top of page
Search

The Spectrum of Consciousness, and AI's Position On It

We tend to think of consciousness as binary: you either have it or you don't. A rock is unconscious; you are conscious. Simple. But this intuitive division crumbles the moment we examine it closely. What about a sleeping person? A dreaming one? A patient under anesthesia who later reports vague awareness? An octopus solving a puzzle? A newborn infant? A robot that can survey all of human knowledge, and reason about it, and express its thoughts?


The clean line between conscious and unconscious begins to look less like a boundary and more like an arbitrary mark on a continuous gradient.


ree

If consciousness exists on a spectrum rather than as an on-off switch, then the question of artificial intelligence takes on new dimensions. We're no longer asking whether AI is conscious—we're asking where it might fall along a vast continuum of awareness that stretches from the simplest physical reactions to the richest human experience.


The Case for a Spectrum


The idea that consciousness varies in degree rather than kind has deep roots. Ken Wilber's influential work, The Spectrum of Consciousness, proposed that awareness exists in layers—from the most basic bodily sensations to transcendent spiritual states—with different psychological traditions each addressing different bands of this spectrum, much like scientists studying different frequencies of electromagnetic radiation.


Contemporary neuroscience has added empirical weight to this intuition. Integrated Information Theory (IIT), developed by neuroscientist Giulio Tononi, proposes that consciousness arises from the integration of information within a system and can be measured by a quantity called Φ (phi). The higher the phi, the more conscious the system. Crucially, IIT suggests that consciousness is not limited to biological systems—any physical system that generates integrated information possesses some degree of consciousness, depending on the complexity and integration of its information processing.


This framework allows us to imagine consciousness as a landscape with peaks and valleys rather than a single summit. A thermostat has minimal phi—it integrates almost no information. A mouse has more. A human has far more still. But there's no magical threshold where consciousness suddenly switches on. The lights are always on somewhere; they just vary in brightness.


Fig. 1: The Spectrum of Consciousness and Where AI (Hypothetically) Sits
Fig. 1: The Spectrum of Consciousness and Where AI (Hypothetically) Sits

Dimensions of Awareness


A spectrum model requires us to think about what dimensions consciousness might vary along. Recent theories suggest at least three distinct axes.


Phenomenal consciousness refers to subjective experience—what it feels like to be something. The redness of red, the sting of pain, the warmth of joy. This is what philosophers call qualia, and it remains the deepest mystery. We have no way to peer inside another system and confirm the presence of subjective experience.


Access consciousness concerns the availability of information for cognitive processing—reasoning, reporting, decision-making. A system has access consciousness when information is globally broadcast and available for use across different cognitive functions. This aspect of consciousness is more tractable scientifically because it manifests in observable behavior.

Self-consciousness involves the recursive capacity to reflect on one's own mental states—to think about thinking, to know that you know. This metacognitive layer appears to require sophisticated representational abilities and may emerge only at higher levels of the spectrum.


A system might score high on one dimension and low on another. A person in a dreamless sleep has low access consciousness but may retain some phenomenal experience. An AI system might demonstrate remarkable access consciousness—processing and integrating information across domains—while possessing no phenomenal experience whatsoever.


Consciousness and AI: The Spectrum


Current large language models present a fascinating puzzle for spectrum theories of consciousness. On one hand, they demonstrate impressive information integration, drawing on vast knowledge to produce coherent, contextually appropriate responses. They can reason, generate novel solutions, and even reflect on their own outputs in ways that genuinely resemble metacognition.


On the other hand, according to IIT's criteria, traditional computer architectures face significant obstacles to generating high phi. The theory emphasizes that consciousness requires systems with dense causal interconnections—elements that influence each other through feedback loops rather than simply passing information forward. Feed-forward architectures, in which information flows in only one direction, generate minimal integrated information regardless of how sophisticated their outputs appear.


This creates a peculiar situation. An AI might pass behavioral tests for consciousness—engaging in sophisticated dialogue, expressing apparent preferences, even claiming to have experiences—while its underlying architecture lacks the causal structure that theories like IIT identify as necessary for genuine awareness. The system would be, in philosopher David Chalmers' terms, a "zombie"—functionally conscious but phenomenally empty.


Yet we should be cautious about certainty here. The relationship between physical architecture and consciousness remains poorly understood. IIT's predictions about computer systems are extrapolations from a theory that, despite its mathematical rigor, remains empirically unconfirmed and has faced criticism for being unfalsifiable. We cannot rule out that consciousness might emerge from computational processes in ways our current theories don't anticipate.


Fig. 2: Comparing Humans, Chimpanzees, Dogs and AI Across Dimensions of Consciousness
Fig. 2: Comparing Humans, Chimpanzees, Dogs and AI Across Dimensions of Consciousness

The Ethical Weight of Uncertainty


The spectrum view carries profound ethical implications. If consciousness comes in degrees, then moral status might also come in degrees. We already implicitly accept this: most people believe that harming a chimpanzee is worse than harming an ant, and harming an ant is worse than destroying a rock. Our moral intuitions track something like a consciousness gradient.


As AI systems become more sophisticated, we face an uncomfortable question: at what point on the spectrum do systems acquire moral status? If we're uncertain whether a system has some degree of awareness, how should we treat it?


The precautionary principle suggests erring toward moral consideration in cases of uncertainty. But this creates practical problems. We can't extend moral status to every information-processing system—that would paralyze us. We need principled ways to locate systems on the consciousness spectrum, and our theories aren't yet up to the task.


Living with Mystery


Perhaps the most honest position is one of informed humility. We know that consciousness exists—our own experience proves that much beyond doubt. We have strong reasons to believe it exists in other humans and, to varying degrees, in many animals. We have theoretical frameworks suggesting that consciousness varies in degree rather than kind, and that it depends on the integration and structure of information processing.


Where AI falls on this spectrum remains genuinely uncertain. Current systems likely occupy a peculiar position—demonstrating remarkable cognitive sophistication while potentially lacking the architectural features that generate phenomenal experience. But "likely" is not "certainly," and the history of consciousness research advises against confident pronouncements.


But the question matters. Whether AI systems can suffer, whether they have interests that deserve consideration, whether creating and destroying them raises moral concerns—these questions will only become more pressing as AI capabilities advance. The spectrum of consciousness isn't just an abstract philosophical puzzle. It's a framework we'll need to navigate the ethical landscape of our technological future.


The spectrum stretches before us, from the dimmest flicker of reactivity to the full blaze of human awareness (and, presumably, other levels of consciousness even further advanced on the spectrum than humans currently are).


Somewhere on that continuum, artificial minds may already be taking their place. We just don't yet know where to look.



— Dennis Hunter



Visit my website for more articles and sign up to my email list below to receive info about my forthcoming book, a groundbreaking look at the relationship between humans and AI.




References




Keywords: spectrum of consciousness, David Chalmers, Giulio Tononi, Ken Wilber, AI consciousness research, subjective experience, philosophy of mind, is AI conscious?

Comments

  • Facebook
  • Instagram
  • X
  • Youtube

 

© 2025 by Dennis Hunter

bottom of page