Algorithmic Culture

In the AI Story

The defining feature of algorithmic culture is the speed and scale at which pattern-processing occurs. A human theoretic thinker—a scientist, a lawyer, a programmer—can manipulate external symbols, can reason through formal relationships, can construct systematic arguments. But the pace of this work is limited by the serial processing constraints of biological attention. Working memory holds seven items; reading occurs at three hundred words per minute; writing accumulates at perhaps fifty words per hour for sustained creative work. AI removes these biological bottlenecks. The system can process millions of tokens per second, can hold in 'attention' (the technical term is context window) the equivalent of hundreds of pages, can generate in minutes what would take a human days or weeks.

This quantitative difference produces qualitative effects. The builder working with AI can explore possibility spaces that biological cognition cannot traverse in available time. The design that would have taken six weeks to prototype can be tested in six hours. The analysis that would have required a research team can be conducted by a single well-directed practitioner. The imagination-to-artifact ratio—the gap between what you can conceive and what you can produce—collapses toward zero for a significant class of creative work. This is not augmentation in the sense of making an existing human capacity slightly stronger. It is the addition of a new capacity operating at a different order of magnitude, reorganizing what is possible.

But Donald's framework insists on a critical distinction that the triumphalist AI narrative elides. The algorithmic layer processes patterns extracted from the products of mimetic, mythic, and theoretic work. It does not and cannot replace the work itself. The AI that generates code has learned from millions of human-written programs, each the product of a programmer's mimetic engagement with tools, mythic understanding of user needs, and theoretic grasp of computational logic. The training data is the crystallized residue of multi-layer human cognition. When current practitioners stop producing that multi-layer work—when the algorithmic layer becomes the only layer actively developed—the training data source begins to degrade, and the degradation propagates through every subsequent generation of models.

The structural risk is that algorithmic culture could become a monoculture, a single-layer cognitive ecology that initially appears more efficient than the hybrid it replaces but that lacks the resilience multi-layer systems provide. The engineer who can code, who understands embodied engagement with systems, who can construct narrative explanations of technical decisions, and who can direct AI tools is more valuable—and more adaptable—than the engineer who possesses only algorithmic facility. When the algorithmic environment changes, as it inevitably will, the multi-layer practitioner can adapt by falling back on lower-layer capacities. The single-layer practitioner has nowhere to fall.

Origin

Merlin Donald did not explicitly propose 'algorithmic culture' as a fourth stage in his original trilogy, but the concept is a natural extension of his framework into the domain of artificial intelligence. In interviews and later essays, Donald acknowledged that digital technologies represent a significant reorganization of the cognitive landscape, though he remained cautious about declaring them a transition comparable to writing. The simulation framework developed in this volume takes the step Donald hesitated to take, arguing that AI's externalization of processing—not merely storage—meets the criteria for a genuine cognitive transition.

The argument rests on the observation that AI reorganizes the relationship between biological and cultural cognition in ways that previous technologies did not. Writing externalized memory; AI externalizes inference. The calculator externalized arithmetic; AI externalizes reasoning across domains. The database externalized structured recall; AI externalizes the discovery of patterns within that structure. Each previous tool required the human to formulate the question and evaluate the answer. AI increasingly handles both ends of the cognitive process, leaving the human to specify goals at a higher level of abstraction. This is not a difference of degree but of kind, and it justifies treating the AI transition as a fourth layer in Donald's evolutionary architecture.

Key Ideas

Processing, not just storage. Previous cognitive technologies externalized memory; AI externalizes inference, pattern-finding, and generation—a qualitative leap beyond passive storage to active processing.

Speed and scale beyond biology. AI removes the serial bottlenecks of biological attention, processing millions of tokens per second and holding context equivalent to hundreds of pages, reorganizing what is cognitively possible.

Built on prior layers. The algorithmic layer processes patterns extracted from the products of mimetic, mythic, and theoretic work—it does not replace those layers but depends on their continued vitality as the source of training data.

Monoculture risk. If algorithmic culture becomes the only actively developed layer, the multi-layer foundations upon which it rests will erode, producing a cognitive ecology that is efficient but fragile.

Genuine transition. AI's reorganization of the relationship between biological and cultural cognition meets the criteria for a fourth major stage in human cognitive evolution, comparable in significance to writing or language itself.