What a new definition of complex systems clarifies about neurodivergence, deficit, and difference
Dave Snowden has published a definition of a complex system (Snowden, 2026). It is worth reading on its own terms. It is careful to be a definition rather than an explanation — rarer than it sounds — and it makes a set of commitments most accounts of complexity avoid: a realist one about where complexity lives, a constraint-based one about where the boundary between complexity and order falls, and a final one about the irreducibility of human judgment. I want to take it seriously, in the way a definition asks to be taken: by carrying it into a domain and seeing what it does. The domain I work in is human neurodevelopment, and the definition does a great deal there — including making a claim about autism that Snowden does not make, but that follows more cleanly from his architecture than from any framework currently used in the field.
Let me start with the part of the definition that turns out to matter most.
Three registers
Snowden distinguishes three irreducibly distinct registers: the nature of a system, our knowledge of it, and our perception of it. Knowledge and perception are observer-relative. The nature register is not. A system’s propensities, its path dependencies, and its constraint structure are features of the system itself, not of the relationship between the system and whoever happens to be looking. The work of inquiry is to generate friction between the registers to reduce the divergence between them — except, and this is the clause that matters, where the divergence is itself an irreducible feature of the system’s nature, in which case it must be acknowledged rather than resolved.
Hold onto that last clause.
Assessment as register confusion
Autism is assessed through instruments — the ADOS, the ADI-R, the CAT-Q — that score behavior against a normative expectation. These instruments operate in the perception register: they record how autistic behavior appears to an observer trained to notice departures from a norm. The deficit model treats those scores as the person’s nature. It collapses the perception register into the nature register. A measurement of how difference appears becomes a statement of what difference is.
Snowden’s architecture names this error more exactly than the field’s usual objections do. “The instruments are biased” is true; “the norms are culturally contingent” is true, but the deeper problem is categorical. A perception-register artifact has been promoted to an ontological claim. And the repair is not better instruments. It is the recognition that some of the divergence between how autistic regulation is perceived and what it is may be irreducible — a feature of the nature register, to be acknowledged, not a gap to be closed by normalization.
This is what “difference, not deficit” has been reaching for without a vocabulary precise enough to hold it. Difference-not-deficit is the recognition of irreducible register divergence. The deficit model is an attempt to resolve a divergence that the system’s nature requires acknowledgment of. The slogan turns out to be a claim in the philosophy of complexity, not in the politics of identity — which is why it has been so easy to dismiss it as the latter.
Snowden does not make this move. He works in organizations and sense-making, not neurodevelopment. But the move is his architecture’s, not mine. I am only carrying it into the room where it was needed.
Enabling, not governing
The definition draws the boundary between complexity and order through constraints. Ordered systems operate under governing constraints that determine outcomes. Complex systems run on enabling constraints, which shape what can happen without determining what does — modulating, in his terms, the plausibility landscape of emergence without specifying any outcome. The distinction is consonant with Juarrero’s account of context-sensitive constraints, as Snowden notes.
Trait psychology, and the deficit model that descends from it, is governing-constraint thinking applied to persons. A trait is a fixed quantity that determines behavior; a deficit is a fixed shortfall that determines an outcome. Both treat the person as an ordered system. But regulation — the thing autism assessment is actually circling — is not a quantity. It is an enabling-constraint structure: it shifts the plausibility landscape of a person’s states without determining any of them. What I have called regulatory bandwidth is a name for that structure. What I have called bio-neurotype is not a trait a person has but a description of which constraint coupling dominates. Neither is a parameter; both are architecture. The deficit model fails here for the same reason governing-constraint thinking fails on any complex system: it looks for a determinate cause where there is only a modulated landscape.
The developmental manifold
The definition gives the nature register its content: propensities and dispositions, shaped by path dependency, such that a system’s current state cannot be understood apart from its historical trajectory, and any understanding is retrospective — coherent in hindsight, unavailable as a basis for linear prediction.
This is, very nearly word-for-word, developmental systems theory. Gilbert Gottlieb called it the developmental manifold: phenotype as the emergent product of reciprocal influences across levels, over time, with no predetermined program. Read through the nature register; neurodivergence is not a category a person falls into. It is a path-dependent propensity structure — a trajectory through a manifold, legible only in retrospect, disposing a person toward certain kinds of emergence under perturbation without specifying them. The nonlinear developmental manifold I work with is the object Snowden’s account describes, reached from the other side: he from complexity, I from stress physiology and development. That the two roads arrive at the same structure is not a coincidence. It is the structure asserting itself.
Judgment, and the clinic
The definition’s final section concerns judgment. Where a system’s nature is modulated rather than determined, where the registers cannot be collapsed, where emergence is available only in retrospect, no rule-based procedure can replace the judgment of an accountable agent entangled with the situation — one who acts, in Snowden’s phrase, from within the entanglement rather than above it.
This is the clinical implication, and it is exact. A deficit instrument scores from above: it stands outside the person and the relationship, returning a number. A clinician practicing regulation-informed formulation is entangled — inside the dyad, accountable for what emerges, exercising judgment no algorithm can bear. The shift the field needs is from scoring to formulation, and this is its epistemology. It is also why the move cannot be automated, including by the AI tools many of us now think with: they can pattern-match across prior cases at scale, but they cannot occupy the entanglement or bear the accountability. The judgment stays human.
What it costs
None of this is free. To say the autistic difference lives in the nature register — real, observer-independent architecture — is to make a realist commitment, and to set neurodivergence-as-real-architecture against two more comfortable positions: the deficit model, which makes difference a real shortfall, and the strong social model, which makes difference largely a construction of context and perception. The view here keeps what each gets right and refuses what each gets wrong. The difference is real — against constructionism. It is not a deficit — against the medical model. The deficit is the real thing: a perception-register artifact mistaken for nature. It is a narrow path, and I am choosing to walk it deliberately because it is the only one of the three that lets you say the difference is genuinely there and genuinely not a lack.
The exchange runs both ways
A definition earns its keep by doing work in cases for which it was not built. Snowden built this one in and for organizations and sense-making; it was not built for autism. Carried into neurodevelopment, it does what the field’s own frameworks have struggled to do: it names why the deficit model is wrong without retreating into pure social construction, it distinguishes the real architecture of difference from the observer-relative perception of it, and it gives “difference, not deficit” a spine — the principle that some divergence is to be acknowledged, not resolved. The definition gains a hard test case; neurodevelopment gains a precision it has lacked. That is the only kind of borrowing worth doing.
Acknowledged, not resolved. It turns out to be the whole argument.
References
- Snowden, D. (2026). Defining a Complex System. The Cynefin Co, 29 June 2026. thecynefin.co/defining-a-complex-system
- Juarrero, A. (1999). Dynamics in Action: Intentional Behavior as a Complex System. MIT Press. (See also Context Changes Everything, MIT Press, 2023.)
- Cilliers, P. (1998). Complexity and Postmodernism. Routledge.
- Axelrod, R. & Cohen, M. D. (1999). Harnessing Complexity. Free Press.
- Gottlieb, G. (2007). Probabilistic epigenesis. Developmental Science, 10(1), 1–11.
- Hogenkamp, L., Sanghavi, D., & Natri, H. (2026). Toward an Emergent Paradigm for Neurodiversity and Health›. Autism in Adulthood. doi:10.1177/25739581261433443
Leave a Reply