Emergence

Working Definition

Emergence is the condition in which a system composed of many parts exhibits properties, regularities, or behaviors that cannot be read off from a description of the isolated parts and their local interactions, even though those properties do not exist without the parts. In its weak version (weak emergence), “cannot be read off” means “cannot be derived without step-by-step simulation,” which reflects a computational limit of the observer rather than a gap in physical law. In its strong version (strong emergence), the whole acquires genuinely new causal powers that act downward on the parts. Colloquially, the word is used for any surprise summed up as “the whole is more than the sum of its parts”; technically, it requires specifying which levels of description, which inter-level relations, and which kind of non-derivability are at stake.

Genesis and Global Lineage

The lineage begins with John Stuart Mill (1843), who distinguished “homopathic” (additive) laws from “heteropathic” ones such as chemical reactions; George Henry Lewes coined the term “emergent” in 1875, and the British emergentists Samuel Alexander (1920) and C. D. Broad (1925) developed it into a doctrine of holistic properties that could not be deduced even by an omniscient calculator with complete knowledge of the parts. The concept receded with the mid-twentieth-century triumphs of quantum chemistry and molecular biology, only to return in 1972 in the condensed-matter physicist Philip W. Anderson’s Science essay More Is Different: spontaneous symmetry breaking in many-body systems means that each rung in the hierarchy of sciences requires its own fundamental principles. From the 1980s onward the concept was adopted by complexity science at the Santa Fe Institute (Holland, Kauffman), while in the philosophy of mind Jaegwon Kim in the 1990s made it the central target of his causal exclusion argument. In parallel, the word entered management and popular-science jargon, where it usually means little more than “unplanned but desirable.”

Central European Lineage

Two thinkers formed in the Austro-Hungarian and interwar Central European academic world gave the concept a shape that predates Anderson’s 1972 manifesto. Michael Polanyi—who earned his medical degree at the University of Budapest in 1914, completed his physical chemistry dissertation in 1916–1917 (doctorate 1919), and emigrated from Hungary after the tumultuous revolutions of 1919—formulated in The Tacit Dimension (1966) and his Science article Life’s Irreducible Structure (1968) a model of “dual control”: a higher level, such as a DNA sequence, the engineering design of a machine, or a living organism, imposes boundary conditions on the laws of physics and chemistry, which it never violates yet cannot be derived from. Ludwig von Bertalanffy, educated in interwar Vienna (doctorate 1926, habilitation 1934, professorship at the University of Vienna 1940; he left the city only in 1948), argued in Zu einer allgemeinen Systemlehre (1945) and An Outline of General System Theory (1950) that the holistic regularities of open systems are objects of exact science rather than vitalist metaphysics—an earlier statement of a kindred idea in the idiom of theoretical biology, without documented influence on Anderson’s physics. Both links rest on established biographical and bibliographic records, while primary-text verification of the paywalled 1950 and 1968 journal papers remains pending.

Why Now

Since 2018, emergence has migrated from philosophy seminars into engineering and regulatory reports. In 2022 a Google Brain team published Emergent Abilities of Large Language Models (Wei et al.), arguing that capabilities such as multi-step arithmetic and in-context reasoning appear abruptly once models cross a scale threshold; in 2023 Schaeffer, Miranda, and Koyejo (Stanford) replied with Are Emergent Abilities of Large Language Models a Mirage?, demonstrating that sharp “jumps” are largely artifacts of nonlinear, all-or-nothing evaluation metrics. That technical controversy carries direct regulatory stakes: it determines whether policymakers should treat larger frontier models as qualitatively unpredictable (the premise behind compute thresholds in the 2024 EU AI Act) or as smooth extrapolations. A second driver is the post-2020 polycrisis, where couplings across climate, finance, and geopolitics generate macro-dynamics absent from any single subsystem. A third is neuroscience, where whole-brain recordings and competing theories of consciousness (from integrated information theory to predictive processing) clash explicitly over whether consciousness is weakly emergent, strongly emergent, or not emergent at all.

Explanatory Power

The concept makes clear that the question “what is it made of?” never replaces the question “how is it organized?” First, traffic flow: a phantom traffic jam forms without an accident or a lane closure, propagating backward as a compression wave driven by drivers’ reaction times; no individual driver “contains” the jam, yet it can be modeled and dissipated at the macro-level. Second, financial bubbles: asset prices that detach from fundamentals are a property of a network of mutual expectations rather than the sum of isolated valuations, which is why micro-prudential supervision focused on individual bank balance sheets misses systemic risk. Third, large language models: when a new capability appears “suddenly” at scale, the concept forces us to ask whether we are witnessing a phase transition in internal representations (such as grokking) or an artifact of how we score the output. Across all three cases, emergence disciplines analysis by forcing us to name the level at which a phenomenon exists and the instruments capable of capturing it there.

Axes of Debate

Underappreciated Angle

Emergence is routinely invoked as an alibi to evade institutional accountability. If a phantom traffic jam, a financial crash, a discriminatory algorithm, or a toxic corporate culture is “emergent,” then no individual intended it—and therefore, the story goes, no one is responsible. Yet Polanyi’s tradition points to the exact opposite conclusion: higher-level patterns exist only because of boundary conditions that someone set—the rules of the market, the architecture of a social platform, speed limits, or the objective function used to train a model. Emergence does not abolish agency; it shifts agency from the level of individual events to the level of constraint design. The practical lesson for institutional and technological analysis is direct: instead of asking who caused a collective outcome, ask who designed the boundary conditions under which that outcome became the natural attractor.