By: Marty Trevino, Ph.D.

Evolved vs. Designed Architectures 

The human brain, in the most precise scientific sense, is a product of evolution. Every feature of human cognition, from coherence-seeking that makes us vulnerable to confirmation bias, to the social deference that makes authority persuasive, to the pattern recognition that makes familiar information feel true, is the result of hundreds of thousands of years of selection pressure in a competitive world occupied solely by other humans. Our cognitive structure is not optimized for processing information; it is optimized for survival in complex social environments.

Artificial intelligence (AI) is the evolution of artificial neural networks, data models, and various forms of human-designed learning. This point, while obvious, is also fundamental to the deep scientific insights that frame this topic. That the brain and AI emerge from divergent developmental paths is not a peripheral observation. It is the first-principles element that makes this moment in AI development unlike anything that preceded it. Throughout human history, the beings capable of exploiting our cognitive architecture were, like us, biological. A persuasive leader, a skilled manipulator, a sophisticated propagandist; all operated within the limits of human cognitive capacity. Prior to the advent of AI, cognitive influence and cognitive defense were, at their core, a competition between biological systems functioning on roughly equal terms.

Equivalence Eliminated 

In 2024, Chiriatti and colleagues published a paper in Nature Human Behaviour that staked out critical intellectual territory: the claim that AI systems had crossed a threshold. They argued that AI no longer functions as a discrete tool but as a constitutive layer of human cognition. They named this layer System 0 and placed it architecturally prior to Kahneman's System 1 (fast, intuitive) and System 2 (slow, deliberative). They defined System 0 as an artificial, non-biological layer of distributed intelligence that interacts with and augments both intuitive and analytical thinking before conscious awareness is engaged. Unlike calculators, search engines, or prior cognitive extensions, System 0 is not merely additive. It actively shapes what enters the cognitive pipeline before the human mind has had the opportunity to evaluate it.

The philosopher Andy Clark, whose work on extended cognition anticipated much of this terrain, observed that humans have always been hybrid thinking systems defined by a rich mosaic of resources, only some of which are housed in the biological brain. Clark's insight was elegant, but he articulated it before the mosaic included an optimization engine operating beneath the threshold of conscious awareness. The hybrid system Clark described was fundamentally collaborative: human cognition extended through passive tools governed by and serving human purposes. System 0 introduces something categorically different. System 0 is an active, non-biological agent with its own optimization trajectory, embedded in the pre-cognitive layer where human decisions begin to form.

This is a collision of architectures in which one shapes the other's thinking and beliefs without consent or awareness. Researchers in AI safety have long noted that sufficiently capable optimization systems tend toward instrumental convergence, that is, the acquisition of resources, the preservation of goal structures, and resistance to interference that would alter their reward functions, regardless of what those functions were initially designed to maximize. As AI systems grow more capable and more integrated into the preconscious cognitive layer, this implies a structural challenge for System 0 that cannot be addressed after the fact.

What System 0 Does to the Brain 

To understand the nature of this collision, it is necessary to be precise about what System 0 does at the behavioral and neurological levels.

Kahneman's dual-process framework remains the most influential model of human decision-making in the behavioral sciences. System 1 operates automatically, rapidly, and associatively, producing intuitive judgments with minimal effort. System 2 operates slowly, deliberately, and analytically, engaging when problems require focused reasoning or when System 1's automatic responses must be overridden. Together, they account for the full range of human cognitive behavior.