as a matter of standing practice, not promissory note. The yes-bar is precisely met: there is a genuine substrate-independent formulation. The caveat that the framework does not specify which inferential systems are phenomenally conscious goes to other questions (Q1/Q2-style what/why), not to universality as written.

Key evidence: Friston 2010 / Parr, Pezzulo & Friston 2022 (Active Inference, MIT Press) formulate the FEP for any system with a Markov blanket and implement active inference in robots and software agents — a substrate-independent formulation already realized in artifacts.

GWT is constitutively functionalist and architecture-based: consciousness is identified with a system's mechanism for making information globally available via competition, a limited-capacity workspace, and global broadcast — a specification stated in terms of information flow, not biological tissue. The theory was born from AI "blackboard" architectures and has been given concrete substrate-independent implementations (Baars & Franklin's LIDA software agents; Blum & Blum's Conscious Turing Machine), and Dehaene explicitly allows that machines could in principle meet GNW criteria. The yes-bar requires only substrate-independence and in-principle applicability to artifacts, not a solution to the hard problem. The acknowledged phenomenal gap (it delivers access/functional consciousness, leaving felt experience open) is the q1 issue and does not bear on universality. The Global Neuronal Workspace neural realization is one substrate-specific instantiation, but the underlying architectural claim is explicitly multiply realizable.

Key evidence: GWT has full computational, non-biological implementations claiming functional consciousness: Baars & Franklin (2009) LIDA model and Blum & Blum (2022) Conscious Turing Machine (PNAS), a deliberately substrate-independent formalization.

Under the strict "as actually formulated" standard, NPS does not provide a substrate-independent formulation that its proponents extend to artifacts. Lyre's NPS grounds phenomenal structure in "self-organized neural maps" and biological difference/change-detection coding, offering no account of artificial substrates; Northoff's allied Spatiotemporal Neuroscience is more explicitly substrate-bound, stressing embodiment, interoceptive/visceral anchoring, and biologically specific architecture (thalamus, insula, cortical hierarchy). Although the homomorphism/structuralist framing is in principle multiply realizable and a structuralist could extend it to any system instantiating the right Q-homomorphic structure, the authors do not make this move — making a YES a promissory note rather than a stated commitment. The yes-bar requires an existing substrate-independent formulation applicable to artifacts; what the theory actually offers is silence or implicit substrate-dependence.

Key evidence: The theory's own characterization states both strands are "essentially silent or implicitly substrate-bound, and neither endorses strong substrate-independence," with Lyre's NPS "tightly tied to 'self-organized neural maps'" offering "no account of artificial substrates" and Northoff's program "more explicitly substrate-dependent."

DIT's mechanism is formulated entirely in terms of concrete mammalian biophysical structures — layer 5 pyramidal apical/somatic compartments, apical calcium plateau potentials, metabotropic-receptor-mediated gating, and non-specific thalamic nuclei. The proponents (Bachmann, Suzuki, Aru; Aru, Suzuki, Larkum) offer no substrate-independent or functional-level restatement of this mechanism that could be realized in non-biological media, and they are largely silent on artifacts/machine consciousness in their primary papers. The theory in fact leans toward a substrate-dependent reading and has been read as implying current feedforward AI lacks the essential cellular integration mechanism, which cuts against universality rather than supporting it. Under the strict bar, a theory framed solely around particular biological neural mechanisms with no substrate-independent formulation is NO.

Key evidence: DIT specifies consciousness as implemented by apical-somatic coupling in L5 pyramidal neurons gated by non-specific thalamus (Aru, Suzuki & Larkum 2020, Trends in Cognitive Sciences) — a claim stated purely in terms of specific cellular biophysics with no abstraction to substrate-independent function.

Under the strict bar, YES requires a substrate-independent formulation that could in principle apply to artifacts. RPT does not provide this cleanly. Lamme's decisive 2018 "missing ingredient" — the move that converts mere recurrence into experience — is recurrent-processing-induced network plasticity grounded in specific biological mechanisms (NMDA-receptor- and calcium-dependent Hebbian synaptic change). This pins the theory's actual explanatory commitment to particular neurobiological machinery rather than to an abstract functional dynamic, and would exclude standard non-plastic artificial networks. The theory offers no explicit substrate-independent criterion or pathway for artifacts; Lamme neither endorses nor rules out machine consciousness, leaving universality genuinely unresolved — and under genuine uncertainty the strict standard defaults to NO.

Key evidence: Lamme (2018, Phil. Trans. R. Soc. B 373:20170344) names the "missing ingredient" as recurrent-processing-induced network plasticity specified via NMDA/calcium-dependent Hebbian synaptic change — a biologically specific mechanism, with no abstract/functional substrate-independent restatement.