The World of SPLectrum
Home > Positioning > Subjects > Complex Adaptive Systems > Adjacent traditions
Adjacent traditions
The intellectual neighbours of CAS as present-day traditions, not as historical feeders. The formation and roots page covers the genealogical story — what flowed into CAS during its formation. This page maps the current boundaries: where they are sharp, where they are fuzzy, and where two traditions share territory without agreeing on what it means.
Cybernetics
Sharp boundary: CAS centres open-ended adaptation — agents changing strategies, populations exploring new regions of possibility space. First-order cybernetics centres homeostasis — feedback loops maintaining a target state. The orientations are opposite: CAS studies how systems change; cybernetics studies how systems stay the same.
Fuzzy boundary: Second-order cybernetics (von Foerster, Maturana, Varela) overlaps with CAS emergentism. The observer enters the system; the system’s organisation is self-produced. Maturana and Varela’s autopoiesis — the claim that living systems produce their own components and thereby maintain their organisation — is a parallel development that shares CAS’s emphasis on self-production without sharing its agent-and-population framework. The two traditions read each other but do not merge: autopoietic theory is about operational closure; CAS is about adaptive openness.
General systems theory
Sharp boundary: CAS specifies its primitive — the adaptive agent. General systems theory (von Bertalanffy) sought universal principles across domains without specifying what the system is made of. The generality is the point for GST; the specificity is the point for CAS.
Fuzzy boundary: The cross-disciplinary ambition is shared. Systems thinking (Peter Senge’s The Fifth Discipline, Donella Meadows’ Thinking in Systems) and system dynamics (Jay Forrester’s stock-and-flow models) sit in the neighbourhood — they study feedback-driven dynamics across domains, often without the adaptive-agent commitment. Some practitioners mix CAS and systems-dynamics vocabulary freely; others maintain the distinction.
Chaos and nonlinear dynamics
Sharp boundary: Chaos describes deterministic systems with sensitive dependence on initial conditions — the system’s rules are fixed; the surprise is in how those fixed rules produce unpredictable trajectories. CAS describes populations of agents whose rules change through interaction. The distinction is in adaptation: chaotic systems don’t learn; CAS agents do.
Fuzzy boundary: CAS systems often exhibit chaotic dynamics at the micro level — individual agent trajectories may be sensitive to initial conditions even as macro-level patterns are robust. The mathematical toolkit overlaps: strange attractors, Lyapunov exponents, bifurcation analysis all appear in CAS work. Popular treatments blur the traditions routinely — “chaos and complexity” is treated as a single subject in bookshops and media. The “complexity” umbrella that groups both under one label doesn’t help the disambiguation.
Cellular automata
A peer rather than a predecessor. Wolfram’s classification of cellular automaton behaviour; Conway’s Game of Life; von Neumann’s self-reproducing automata.
Sharp boundary: Cellular automata have rules, not strategies. The cells do not adapt, learn, or adjust — they execute the same rule at every step. The richness comes from the interaction topology, not from the components’ internal dynamics. CAS agents have internal states that change through experience; CA cells do not.
Fuzzy boundary: Artificial life bridges the two. Langton’s work at SFI treated cellular automata as the substrate for CAS-like dynamics — simple local rules producing lifelike behaviour. When CA rules are allowed to evolve (as in genetic programming applied to CA rule spaces), the boundary with CAS dissolves.
Evolutionary theory
Ancestral rather than adjacent. Darwin, population genetics, the modern synthesis — CAS draws on variation-and-selection as its core adaptive mechanism. The relationship is not one of neighbourhood but of inheritance.
Where CAS extends evolutionary dynamics: into cultural evolution (memes, social learning, norm transmission), technological evolution (Arthur’s increasing returns, innovation as recombination), and market evolution (competing strategies in economic populations). These extensions generalise the variation-and-selection mechanism beyond biology and into domains where the “agents” are firms, ideas, or technologies rather than organisms.
Where CAS stays close to the biological original: Kauffman’s work on gene regulatory networks, NK fitness landscapes, and the order-for-free programme operate directly within evolutionary biology, contesting the sufficiency of selection as the sole source of biological order.
Network science
A substantial peer. Barabási’s scale-free networks, Watts and Strogatz’s small-world networks, Newman’s community detection — network science developed its own identity in the late 1990s and 2000s, partly out of CAS and partly out of graph theory and statistical physics.
Sharp boundary: Network science foregrounds structure — degree distributions, clustering coefficients, path lengths, community structure. CAS foregrounds adaptation — agents changing strategies, populations evolving. A network scientist asks what the topology looks like; a CAS researcher asks what the agents do on it.
Fuzzy boundary: Adaptive networks — networks whose topology coevolves with the dynamics running on them — dissolve the distinction. Much CAS work uses network analysis as a tool; much network science studies adaptive phenomena. The two traditions share conferences, journals, and personnel.
Statistical mechanics
The physics tradition that informs CAS most directly. Philip Anderson’s “More Is Different” (1972) argued that each level of organisation has its own laws, not derivable from the level below — a founding statement for the view that emergence is real and not merely apparent. Anderson was an SFI co-founder; statistical mechanics supplied CAS with phase transitions, scaling laws, mean-field approximations, and the renormalisation group as conceptual tools.
Sharp boundary: Statistical mechanics studies systems of identical or near-identical components (gas molecules, spins on a lattice). CAS studies heterogeneous adaptive agents. The mathematics transfers; the ontology does not.
Fuzzy boundary: Econophysics applies statistical-mechanics methods directly to financial markets — treating price movements as phase transitions, volatility clustering as critical phenomena. Some of this work sits comfortably within CAS; some treats the economy as a physical system without the adaptive-agent commitment.