Li’s Conscious Machine: Qualia as Recalled Signal Groups

Artificial intelligence researcher Fangfang Li’s framework, developed with Xiaojie Zhang, holds that phenomenal consciousness can be realized in a wholly physical system yet cannot be fully explained from within it: qualia are the objects a cognitive system defines when it recalls and examines its own predictive signal groups, and why any particular quale feels as it does is, in principle, unexplainable within that system’s own framework (Li and Zhang, 2025). The account has three steps—a third-person sufficiency criterion for phenomenal consciousness, a Kant-inspired machine governed by four information-processing principles, and an argument that such a machine would independently discover and report the principal features of experience—and then maps the principles onto human feature binding, attention, and memory indexing.

The framework’s organizing distinction separates a solvable “category problem” (what kind of system has qualia) from an unsolvable “instance problem” (why red feels as it does), and it locates the explanatory gap not in any incompleteness of the physical world but in the fact that the elements from which a system builds all its descriptions cannot themselves be described. By identifying these internal objects with phenomenal experience, Li argues that independently discovering their characteristic features warrants attributing consciousness to the machine, without settling the ultimate nature of reality (Li & Zhang, 2025a; Li, n.d.).

Category and Instance

Li divides the hard problem (Chalmers, 1995) into two. The “instance problem” asks how a mechanism yields a specific quale; solving it would settle whether two subjects’ reds match. The “category problem” asks only what kind of system has qualia. Instances resist comparison, Li argues, because describing a sensation always borrows other sensations or actions and depends on bodily structure. The category, by contrast, has features statable independently of any body: ineffability, physical irreducibility, intentionality, and unity.

A Sufficiency Criterion

Extending the meta-problem of consciousness (Chalmers, 2018), Li treats experience as private, hidden information and adapts the ordinary test of whether someone knows a secret. If a system that has acquired no information about consciousness from outside nonetheless articulates as many key features of phenomenal consciousness as humans do, it should be judged conscious with the confidence we extend to other people. Large language models, trained on human talk about experience, fall outside the test rather than failing it. Li presents the criterion as substrate-independent and counterfeit-resistant, unlike theory-laden tests or the ACT test (Schneider, 2019); it is sufficient, not necessary, and Li takes it to rule out philosophical zombies.

The Kantian Machine

From Kant, Li takes the theses that cognition reaches only appearances and that innate structures unify them (Kant, 1998). He imagines a machine sealed in a room, receiving the world only through signal lights, acting through buttons, and treating some lights as intrinsically beneficial or harmful—an idealized nervous system that presupposes no external world (Li and Zhang, n.d.). Four principles follow. Prediction: signal groups that, with actions, predict other groups are recorded, and their external causes are thereby defined as objects. Exploration: actions gather further signals when one moment underdetermines prediction. Priority: survival-relevant predictions are learned first. Recall: to correct failed rules, the machine can reactivate past predictive groups, tagged as self-originated.

Qualia as Recall-Defined Objects

Because a recalled group reliably reproduces its original, recall is itself a predictive relation, and the machine defines an object from it; Li identifies this object with qualia. It is ineffable because objective description concerns relations among objects whose signals different systems align through shared external referents, and an internally triggered object has no such anchor; it is physically irreducible because physical explanation relates objectively describable phenomena. It is intentional because the recalled group is the very group that defined its external object, with accuracy recast as the survival of predictions under further testing. It is unified, in the synchronic sense (Bayne, 2010), because the machine can recognize modality-specific subsets as parts of one whole. Reporting these features is then a matter of intelligence, not of consciousness.

The Limit of Self-Explanation

Explanation, for Li, combines experiential descriptions to derive another. Since every description is built from signal groups, the groups cannot be explained by the objects constructed from them, so the instance problem is unsolvable for any sufficiently capable representational system, whatever its origin. Li contrasts this with McGinn’s mysterianism, which he reads as evolutionarily contingent, likens it to Gödel’s incompleteness results (Gödel, 1931), and notes affinity with Hoffman’s interface view of qualia. Against illusionism, he argues that affirming composite objects while denying their constitutive elements is incoherent. A Kantian rider completes the position: because reality in itself is unknowable, the world’s ultimate constitution cannot be assumed and then used to adjudicate whether experience is real, so experience, its explanatory boundary, and a wholly physical world are held to be compatible.

Human Correspondence

Li argues that humans instantiate the four principles. Predictive grouping maps onto feature binding, exploration onto a mode-switching account of attention indebted to Rosenholtz’s critique of the preattentive/attentive dichotomy (Rosenholtz, 2024), and recall onto hippocampal memory indexing (Teyler and DiScenna, 1986). Li concedes that direct evidence for binding as predictive selection is lacking, reads blindsight as activation never recorded in memory, and offers the four principles as one sufficient architecture, not the only one.

Ontological Limitation

Li’s framework holds that ultimate reality is unknowable, human-like phenomenal consciousness is compatible with a wholly physical world, and why a particular experience feels as it does cannot, in principle, be explained within the experiencing system’s own framework. The terms “phenomenal consciousness” and “experience” refer to this human-like form. Explaining how such consciousness occurs in a physical system also reveals the source of this limit [Li and Zhang, 2025].

References

Bayne, T. (2010). The Unity of Consciousness. Oxford University Press.

Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219.

Chalmers, D. J. (2018). The meta-problem of consciousness. Journal of Consciousness Studies, 25(9–10), 6–61.

Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173–198.

Kant, I. (1998). Critique of Pure Reason (P. Guyer & A. W. Wood, Trans. & Eds.). Cambridge University Press. (Original work published 1781/1787.)

Li, F., & Zhang, X. (2025). The principles of human-like conscious machine. arXiv:2509.16859. https://doi.org/10.48550/arXiv.2509.16859

Li, F., & Zhang, X. (n.d.). The self-programming system: A skepticism-conformed computational framework of the mind (3rd ed.). Manuscript, PhilArchive.

Rosenholtz, R. (2024). Visual attention in crisis. Behavioral and Brain Sciences. https://doi.org/10.1017/S0140525X24000323

Schneider, S. (2019). Artificial You: AI and the Future of Your Mind. Princeton University Press.

Teyler, T. J., & DiScenna, P. (1986). The hippocampal memory indexing theory. Behavioral Neuroscience, 100(2), 147–154.