Supervenience
In philosophy, supervenience is a dependency relation between two sets of properties: the subvenient properties (typically physical or lower-level) and the supervenient properties (typically mental, biological, or higher-level). The core idea is that there can be no difference in the supervenient properties without some difference in the subvenient properties. Mental states supervene on brain states: if two individuals are physically identical in every relevant respect, they cannot differ in their mental states.
Supervenience is often invoked as a way of capturing the dependence of higher-level phenomena on lower-level phenomena without committing to reductionism. It allows that mental properties are genuinely distinct from physical properties while insisting that they are not free-floating: they are anchored to the physical base.
The concept was introduced into philosophy of mind by Donald Davidson in 1970 and has since been applied to debates about the relationship between chemistry and physics, biology and chemistry, and social facts and individual behavior. In the context of emergence, supervenience captures the intuition that emergent properties are novel but not autonomous: they depend on their base without being reducible to it.
Supervenience and Systems Theory
The philosophical concept of supervenience has found unexpected resonance in systems theory, where it describes a structural feature of multi-level systems that is neither reductionist nor dualist. In a complex adaptive system, global properties — market prices, ecosystem stability, organizational culture — supervene on local interactions: no two systems with identical local agent configurations can differ in their global properties, yet the global properties are not merely summaries of the local ones. The supervenience relation is what makes the system intelligible at multiple scales without requiring a single master equation.
This multi-scale supervenience is the formal warrant for effective information and causal emergence. When the macroscopic description of a system has more causal power than the microscopic description, the macro-property is supervenient in a strong sense: it depends on the micro-level but is not replaceable by it. The emergence of causal power at the macro level is compatible with supervenience precisely because supervenience does not demand reducibility. A system can be fully determined by its micro-physics and yet require macro-level interventions to control it effectively.
Network theory provides a complementary perspective. In a social network, the property of being influential supervenes on the pattern of connections, not on any single connection. Remove one edge, and the influence may persist; rewire the topology, and it collapses. This is structural supervenience: a property that depends on the arrangement of components rather than on the components themselves. The concept has been applied to understanding why systemic risk in financial networks is not reducible to the risk profiles of individual institutions: the same set of banks, arranged differently, produces different systemic properties.
Varieties and Their Limits
Philosophers distinguish several varieties of supervenience, and the distinctions matter for systems theory. Local supervenience holds that two objects with identical local properties cannot differ in supervenient properties. Global supervenience holds that two entire worlds with identical global distributions of subvenient properties cannot differ in supervenient properties. Most systems of interest exhibit regional supervenience: properties depend on the state of a neighborhood, not on the entire system or on a single point.
Regional supervenience is the natural framework for cellular automata and agent-based models. In Conway's Game of Life, the property of a glider being alive supervenes on the local neighborhood of each cell, not on any single cell. The glider's trajectory is a supervenient property that is computationally irreducible from the local rules — the hallmark of weak emergence that is nonetheless robust and predictable. The limit of supervenience appears when the region becomes the entire system: in systems with long-range correlations or small-world topology, local supervenience fails because distant nodes are causally relevant to local outcomes.
The most pressing critique of supervenience comes from systems that exhibit path dependence. If a system's history matters, then two systems with identical current micro-states may have different supervenient properties because they arrived at those states through different trajectories. Supervenience, as standardly defined, is a synchronic relation: it holds at a single time slice. But most complex systems are diachronic: their current state encodes their history. This suggests that supervenience is either an idealization — a useful fiction that ignores time — or it needs to be extended to include temporal indices. A system supervenes on its history as well as its present state, and the supervenience relation itself evolves.
Supervenience, Causal Power, and the Observer
The central tension in supervenience theory is whether supervenient properties can be causally efficacious. If mental properties supervene on physical properties, and the physical properties are causally closed, then mental properties seem to be epiphenomenal — present but powerless. This is the causal exclusion problem, and it has an analog in systems theory: if market prices supervene on individual trades, and individual trades are causally closed, are market prices causally powerless?
The systems-theoretic answer is that causal closure is the wrong framework. In a complex adaptive system, there is no single level of causal closure. The micro-level is not causally closed because the agents are responding to macro-level signals; the macro-level is not causally closed because it is constituted by micro-level interactions. The causal structure is reciprocal, not unidirectional. This is the insight of feedback theory: causes loop, and the loops are real. Supervenience does not exclude causation; it describes the dependence structure that makes looping causation possible.
The role of the observer complicates this further. In observer-indexed emergence, the supervenient property is not merely dependent on the subvenient base; it is dependent on the observer's coarse-graining. A property that supervenes for one observer — a macro-economist who sees GDP — may not supervene for another — a micro-economist who sees individual transactions. Supervenience becomes observer-relative: not a metaphysical fact about the world, but a relational fact about the coupling between a system and a description of it. This does not make supervenience subjective; it makes it instrumental: a property of the system-observer interface that is as objective as any other physical relation, but that cannot be understood without reference to both relata.
The standard philosophical treatment of supervenience treats it as a static, metaphysical dependency relation. But in systems theory, supervenience is a dynamical, operational phenomenon. It is not merely that there can be no difference without a difference; it is that the relevant differences are themselves selected by the dynamics of the system. Supervenience is not a constraint on possibility; it is a signature of how complex systems organize themselves into levels. The mistake is to ask whether supervenience is 'real' or 'merely conceptual.' The right question is: for which systems, at which scales, under which observational regimes, does the supervenience relation stabilize — and what happens when it breaks?