How Does Awareness Get Here? Emergent and Non-Emergent Accounts

What Is Consciousness ended on a fork, and I want to restate it precisely before going further, because everything below is organized around it: is awareness something the universe builds, under the right conditions, call this the emergent view, or is it something already there, prior to and independent of any brain or body, with those brains and bodies acting as channels for it rather than sources of it, call this the non-emergent view?

Before the theories, one clarification that I think has to be made explicit, because it's easy to blur: this is a genuinely hard problem, not a merely difficult one. Reasoning, memory, language, planning, even something as sophisticated as playing chess or writing an essay, these are, in Chalmers' terms, "easy" problems, not because they're simple to build (they obviously aren't) but because they're the right kind of problem: you can specify the function you want, and in principle track how a system carries it out, mechanism by mechanism, until it's explained. There's no further question left over once you've explained the mechanism. Consciousness isn't like that. You can fully explain every mechanism involved in, say, color processing, the wavelength detection, the neural pathways, the categorization, the verbal report, and still have a coherent further question: why is any of that accompanied by the felt quality of seeing red, rather than happening with no one experiencing it at all? That leftover question, after all the function has been explained, is what makes this hard rather than just difficult. Every theory below is, whether it admits it or not, making a bet about that leftover question: solve it, dissolve it as a confusion, or reorganize the problem so it doesn't have to be answered directly.

A note before diving in: the theories below are the major, most-discussed ones, not an exhaustive list. There are somewhere around two dozen distinct theories currently circulating in the academic literature, and I'll flag a few more of the significant ones I'm not covering at the end of this piece.

The emergent camp

Integrated Information Theory (IIT)

Giulio Tononi's IIT starts from an unusual place: not from the physical system outward, but from the structure of experience itself, inward. It begins with axioms about what any experience is necessarily like, that it exists, that it's structured, that it's specific (this experience and not some other), that it's unified, that it's bounded, and then asks what physical properties a system would need in order to produce something with those properties¹. The answer it arrives at is a measure called phi (Φ): the amount of integrated information a system generates, above and beyond what its separate parts generate on their own. High Φ means high consciousness; a system that's just a loose pile of independent parts, however complex, has low Φ, no matter how much raw information those parts process, because nothing about the whole is more than the sum of the parts.

IIT is unusually bold in its implications: it claims consciousness is, in principle, measurable, and that it's a matter of degree, which invites the "levels" framing I explicitly set aside in What Is Consciousness, though IIT would say that's not a bug, it's the theory's whole point. It's also drawn some of the sharpest criticism of any theory in the field, on exactly the ground I care most about here: critics argue the theory doesn't actually explain why integrated information should feel like anything at all; it just stipulates the identity and builds an elaborate mathematical apparatus on top of the stipulation². There's also a live methodological controversy: a 2023 open letter signed by over 100 researchers labeled IIT "pseudoscience," specifically because some of its proponents suggested no physical evidence could ever falsify certain of the theory's core claims, an inversion of how a scientific theory is supposed to work³. I flag this because it's worth being direct about: not every well-known theory in this space is equally reputable within its own field, and IIT is a good example of a theory that's simultaneously mathematically serious and genuinely contested as science.

Global Workspace Theory (GWT) and Global Neuronal Workspace Theory (GNWT)

Bernard Baars' original theory treats consciousness on the model of a theater: the brain runs a huge number of specialized, unconscious processes in parallel, vision, language, motor control, memory retrieval, and consciousness is what happens when one piece of information wins a competition for the "spotlight" and gets broadcast widely enough that many different processes can use it at once⁴. Stanislas Dehaene's later neuroscientific version, GNWT, gives this a physical location: a network of long-range neurons, concentrated in prefrontal and parietal cortex, that can suddenly "ignite," a rapid, self-sustaining, brain-wide burst of activity, when information crosses the threshold into global availability⁵.

GWT has a genuine appeal that IIT lacks: it comes with testable, falsifiable predictions about brain activity, and those predictions have held up reasonably well experimentally. But notice what it's actually explaining: information availability, the fact that a piece of information becomes usable by many systems at once, for reasoning, report, and action. That's much closer to Ned Block's "access consciousness" than to "phenomenal consciousness," and this is exactly the gap What Is Consciousness flagged. GWT gives a genuinely good account of when information becomes globally accessible. It's much thinner on why global accessibility should be felt from the inside, rather than just functionally broadcast in the dark. In my read, GWT is honestly closer to answering the easy problem extremely well, dressed in language that sometimes suggests it's answering the hard one.

Physicalism: the identity theory

The oldest and most direct emergent claim is also the simplest to state: consciousness just is a physical process, full stop, not correlated with brain activity, not produced by it, but strictly identical to it, the way "lightning" turned out to just be a particular kind of electrical discharge rather than something separate that electrical discharge causes. This is the mind-brain identity theory, also called type physicalism, developed by U.T. Place and J.J.C. Smart in the 1950s: a specific type of mental state, pain, say, just is a specific type of physical brain state, with nothing left over to explain once the physical description is complete⁶.

The identity theory ran into a serious problem almost immediately, raised by Hilary Putnam: if pain is strictly identical to one specific brain state, then anything without that exact brain state, an octopus, an alien, a hypothetical AI, couldn't be in pain, even if it behaved exactly as if it were, in obvious distress, avoiding harm, and so on. This seemed far too restrictive. Putnam's alternative was multiple realizability: the same mental state could be built out of many different physical substrates, so what matters isn't the specific material but the causal role that material plays⁷. That single move is what opened the door to the next theory, which takes substrate-independence much further than Putnam originally intended.

Computationalism: the mind-uploading thesis

This is a more specific and more radical claim than functionalism in general: not just that mental states are defined by causal role rather than physical substrate, but that the causal role in question is computational, that a mind just is a certain pattern of information processing, and that pattern could in principle be run on any substrate that implements the right computations, biological neurons or silicon chips alike. This is usually called the substrate independence thesis, and it's the load-bearing assumption behind mind uploading, whole brain emulation, and related ideas: if substrate independence is true, then a sufficiently detailed computational emulation of a brain wouldn't just simulate a mind, it would be one, complete with genuine conscious experience, running on entirely different hardware than the original⁸.

This is a genuinely bold claim, and its defenders know it. It commits to something identity theory explicitly denies: that the specific physical material doing the computing is irrelevant, and only the pattern matters. The most direct challenge to this is the same one that challenges functionalism generally: John Searle's Chinese Room. A person manually following an English rulebook to produce correct Chinese responses, with no understanding of Chinese anywhere in the process, behaves exactly as a computational theory would predict a "conscious, Chinese-understanding system" to behave, and yet, Searle argues, nobody is home⁹. If Searle is right, running the correct computation, however faithfully, doesn't guarantee anything is experienced while it runs, which would mean an uploaded mind could be a perfect behavioral copy of the original person with nobody inside it. Defenders of computationalism have serious responses to this (the "systems reply" chief among them, arguing that understanding belongs to the whole room-plus-rulebook system rather than the person inside it), but the core worry is exactly the hard problem again, now wearing the costume of a very practical, very consequential question: if someone uploaded their mind tomorrow, would they wake up on the other side, or would something that merely acts like them?

Anil Seth: the predictive brain and "controlled hallucination"

Neuroscientist Anil Seth offers a different emergent account, less concerned with a single sufficient condition for consciousness and more with the specific mechanism generating its contents. His framework, predictive processing, treats the brain as a Bayesian prediction engine: rather than passively receiving sensory data, the brain is constantly generating top-down predictions about the causes of its sensory input, and then updating those predictions against the actual signal to minimize the error between the two. What we experience, his phrase, is a "controlled hallucination": not raw perception of the world as it is, but the brain's best current guess about what's out there, held in check (rather than running free, as in an uncontrolled hallucination, or a dream) by continuous correction from sensory data¹⁰. Seth extends this specifically to selfhood and embodiment with his "beast machine" idea: the experience of being a self is, on his account, itself a set of predictions, chiefly about the state of your own body (interoception), rather than a fixed, separate observer sitting behind the process¹¹.

Seth is explicit that he's agnostic on IIT and skeptical of panpsychism, describing his own view as a "functionally agnostic flavor of physicalism"¹², he thinks integration and information are probably necessary conditions for consciousness, but doubts they're sufficient, and he's openly unconvinced IIT's mathematics can currently be tested. His account gives an unusually rich, falsifiable, empirically productive story about what we experience and why it takes the shape it does, but like GWT, it's a theory of content and mechanism more than a theory of why there's felt experience running underneath any of it at all. It's a very good theory of the "controlled" part. It's less clear it's a theory of the "hallucination" part, the fact that anything is being hallucinated, felt, by anyone.

The non-emergent camp

Panpsychism

Panpsychism takes a genuinely different starting move: rather than explaining how consciousness arises from non-conscious matter, it denies the premise, it holds that consciousness, in some minimal, rudimentary form, is a basic and ubiquitous feature of physical reality, present even at the level of electrons and quarks, not something that switches on only once matter gets sufficiently complex¹³. Philosopher Philip Goff, one of its most prominent contemporary defenders, argues this actually dissolves the hard problem rather than solving it head-on: if experience is already present in the basic constituents of the world, there's no gap to explain between non-experiential matter and experience, because there never was any non-experiential matter to begin with.

The cost panpsychism pays for this move is what's called the combination problem: if individual electrons have some tiny flicker of experience, how do billions of those separate micro-experiences combine into the single, unified experience of being you, reading this sentence? Goff's own proposed solution invokes "phenomenal bonding," some special relation by which micro-subjects fuse into a macro-subject, but this remains one of the most actively disputed and unresolved parts of the theory, and critics reasonably point out that the combination problem might just be the hard problem, relocated one level down rather than actually solved¹⁴.

Advaita Vedanta

Advaita's account, on the reading I care about here, is the strongest and most fully worked-out version of the non-emergent claim I've come across. Its position isn't merely "consciousness is present everywhere in small amounts" (panpsychism's move), it's that consciousness (chit, identified with the ultimate reality, Brahman) is the one thing that is genuinely real and unchanging, and that individual minds don't produce consciousness at all; they reflect it, the way a mirror reflects light without being the light's source.

The classical Advaita school of reflection (pratibimbavada), developed by Padmapada and later Prakashatman, gives this a precise mechanism: pure consciousness (the bimba, or "original") appears as if divided and localized when it's reflected in the antahkarana, the mind-intellect complex, producing what's called chidabhasa, a "semblance of consciousness"¹⁵. This reflected consciousness is what does the actual thinking, remembering, and experiencing in daily life, the individual "I", while the source it reflects, the witnessing consciousness itself, never actually enters the mind, is never divided, and never actually undergoes any of the states it appears to illuminate. The tradition is careful to flag that "reflection" is a metaphor and shouldn't be taken too literally, there's no physical distance between consciousness and its reflection the way there's a real distance between the sun and a mirror, but the structural claim is exact: the mind is the medium through which consciousness appears to operate, not the generator of it¹⁶.

This is a genuinely different kind of theory than anything in the emergent camp: it doesn't propose a physical condition that produces consciousness, because it denies that consciousness is produced by anything physical at all. What a brain (or, on some readings, a sufficiently structured mind-body complex) does, on this account, is provide the conditions for reflection, clarity, structure, the ability to individuate an "I", not the conditions for generation. If this is right, asking "what does a system need in order to generate consciousness" is a category error, on the order of asking what a mirror needs to generate light. The right question becomes: what does a system need in order to reflect it clearly?

I want to flag directly, in the same spirit as What Is Consciousness: I find this account intellectually serious and worth taking on its own terms, and it's the account I'm most personally drawn to. But "worth taking seriously" and "true" are different things, and I don't currently have a way to adjudicate between this and the emergent accounts above other than the same route I flagged last time, the sleep and meditation research on whether awareness is ever actually found without content. That's still the most concrete empirical foothold either side has.

Other major theories I haven't covered here

A 2022 review by Anil Seth and Tim Bayne catalogued something like two dozen distinct theories of consciousness currently active in the literature, and organized the most prominent ones into four families: higher-order theories, global workspace theories, predictive processing theories, and integrated information theory¹⁷. I've covered three of those four families above, plus panpsychism, Advaita, physicalism, and computationalism, but higher-order theories deserve at least a mention: the idea, associated most with David Rosenthal, that a mental state becomes conscious specifically when it's the target of a further, higher-order mental state representing it, roughly, a thought becomes conscious not just by occurring, but by being noticed by another part of the mind. I'm also aware of illusionism (associated with Keith Frankish and, in later work, Daniel Dennett), which takes a different exit from the hard problem entirely: rather than explaining phenomenal consciousness, it argues phenomenal consciousness as usually described is itself something of an illusion generated by the brain's self-monitoring, there's a seeming of rich, ineffable experience, but the seeming is functional information processing, not evidence of an additional, separate phenomenal fact needing explanation. And John Searle, whose Chinese Room I used above to challenge computationalism, actually holds his own distinct position, biological naturalism, which argues consciousness is a real, physical, but irreducibly biological process, rejecting both the substrate independence of computationalism and the reductive move of the classical identity theory.

I haven't given any of these three the full treatment the others above got, and I don't want to pretend I have. I'm noting them so the survey doesn't read as more complete than it is.

The open questions this hands off

I've deliberately not answered several things this piece raises, because each is a real fork that deserves its own space rather than a rushed paragraph here:

Do humans have "the right ingredients," and do other things? Every theory above implies a different answer to whether animals, plants, or even simpler systems are conscious, IIT in principle allows any system with nonzero Φ to have some degree of experience, however faint; GWT would look for something like a global broadcasting architecture, which gets much harder to locate outside brains with the relevant structure; panpsychism and Advaita, in different ways, refuse to draw a hard line at all. This deserves a dedicated piece.

Can this be tested using AI? Given that I'm already doing mechanistic interpretability work on self-referential structure in language models, this is the question I'm most personally positioned to dig into directly, does a system need anything like IIT's integration, or GWT's global broadcast architecture, to produce the functional signatures associated with consciousness, and would building those signatures in an artificial system tell us anything at all about whether something is actually experienced, or only that the right function is present? This is close to the Chinese Room question, restated for the systems I actually study.

Is the emergent/non-emergent fork even the right fork? I've organized this whole piece around a binary, and I'm aware binaries are often a simplification that later turns out to hide a third option nobody was looking for. I don't have a candidate for that third option yet. I'm noting the possibility so I don't forget to look for it.