Emergence is the phenomenon by which properties, patterns, or capacities arise at the level of a system that are absent at the level of its components, and cannot be straightforwardly derived from those components even if their properties and interactions are fully known. It is, in a precise sense, the appearance of genuine novelty through organisation.
The philosophical provenance of the concept is long. John Stuart Mill distinguished between homopathic effects, where the whole resembles its parts, and heteropathic effects, where new properties arise from combination. C.D. Broad, writing in the 1920s, introduced the vocabulary of emergence into analytic philosophy, distinguishing between resultant properties (predictable from components) and emergent properties (irreducible to them). George Henry Lewes coined the term itself in the nineteenth century. What has changed since is not the fundamental intuition but the scientific and mathematical vocabulary for describing it with rigour.
To understand emergence is to understand why the classical reductionist programme, i.e., explain everything by explaining its parts, is always incomplete. It is not that reductionism is wrong; it is that it is insufficient. The properties of water cannot be understood from the quantum mechanics of hydrogen and oxygen atoms alone: the properties of liquid water require knowing that there are many molecules interacting, and how they interact. The economy cannot be understood from individual preferences alone: it requires knowing how agents interact in markets, institutions, and networks. Emergence is the reason that each level of organisation requires its own conceptual vocabulary and its own science.
Philosophy and Formulation
Two broad philosophical positions structure the debate about emergence. The first holds that emergence is epistemological: emergent properties are real and irreducible in practice, because computation or prediction from lower-level descriptions is intractable, but are in principle explicable by the lower-level facts. The second holds that emergence is ontological: some emergent properties are genuinely irreducible even in principle, representing a real addition to the causal fabric of the world.
The physicist Philip Anderson’s landmark 1972 essay “More is Different” gave the epistemological position its most influential scientific formulation. Anderson argued that each level of complexity involves “entirely new laws, concepts, and generalisations,” making the reductionist hypothesis (that everything follows from particle physics) compatible with the practical impossibility of deriving psychology from neuroscience, or economics from physics. The sciences of each level are autonomous disciplines, not mere applied lower-level physics.
More recently, researchers including Erik Hoel have developed information-theoretic approaches to formalise emergence. Hoel’s causal emergence framework uses the measure of effective information, derived from Judea Pearl’s do-calculus, to quantify whether a macro-scale description of a system has greater causal power than a micro-scale description. On this view, emergence is not merely about unpredictability; it is about causal efficacy. A macro-level description may genuinely be more causally predictive than any micro-level description. In a precise technical sense, the whole has more causal power than its parts. This constitutes one of the most significant recent advances in the formalisation of emergence.
Consciousness remains the most contested domain of emergence. The philosopher David Chalmers’ “hard problem,” namely why physical processes give rise to subjective experience, is in essence a question about whether phenomenal consciousness is an emergent property in the strong, ontologically irreducible sense. This question remains open and is one of the deepest in all of philosophy.
Key Features
Irreducibility
Emergent properties cannot be deduced from a full description of the components alone. Something is gained, some property, capacity, or behavioural pattern, in the transition from parts to whole. The whole is, in the relevant sense, more than the sum of its parts.
Downward Causation
Emergent properties do not merely arise from lower-level interactions; they also constrain and influence lower-level behaviour. The existence of a traffic jam (a macro-level emergent pattern) causally influences the behaviour of individual drivers (micro-level agents). The norms of an organisation shape the decisions of its members. The concept of a species constrains what individual organisms can do. This bidirectional causation, upward emergence and downward constraint, is a defining feature of complex adaptive systems.
Unpredictability
Even given complete knowledge of a system’s components and interaction rules, emergent properties may be computationally or practically unpredictable without actually running the system forward in time. This is related to, but distinct from, chaos: not all unpredictable emergent behaviour is chaotic, and not all chaotic behaviour is emergent in the philosophically interesting sense.
Context-Dependence
Emergent properties depend on context: on the size of the system, the density of interaction, the topology of the network, and the rate of information flow. The same components can give rise to radically different emergent properties depending on how they are organised. This is why the question “what emerges?” cannot be answered without specifying the conditions of organisation.
Novelty
Emergence generates genuinely new properties, new not in the sense that they were latent in the components waiting to be expressed, but new in the sense that they come into being through organisation. This is the sense in which emergence is the source of creativity and innovation in both biological evolution and economic life.
Types of Emergence
Weak Emergence
Weakly emergent properties arise from lower-level interactions in a way that is in principle traceable: given sufficient computational resources, one could derive the emergent property from the micro-level description. The emergence is real and practically irreducible (the computation may be intractable), but not ontologically irreducible. Most emergent phenomena studied in complexity science are weakly emergent. Traffic jams, market prices, collective animal behaviour, and most social norms fall into this category.
Strong Emergence
Strongly emergent properties cannot be derived from lower-level descriptions even in principle; they represent genuine ontological additions to the causal structure of the world. Strong emergence is philosophically controversial: critics argue that any apparently strongly emergent property is ultimately weakly emergent, and that strong emergence would require a kind of nomological miracle. Consciousness is the most commonly cited candidate. Some philosophers and scientists take the view that consciousness is indeed strongly emergent; others hold that this merely reflects our current ignorance of the relevant lower-level mechanisms.
Causal Emergence
A formally distinct and more recently proposed category, causal emergence identifies emergence not by the unpredictability of macro-properties but by their causal power relative to micro-level descriptions. A macro-level description is causally emergent if it has greater effective information, i.e., greater capacity to predict the system’s future states, than any micro-level description. This framework is empirically applicable and has been used to identify causally emergent levels in neural systems.
Functional Emergence
A category of practical importance in organisational and engineering contexts: properties that arise from the functional organisation of components rather than their intrinsic properties. A committee has capabilities, deliberative, coordinative, and decision-making, that its individual members lack. A market has price-discovery capacities that no individual participant possesses. The function is emergent from the organisation.
Epistemic Emergence
Sometimes distinguished in philosophical literature: the irreducibility is entirely in the observer’s representational system rather than in the system itself. What appears emergent is a consequence of using the wrong conceptual vocabulary at the micro-level. This is the most deflationary view of emergence and is contested.
Emergence vs Self-Organisation
Emergence and self-organisation are closely related but distinct. Self-organisation is the process by which a system develops structure and order without external direction; emergence is the outcome, i.e., the appearance of properties at a higher level of organisation. Self-organisation typically produces emergence, but emergence does not require self-organisation: a system could exhibit emergent properties even if its structure were externally imposed, provided those properties are not derivable from the component-level description.
In practice, the two concepts are almost always found together in the CAS literature, and for good reason. The most interesting and scientifically tractable examples of emergence, including ant colony foraging, financial market dynamics, the structure of the internet, and cultural norm formation, are also cases of self-organisation: the patterns arise from the local interactions of agents following relatively simple rules, without a designer and without a plan.
The productive tension between the two concepts lies here: self-organisation can be designed for, in the sense that one can create the conditions (the rules, the information flows, the agent diversity, the feedback mechanisms) that make self-organisation more or less likely. But the emergent properties that result cannot be fully specified in advance. This is the fundamental challenge and the fundamental opportunity for managers, engineers, and policymakers working with complex systems: you can cultivate the conditions for emergence, but you cannot blueprint the outcomes.
Emergence Across Domains
Biology
The central narrative of biological organisation is emergence: molecules form cells, cells form tissues, tissues form organs, organs form organisms, organisms form ecosystems. At each level, genuinely novel properties arise. The cell membrane is not a property of its constituent lipids; metabolism is not a property of individual enzymes; immune memory is not a property of individual lymphocytes. Evolution itself is an emergent process: the adaptation of populations arises from the interactions of random mutation, differential reproduction, and selection, none of which are directed toward adaptation as a goal.
Neuroscience and Cognition
Consciousness, perception, memory, and emotion are emergent properties of neural activity. Despite decades of remarkable progress in neuroscience, the relationship between neural firing patterns and phenomenal experience remains one of science’s deepest mysteries. More tractable, but still surprising, is the emergence of cognitive capacities in artificial neural networks: language understanding, reasoning, and creative generation appearing as emergent properties of prediction training at scale.
Economics and Finance
Market prices, asset bubbles, financial contagion, and the emergence of dominant platforms are all examples of economic emergence. Perhaps the most striking current example is the emergence of the gig economy, a new economic form that was not designed by any single actor but emerged from the interaction of digital platforms, changing labour preferences, smartphone ubiquity, and regulatory inertia.
Social Systems
Social norms, institutions, languages, and cultures are emergent phenomena: they arise from the interactions of individuals, persist beyond any individual, constrain individual behaviour, and change as the interactions that sustain them change. The emergence of political polarisation in networked media environments is one of the most consequential emergent social phenomena of the current era. No individual or algorithm explicitly designed this pattern; it emerges from the interaction of recommendation algorithms, human psychology, and network structure.
Technology and AI
The emergent capabilities of large language models, including reasoning, cross-domain analogy, and apparent common sense, have surprised even their creators. These capabilities arise not from explicit programming but from the organisation of billions of parameters trained on vast corpora. Understanding which capabilities will emerge at which scales of training, and why, is a central open problem in AI research and a quintessentially complexity-scientific question.
