The World of SPLectrum
Home > Positioning > Subjects > Complex Adaptive Systems > Formation and roots
Formation and roots
How the CAS tradition came together — the intellectual traditions it drew on, the research programs that arrived ready-made, and the institutional moment that gave them a shared language.
Intellectual roots
Cybernetics
The Macy conferences (1946–1953) were the formation moment — mathematicians, engineers, biologists, and social scientists working out the implications of feedback, circular causality, and self-regulation. Norbert Wiener’s Cybernetics (1948) gave the field its name and its central concept: the feedback loop as a universal mechanism of control and communication. W. Ross Ashby’s An Introduction to Cybernetics (1956) formalised the ideas around the homeostat — a device that self-regulates toward stability regardless of perturbation.
First-order cybernetics studied feedback and regulation from an external observer’s position. The second-order turn — von Foerster, Maturana — brought the observer into the system: the act of observation is itself part of the dynamics being observed. CAS inherits the feedback concept; it does not centre the homeostatic emphasis, preferring adaptation and open-ended change over stability-seeking.
General systems theory
Ludwig von Bertalanffy’s General System Theory (1968) proposed that systems across domains — biological, social, physical — share structural principles transferable from one to another. The open-systems framing was the key move: systems exchange matter and energy with their environments, and their organisation is maintained through that exchange rather than in spite of it. CAS shares the cross-disciplinary ambition but is more specific about its primitive — the adaptive agent, not the generic system.
Chaos and nonlinear dynamics
Edward Lorenz’s 1963 paper on deterministic non-periodic flow showed that simple deterministic systems can produce behaviour indistinguishable from randomness — sensitive dependence on initial conditions, the signature of chaos. Ilya Prigogine’s work on dissipative structures (Nobel 1977) showed the other side: order arising far from equilibrium, sustained by the throughflow of energy. Order Out of Chaos (Prigogine and Stengers, 1984 — coincidentally the year of SFI’s founding) synthesised this for a broad audience. CAS shares non-linearity with the chaos tradition; it adds the adaptive agent that changes its strategies through interaction, which chaos does not centre.
Artificial life
A distinct feeder tradition. John von Neumann’s work on self-reproducing cellular automata (1940s–50s, posthumously published as Theory of Self-Reproducing Automata, 1966) posed the question of what formal conditions allow a machine to reproduce itself. Stanislaw Ulam contributed the cellular automaton framework. John Conway’s Game of Life (1970) demonstrated that a handful of local rules can generate structures of extraordinary variety. Christopher Langton organised the first artificial life workshop at Los Alamos in 1987, bridging cellular automata, biology-inspired computation, and the emerging CAS framework. Artificial life and CAS overlap substantially and feed each other at SFI; their pre-histories are distinguishable.
Pre-SFI programs
Holland at Michigan
John Holland began developing genetic algorithms in the early 1960s at the University of Michigan. Adaptation in Natural and Artificial Systems (1975) was the founding text: populations of candidate solutions evolving through selection, crossover, and mutation, with the schema theorem providing a mathematical account of why the process works. The BACH group (Burks, Axelrod, Cohen, Holland) at Michigan formed the early intellectual community. Classifier systems extended the framework toward learning agents — rule-based systems that adapt through experience. Holland arrived at SFI with a developed framework and a generation of students.
Kauffman’s autocatalysis and order-for-free
Stuart Kauffman’s 1969 paper on gene regulatory networks proposed that random Boolean networks exhibit ordered behaviour under certain conditions — the system self-organises without selection. The suggestion was radical: biology lives at the edge of chaos, and network topology generates order independently of natural selection. The shift toward “order for free” put Kauffman in tension with the neo-Darwinian mainstream, which centres selection as the source of biological order. The Origins of Order (1993) consolidated the programme: NK fitness landscapes, autocatalytic sets, the conditions under which self-organisation and selection interact. Kauffman arrived at SFI with a programme already in productive conflict with the prevailing evolutionary synthesis.
The founding moment
SFI emerged partly from Los Alamos veterans who wanted an institute for cross-disciplinary work that the national labs could not house. George Cowan, a nuclear chemist and Los Alamos administrator, drove the early organisation. The founding circle included Murray Gell-Mann (particle physics, Nobel 1969), Philip Anderson (condensed-matter physics, Nobel 1977), and Kenneth Arrow (economics, Nobel 1972). The cross-disciplinary brief was deliberate and architectural — not a feature added to a discipline but the organising principle of the institution.
The early workshops shaped what CAS would become. David Pines ran condensed-matter-and-biology workshops exploring shared structures across physics and the life sciences. Brian Arthur brought economists into conversation with physicists and biologists, developing the complexity-economics programme that would challenge equilibrium assumptions. Langton’s artificial life conferences (from 1987) gave CAS a computational wing. Early funding came from Citibank and the MacArthur Foundation, among others — institutional backers willing to fund cross-disciplinary work without an immediate disciplinary home.
What CAS added
The traditions above had pieces. Cybernetics had feedback. General systems theory had cross-disciplinary universality. Chaos had non-linearity and sensitive dependence. Evolutionary biology had variation and selection. None centred an agent that learns, adjusts internal models, and reshapes the landscape it moves through. Holland’s contribution was to treat the adaptive unit as the primitive object of study — not deriving its behaviour from below (as physics would) or cataloguing it from above (as natural history would), but modelling it as a component that interacts, learns, and changes. This move organised what CAS would become.