Consciousness may have evolved to help living things resolve competing needs when their survival is at stake, rather than simply emerging from greater intelligence. This theoretical account was published in Frontiers in Neuroscience.
How did conscious experience emerge from the chemistry of life? Existing theories examine how brains broadcast information, integrate signals, and generate predictions. But explaining these mechanisms leaves an evolutionary question: what problem did consciousness help organisms solve?
Jeff Stibel, a cognitive scientist and author who serves as a trustee of the Natural History Museum of Los Angeles County and Tufts University, proposes that cognition and consciousness arose from different evolutionary pressures. Cognition helped organisms navigate uncertainty, while the regulatory processes underlying consciousness helped them preserve themselves and manage competing demands.
“This paper was deeply personal. It grew out of years of conversations with my late mentor, Dan Dennett, who pushed me to formalize some of my thoughts on the evolutionary differences between cognition and consciousness,” explained Stibel. “When Dan passed, I felt an obligation to revisit our discussions, and it became even more pressing given the advances we have seen across AI. What pulled me in was a growing suspicion that we have been asking the wrong question: treating consciousness as the top rung of intelligence, when it looks more like a separate adaptation with its own evolutionary job. The timing felt urgent as AI grows more capable, alongside a widespread assumption that a clever enough machine will simply ‘wake up’ or has already awoken. That narrative is everywhere, and I think it rests on the same confusion.”
Stibel draws together existing research in cell biology, evolutionary theory, neuroscience, and artificial intelligence. He traces a possible progression from cells protecting their boundaries to organisms coordinating multiple needs under conditions of uncertainty, limited resources, and potentially irreversible consequences.
The author outlines six predictions to make the framework testable. These concern whether conscious engagement increases with competing demands; whether different needs must become comparable; whether attention, bodily state, memory, and readiness to act change together; whether consciousness diminishes when this coordination is disrupted or less necessary; how these capacities differ across species; and whether artificial systems require genuine self-maintenance pressures.
Suggested experiments would vary conflict or risk while holding sensory input constant, disrupt coordination while preserving simpler processing, or compare artificial systems with different self-maintenance requirements. The paper also identifies potentially disconfirming findings, including consciousness persisting despite disrupted settlement capacity.
The central argument describes three levels of regulation. “Proto-control” protects an organism’s integrity, such as by preserving a cell’s membrane and limiting destructive processes; the paper explicitly does not describe this as conscious. “Meta-control” manages conflicts among separate regulatory processes. “Settlement control” makes competing needs comparable and coordinates them into a shared course of action. For example, a hungry animal encountering food near a possible predator must weigh feeding against danger. A decision to retreat would involve coordinated changes in attention, bodily readiness, movement, and learning. Stibel identifies consciousness with this coordinated state in a system whose continued existence depends on its own regulation.
“Intelligence and consciousness are not the same thing, and they did not evolve for the same reason. Intelligence is outward facing: it is for modeling the world, predicting, planning, solving problems. Consciousness evolved for something more internal: holding a fragile organism together when survival is on the line. The analogy I use in the paper is cannibalism. Intelligence helped early life forage efficiently, but without a self to protect, the most efficient meal was often the nearest neighbor, or the organism itself. The earliest precursors of consciousness helped define biological boundaries, which likely protected early cellular life from eating itself.”
Applied to artificial intelligence, the argument is that greater capability alone would not produce consciousness: a system would also need to maintain itself and resolve competing demands with genuine consequences for its continued existence.
“The upshot for AI is blunt. Scaling today’s models makes them better predictors, not systems with anything of their own at stake. LLMs have no boundaries and no self to protect, and humans provide everything they need,” Stibel told PsyPost. “In that respect, the problem isn’t that they are too simple or fall short of AGI. They are too coddled. Without real skin in the game, no amount of capability will magically cross over into consciousness.”
The paper does not restrict consciousness to biological organisms. However, an artificial system would need more than programmed self-preservation, such that its continued operation would have to depend on conditions that it maintains itself.
Stibel emphasized the implications: “The most important takeaway is that biology and evolution reframe the AI debate. The fear that we will accidentally scale our way into a suffering machine, and the engineering hype that digital minds are just around the corner, both rest on conflating intelligence with consciousness. If those are genuinely separate things, then consciousness is not something that sneaks up on us through sheer capability. It is a specific kind of organization we would have to deliberately build, and I would argue, one we have good reason not to. Getting that distinction right is how we choose our ethics rather than stumble into them.”
Testing this account presents a fundamental challenge: researchers may be able to examine whether a system has the proposed capacities for self-maintenance and coordination, but they cannot directly observe whether it has subjective experience.
“The biggest unanswered question is measurement,” Stibel said. “It is impossible to observe consciousness in anyone other than yourself, so the paper proposes testing the necessary precursors instead. At the heart of the test is whether a system genuinely works to maintain its own existence under constraint, rather than just reacting to threats it was built to detect. Turning that into a real experiment, in both biological and artificial systems, is the hard road ahead.”
Stibel also identified an unresolved ethical question: “There is also a deeper question the paper raises but doesn’t settle. If we ever did build a system with genuine stakes in its own survival, we would have created something that can be harmed, a moral patient whose interests may not align with ours. That is a line worth understanding well before we approach it, not after.”
The paper, “From chemistry to cognition: the adaptive origins of consciousness under constraint,” was authored by Jeff Stibel.