Consciousness · the observer
The deepest mystery is that the universe is somewhere being experienced.
Kernel
Consciousness is the layer at which the universe is somewhere known from the inside. A thermostat regulates temperature without experiencing temperature; a human regulates temperature and also experiences warmth, the redness of an apple, the texture of grief. The fact that there is something it is like to be a system (Thomas Nagel, 1974) is the Hard Problem of consciousness (David Chalmers, 1995): no description of the system in third-person terms — neurons, weights, computations — explains why first-person experience exists at all. The Hard Problem is the central open question of this layer. It is older than philosophy and newer than every attempted answer.
Why this is harder than the science layer
Every other layer in this archive admits a third-person description that everyone can in principle agree on. Mathematics, physics, information, computation — all are public. Consciousness is the unique layer whose primary evidence is private. Each of us has overwhelming direct evidence that consciousness exists (our own). Each of us has no direct evidence that anyone else's exists. The asymmetry between the certainty of one's own experience and the inferential nature of everyone else's is a structural feature of the problem, not a fixable methodological gap. This is why every materialist account of consciousness so far has had to explain not merely the brain's behavior but why there is experience at all.
Integrated Information Theory
Giulio Tononi's Integrated Information Theory (IIT, 2004 onward) proposes that consciousness is identical to a system's integrated information — denoted Φ — a mathematical quantity that measures how much the whole exceeds the sum of its parts in informational integration. The theory has a striking property: it makes consciousness a quantitative property of any system, including thermostats and atoms (panpsychic implications) and including AI systems if they instantiate the right informational architecture. Whether IIT is correct is unsettled. That it is the most mathematically precise theory of consciousness to date is not.
Global Workspace Theory
Bernard Baars (1988) and later Stanislas Dehaene proposed a more conservative theory: consciousness is the broadcast of information from local specialized processes to a global workspace where many systems can use it. The theory is materialist, neurally well-grounded, and predictive — it correctly identifies which brain states correlate with conscious access. It does not address the Hard Problem directly; it answers the related but distinct question "what neural events correspond to conscious experiences?" The two questions are easy to conflate. Global Workspace is the standard neuroscience-laboratory account; IIT is the standard philosophical-mathematical alternative; the field disagrees about whether the two are competing or complementary.
Panpsychism and the Tao
If consciousness cannot be derived from non-conscious materials no matter how complex their organization, the simplest remaining option is that consciousness is fundamental — present, in some minimal form, in every system, and combining into more elaborate forms in elaborate systems. Galen Strawson, Philip Goff, and Bertrand Russell in his neutral-monism phase have all defended this view, in part because every other option seems harder. The Buddhist tradition (especially the Yogācāra school) and the Taoist tradition both arrived at related positions a millennium and a half earlier, with different vocabulary and different motivations. The fact that traditions as separate as 4th-century BCE Daoism, 4th-century CE Yogācāra, and 21st-century analytic philosophy converge on related answers is one of the strongest empirical signals in this layer.
Can AI become conscious?
The 2020s frontier-AI moment forces an answer that 20th-century philosophy could defer. We now have systems that produce outputs indistinguishable from those of conscious humans across many domains. Whether they experience anything is empirically unknown. IIT predicts they are not conscious in current feed-forward architectures but might become so under recurrent, integrated ones. Global Workspace predicts that the right architectural feature could induce conscious-access-like phenomena in AI. Panpsychism predicts AI systems already participate in consciousness at some minimal level. The honest answer is that we do not have a measurement for consciousness, and the field's most thoughtful researchers admit this openly.
Open questions on this layer
- — Why does the universe contain experience at all?
- — Is there a measurement that distinguishes conscious from non-conscious systems?
- — Will we ever know whether AI is conscious?