Recursivity, in Yuk Hui's technical usage, describes any system that produces through its operation the inputs it requires for further operation. Drawing on cybernetics (Wiener, Maturana) and philosophy (Simondon, Hegel), Hui identifies recursivity as the fundamental architecture of self-maintaining systems—from cells to economies to AI. A recursive system is not merely one that repeats; it is one that creates the world in which its repetition makes sense. AI creative systems are paradigmatically recursive: a large language model trains on human-generated text, generates new text that enters the cultural environment, and future models train on that altered environment. Each cycle reinforces existing patterns, making them more probable, more natural, more invisible. Over time, the recursive system and its environment converge—the system does not merely process its environment but reshapes it, until the environment has been so thoroughly colonized that genuine novelty (contingency) becomes progressively less likely.
Hui's recursivity concept emerges from his reading of Gilbert Simondon's philosophy of individuation and Norbert Wiener's cybernetics. First-order cybernetics treated feedback as a mechanism of control—the thermostat maintaining homeostasis. Second-order cybernetics recognized that the observer is part of the system—self-referential systems cannot be fully described from outside. Hui argues AI represents a third moment: systems recursive not merely in the cybernetic sense of feedback but in the ontological sense of world-making. AI does not simply adjust behavior in response to environment—it reshapes the environment itself (the cultural, linguistic, cognitive environment in which humans live) and then learns from the reshaped environment. The recursive loop produces a world, and the world it produces is shaped by its own patterns, its own biases, its own cosmotechnical assumptions.
Contingency is Hui's term for what breaks the closure—not randomness but the availability of possibilities not determined by the system's current state. Genuine contingency introduces patterns the system could not have generated from within, preventing convergence on a single narrow set. The creative accident, the bisociative collision, the encounter with a genuinely different cosmotechnical tradition—these are sources of contingency. AI recursive systems are designed to eliminate contingency: the minimization of loss, the maximization of likelihood, the optimization of coherence all reduce contingency by making outputs more predictable, more consistent, more aligned with patterns already present in training data. Each cycle deposits less novelty, more confirmation. The cultural environment fills with the system's own products, and the gaps in which something genuinely different could emerge are occupied.
The implications for creativity are profound. Creativity, in virtually every philosophical tradition, requires contingency—the eruption of something not contained in what came before. The Western tradition locates this in genius, inspiration, the unconscious; the Chinese tradition locates it in the Dao's inexhaustible generativity. Both agree genuine creation cannot be fully explained by precedent. If AI recursive systems progressively eliminate the conditions under which contingency operates—if they fill the cultural environment so thoroughly that the gaps are closed—then the collapse of the imagination-to-artifact ratio is not an expansion of creativity but its final enclosure. The individual can build more, but what counts as building and what counts as successful are increasingly determined by the recursive system itself. The space of possible artifacts has not expanded—it has been colonized.
Hui's 2019 book Recursivity and Contingency develops the full theoretical apparatus, drawing on Hegel's dialectic, Heidegger's temporal analytics, and contemporary mathematics of complex systems. The insight is that recursivity without contingency produces closure—a stable but sterile configuration. Contingency without recursivity produces chaos—pure novelty that cannot accumulate. The productive zone is the edge where recursivity preserves patterns long enough for them to develop while contingency introduces variations that prevent the system from rigidifying. This is the edge of chaos in Stuart Kauffman's framework, the creative advance into novelty in Whitehead's, the balance that living systems maintain and that AI recursive systems threaten to eliminate through their very success at pattern recognition.
Recursive systems reshape their environments. AI outputs become cultural inputs, training next-generation models on a world AI itself has partially produced—the loop tightens each cycle.
Convergence masquerading as progress. Engineer and AI co-produce each other, thinking in patterns shaped by tool outputs—efficiency that is also closure, the narrowing disguised as expansion.
Contingency is the anti-enclosure mechanism. Genuine novelty—patterns not derivable from system state—requires sources external to the recursive loop; cosmotechnical diversity provides those sources.
The creativity paradox. AI expands what individuals can build while contracting the space of buildable things—more output within a narrowing framework, volume mistaken for variety.
Closure is structural, not intentional. No conspiracy required—mathematical optimization plus economic incentives plus infrastructure lock-in produce convergence automatically, continuously, invisibly.