Consciousness: A Definition Science Can Use
In October 2026, New Scientist magazine published the article, “Do AIs already process information in a way that gives rise to consciousness?“
For more than two thousand years, philosophers have debated the nature of consciousness. More recently, neuroscientists and artificial intelligence researchers have joined the search. Yet much of that effort remains entangled in a fundamental problem: Consciousness often is defined in terms that resist scientific observation, measurement, or verification.
The New Scientist article, “Do AIs already process information in a way that gives rise to consciousness?“, illustrates this difficulty. It explores whether artificial intelligence systems might possess consciousness because their information-processing mechanisms resemble certain processes in the human brain.
The investigation is worthwhile. The difficulty isn’t in studying information processing, but in assuming that information processing must somehow produce an additional phenomenon called “subjective experience.”
Science has spent centuries trying to explain consciousness while defining it in terms that resist scientific investigation. Rather than search for a physical explanation of an undefined subjective essence, we should define consciousness as a physical process and investigate its manifestations.
The Article’s Central Claims
The article examines whether artificial intelligence systems may process information in ways that give rise to consciousness. It draws on distinctions between kinds of consciousness and theories of how the brain makes information available for thought and action.
The following three claims summarize the conceptual approach discussed in its accessible material. They are interpretations, not direct quotations from the author.
Claim A: Consciousness involves subjective experience.
The article discusses what philosophers call phenomenal consciousness: the supposedly private, subjective character of an experience.
For example, a person sees a red apple. Scientists can investigate the light entering the eye, the neural signals traveling to the brain, the recognition of the apple, and the person’s responses. But philosophers ask an additional question: What is it like to see red?
That question introduces the notion of qualia—the supposedly distinctive subjective qualities of experiences. But what, exactly, are they proposing?
QUALIA: The individual, subjective, phenomenal instances of conscious experience, essentially, “what it is like” to feel, taste, see, or hear something.
Key Characteristics:
- Subjective and Private: They belong strictly to the person experiencing them and cannot be directly observed or measured by an outside observer.
- Intrinsic: They are built directly into the texture of the experience rather than being a secondary thought or judgment.
- Introspectively Accessible: You become aware of them the moment you turn your attention inward to examine your own mind.
If qualia are physical processes, we can investigate them as such. If they aren’t physical, what evidence establishes their existence, and how could science investigate them? Merely naming an unexplained phenomenon doesn’t explain it.
Claim B: Certain information-processing mechanisms may produce consciousness
The article discusses theories in which information becomes widely available across a system, enabling memory, reasoning, decision-making, and coordinated action.
Global workspace theory is one example. These mechanisms are legitimate subjects of scientific investigation. Researchers can study how information travels, how memories affect decisions, and how different system components interact.
The difficulty arises when researchers ask whether these processes give rise to consciousness, understood as an additional subjective phenomenon. Why assume that the observable processes must produce something beyond themselves?
This is like studying the chemical reactions involved in life and then asking whether those reactions produce an invisible vital force. Before investigating its production, we must first define the additional phenomenon.
Claim C: Artificial intelligence may or may not be conscious, depending on whether it possesses the necessary mechanisms
The article considers whether artificial systems possess information-processing architectures sufficiently similar to those associated with human consciousness. That is a legitimate engineering and scientific question.
But why should an AI system need to resemble a human brain to be conscious? A bird flies by flapping its wings. An airplane flies without flapping wings. Both fly. Likewise, a human responds to stimuli through biological neurons. A computer responds through electronic circuits. A thermostat responds through temperature-sensitive mechanisms.
Their capabilities differ enormously. Their structures differ. Their histories differ. But why should those differences determine whether they qualify as conscious?
That question becomes particularly troublesome when consciousness is defined by something that nobody has established how to identify independently.
An alternative scientific approach
Rather than begin with subjective experience and search for its physical causes, we can begin with an observable physical process.
Proposition D: Define consciousness physically
Consciousness is the process by which an entity’s history shapes its responses to internal and external stimuli.
This definition requires no supernatural substance, mysterious inner observer, or unmeasurable subjective ingredient.
It identifies three essential elements:
- History: The entity’s accumulated physical state, including its structure and the effects of previous interactions.
- Stimulus: Any internal or external influence that produces a response.
- Response: The resulting physical activity or change. Each response becomes part of the entity’s history, shaping its responses to future stimuli.
The process can be expressed as: History + Stimulus —> Response —> New History.
This is a recursive physical process. No qualia are required.
Proposition E: Consciousness is universal, but its manifestations differ
Under this definition, consciousness isn’t a special possession of humans or a privilege granted to sufficiently sophisticated organisms. It is a process occurring throughout the physical universe.
An atom responds according to its physical properties and history. A cell responds through biochemical processes. A human responds through extraordinarily complex biological systems. A computer responds through electronic processes and its programmed and accumulated states. The differences are in the structures, histories, stimuli, and responses.
Consider a two-year-old child and an adult mathematician. The mathematician can solve equations that the child can’t. But this doesn’t establish that the mathematician is conscious while the child isn’t.
Similarly, a dog detects odors humans can’t. This shows a difference in sensory capabilities, not the presence of some mysterious ingredient.
One entity may exhibit more elaborate forms of consciousness than another, but complexity isn’t a prerequisite for the fundamental process.
Nor must all parts of an entity respond identically or communicate continuously. We can examine a human heart, its individual cells, and the human body as a whole as entities at different organizational levels. Each has its own history-dependent responses.
A crowd provides another example. When hundreds of people witness a frightening event, their individual responses may combine into a collective response. They don’t need a mysterious shared mind. Researchers can investigate their collective behavior through the histories, connections, stimuli, and responses of the participating individuals.
In this formulation, consciousness can be examined at whatever organizational level is scientifically useful.
Proposition F: Study the processes, not an assumed invisible product
This approach changes the questions scientists ask. Instead of asking, “Does this computer really experience anything?” we can ask: How does its history influence its responses? How does it retain information? How do internal processes affect its behavior? How do its responses modify its future behavior?
Instead of asking, “Does an anesthetized patient still possess qualia?” we can investigate which neurological processes continue, which are suppressed, and which responses become possible when anesthesia ends.
Instead of asking, “Does a thermostat have subjective awareness?” we can investigate how its physical structure and previous states determine its responses to temperature.
These are scientifically answerable questions. They don’t require us to pretend that all entities have identical capabilities. Nor do they require us to establish an arbitrary threshold beyond which consciousness supposedly appears. They require us to investigate observable physical processes.
A predictable objection is that the proposed definition is too broad. If consciousness is history-dependent responsiveness, does that include virtually every physical interaction? Yes, but broadness alone doesn’t invalidate a definition.
Gravity operates throughout the universe. Its universality doesn’t make it scientifically useless. Similarly, chemistry describes processes occurring in living and nonliving matter. Biology investigates particular organizations of those processes.
A universal definition of consciousness doesn’t eliminate specialized investigations of memory, perception, learning, intelligence, emotion, or self-recognition.
It can provide a common framework for studying them.
Yes, this proposed definition departs significantly from how philosophers and neuroscientists traditionally use the term. It doesn’t aim to invalidate existing philosophical discourse. Rather, it suggests replacing an elusive, non-operational concept with a physically testable framework, ultimately clarifying the scope of scientific inquiry into consciousness.
The lesson from vitalism
For centuries, some philosophers and scientists believed living organisms possessed a special vital force that distinguished them from nonliving matter. Life supposedly required something beyond ordinary chemistry and physics.
Modern biology advanced by investigating the physical processes of living systems without invoking an additional force.
Scientists discovered increasingly detailed mechanisms of metabolism, reproduction, heredity, and cellular organization.
Those explanations made the concept of a vital force unnecessary.
Perhaps consciousness research faces a comparable conceptual obstacle.
We investigate neural activity, perception, memory, emotions, and responses. Yet we continue asking whether these processes somehow generate an additional, privately experienced phenomenon. But what evidence requires that additional phenomenon?
The fact that humans describe their experiences doesn’t establish a separate, nonphysical ingredient. Those descriptions themselves are physical responses, produced by physical systems with particular histories. The scientifically useful question is how those responses arise.
We can study every aspect of love without postulating a nonphysical essence of love. It should likewise be possible to study consciousness without postulating a nonphysical essence of consciousness.
What this means for artificial intelligence
The article’s question—whether artificial intelligence already processes information in ways that give rise to consciousness—can be reformulated. Rather than asking whether AI produces an additional phenomenon called consciousness, ask how its history shapes its responses to stimuli.
An AI system’s history may include its hardware design, training, learned parameters, current inputs, and information retained during operation. These influence its responses. Under the proposed definition, that is consciousness.
Scientifically interesting questions then concern the characteristics of that consciousness:
- How does it differ from human consciousness?
- What kinds of information can it retain?
- How does it modify its responses?
- What limitations arise from its structure?
- How does it interact with other systems?
We can study these questions without first determining whether an AI has a metaphysical, magical, undefined, inner feeling.
Importantly, calling an AI conscious under this definition doesn’t establish that it feels pain, experiences love, possesses humanlike awareness, or deserves any particular moral status. Those are separate questions requiring their own definitions and evidence.
If, for instance, feeling pain is a criterion for consciousness, then virtually all animals would qualify, and we don’t know whether plants would. We also know that we could program a computer to physically respond in ways that mimic human pain, and by the definition of “inner feelings,” we would have no way to determine whether the feeling is “real.”
The same goes for love, awareness, and morals. In each case, an arbitrary definition might determine inclusion, yet we still would lack evidence.
Conclusion: We should stop searching for an undefined “subjective essence” and instead ground our research in observable reality. The New Scientist article highlights both the potential and the pitfalls of current consciousness studies. While its investigation into the physical mechanisms of information processing is promising, its lingering tendency to frame consciousness as a mysterious, untestable byproduct of those mechanisms remains a significant obstacle to scientific progress.
The alternative proposed here is simpler and testable: Consciousness is a physical process: the interaction of history and stimuli. The definition establishes a physical subject for scientific investigation:
History + Stimulus —> Response —> New History
Science contains many examples of progress achieved by abandoning unnecessary assumptions. The rejection of vitalism is one. Replacing supernatural explanations with physical mechanisms is another.
We don’t deny the reality of pain, love, perception, thought, or emotion. We only need to investigate them as physical processes rather than treating them as evidence of an additional, undefined substance or property.
Science has spent centuries trying to explain consciousness while defining it in terms that resist scientific investigation. The proposal is that scientists investigate consciousness without first assuming something that can’t be defined or measured.
Researchers sometimes investigate measurable physical processes while retaining an additional, poorly specified criterion for deciding whether those processes constitute consciousness. Instead, we challenge the necessity of that criterion. We have replaced an argument about an inaccessible property with an investigation of observable processes.
Some might claim that a computer isn’t conscious because it lacks subjective experience, or qualia. The response: “How do you know?” No one can observe an AI’s supposed subjective experience directly. Nor can they inspect a human’s. They can investigate physical structures, processes, and responses, but those observations don’t independently establish whether qualia are present.
Some might argue that an AI lacks the necessary biological mechanisms. But that assumes biology is necessary for consciousness—the very proposition they would need to demonstrate.
Others assume biology alone isn’t enough, and that many biological entities aren’t conscious.
Under the suggested definition, consciousness is observable. Under the qualia-based definition, we can’t conclusively demonstrate either AI consciousness or its lack.
Rodger Malcolm Mitchell
