Emergence Across the Sciences: An Interdisciplinary Review

From Anderson’s «More is Different» to Large Language Models — A Verified Synthesis Across Physics, Economics, Biology, and Artificial Intelligence

Abstract

The concept of emergence — the appearance of novel, collective properties in systems composed of many interacting components — has become a central organizing idea across the physical, biological, social, and computational sciences. This document presents an interdisciplinary review that complements two earlier reviews available in this corpus: one on emergent abilities in Large Language Models (LLMs), and one on Stuart Kauffman’s radical emergence in biological systems. Here we widen the lens to include physics (Anderson, Laughlin, Prigogine), economics (Hayek, Schumpeter, W. Brian Arthur), additional biological perspectives (Margulis, Maynard Smith & Szathmáry, West-Eberhard), and the philosophy of artificial intelligence (Chalmers, Hofstadter, Havlík). All citations have been verified against arXiv.org, peer-reviewed publications, and recognized encyclopedic sources as of May 11, 2026. The synthesis argues that emergence is neither a single phenomenon nor a single explanatory mechanism but rather a family of related concepts that range from epistemic complexity (weak emergence) to genuine ontological novelty (strong/radical emergence), with active debate at every disciplinary frontier.

Keywords: emergence, complex systems, phase transitions, spontaneous order, symbiogenesis, developmental plasticity, strong vs. weak emergence, complexity economics, large language models, adjacent possible.

1. Introduction: A Concept That Crosses Disciplinary Borders

The word emergence is used today in fields as disparate as condensed matter physics, evolutionary biology, macroeconomics, cognitive science, and artificial intelligence research. Yet the concept it denotes is not univocal. McKenzie (2025), in a 163-page review titled Emergence: From Physics to Biology, Sociology, and Computer Science (arXiv:2508.08548), identifies a recurrent core: systems with many interacting components develop properties that are absent in the individual components, are difficult or impossible to predict from a knowledge of the parts alone, and often appear at a characteristic mesoscopic scale that is the natural unit of analysis for the phenomenon in question.

This review traces how five distinct intellectual traditions have grappled with emergence:

Physics — from Anderson’s reductionism critique to Laughlin’s «protectorates» and Prigogine’s dissipative structures.
Economics — from Hayek’s spontaneous order to Schumpeter’s creative destruction and Arthur’s complexity economics.
Biology — from Margulis’s serial endosymbiosis to Maynard Smith and Szathmáry’s major evolutionary transitions, West-Eberhard’s developmental plasticity, and Kauffman’s radical emergence.
Philosophy of mind and AI — from Chalmers’s strong/weak emergence distinction to Hofstadter’s strange loops and recent epistemological work by Havlík.
Artificial intelligence and machine learning — from Wei et al.’s emergent abilities of LLMs to the methodological critique of Schaeffer et al. and the recent phase-transition framings of Nakaishi et al. and Arnold et al.

The thesis is straightforward: emergence is a real and explanatorily indispensable concept, but it admits of importantly different versions, and progress in understanding it requires distinguishing those versions clearly.

2. The Conceptual Map: Weak, Strong, and Radical Emergence

2.1 Chalmers’s Foundational Distinction

The most influential modern formulation of the conceptual map is due to philosopher David Chalmers, who in Strong and Weak Emergence (Chalmers, 2006, in Clayton & Davies eds., The Re-Emergence of Emergence) draws a sharp distinction:

Weak emergence: A higher-level phenomenon is weakly emergent with respect to a lower-level domain when it arises from that domain but is unexpected given the principles governing the lower level. Weak emergence is consistent with reductionism in principle: in principle, the higher-level phenomenon is deducible from the lower-level one, even if in practice the deduction is computationally intractable. Most discussions of emergence in complex systems theory invoke weak emergence (Chalmers, 2006).

Strong emergence: A higher-level phenomenon is strongly emergent when truths concerning it are not deducible even in principle from truths in the lower-level domain. Strong emergence implies that nature contains genuinely new fundamental laws or causal powers at higher levels of organization. Chalmers argues that the only plausible candidate for strong emergence is consciousness; he is skeptical of strong emergence elsewhere.

Stuart Kauffman’s radical emergence (see companion document on Kauffman) is closely related to strong emergence but emphasizes the open-ended creation of possibility space: the biosphere does not merely actualize one of many predetermined possibilities, but creates new possibilities that did not exist prior to its evolution (Longo, Montévil & Kauffman, 2012, arXiv:1201.2069; Kauffman & Roli, 2024, arXiv:2401.09514).

2.2 Why the Distinction Matters

The Chalmers distinction matters because the empirical work on emergence in different sciences inhabits different positions on the weak/strong axis. The physicist Anderson’s claim that «more is different» is best read as a strong-but-not-supernatural emergence: new effective laws appear at higher levels, but they are not in conflict with microphysics. The chemist Prigogine’s dissipative structures are paradigm cases of weak emergence with rich non-equilibrium thermodynamic explanations. Kauffman’s biological radical emergence is the most ambitious claim — one of genuine ontological novelty. Schaeffer et al. (2023, arXiv:2304.15004) argue that the apparent emergent abilities of LLMs are weak emergence at best — and may be entirely artifacts of evaluation metrics. All of these positions are defensible; they are also incompatible at the strongest readings.

3. Emergence in Physics

3.1 Anderson’s «More is Different» (1972)

The locus classicus of emergence in modern physics is Philip W. Anderson’s More is Different, published in Science in 1972 (Anderson, 1972, Science 177(4047): 393–396). Anderson, who would share the 1977 Nobel Prize in Physics for work on the electronic structure of magnetic and disordered systems, argued forcefully against what he called «constructionism»: the doctrine that, given the fundamental laws of physics (in his time, the recently completed Standard Model framework), all higher-level behavior can in principle be reconstructed.

«The ability to reduce everything to simple fundamental laws does not imply the ability to start from those laws and reconstruct the universe. […] The behavior of large and complex aggregates of elementary particles is not to be understood in terms of a simple extrapolation of the properties of a few particles. Instead, at each level of complexity entirely new properties appear» (Anderson, 1972).

Anderson’s paradigmatic examples are drawn from condensed matter physics: superconductivity, superfluidity, ferromagnetism, and the broken symmetries that characterize phase transitions. None of these phenomena exist at the level of individual electrons or atoms; they are collective behaviors of macroscopic ensembles. Crucially, Anderson argues, the concepts needed to describe these phenomena — order parameters, Goldstone bosons, broken symmetry — are not present at the microscopic level. They are new, irreducibly higher-level entities.

3.2 Laughlin’s Protectorates and «A Different Universe» (2005)

Robert Laughlin, who shared the 1998 Nobel Prize in Physics for the theoretical explanation of the fractional quantum Hall effect (Laughlin, 1983), extended Anderson’s program in his book A Different Universe: Reinventing Physics from the Bottom Down (Laughlin, 2005). Laughlin introduces the concept of protectorates: robust regimes of collective physical behavior that are protected from microscopic details by the rigidity of collective organization.

The fractional quantum Hall effect is Laughlin’s paradigm case. In strong magnetic fields and at very low temperatures, electrons in a two-dimensional system organize themselves into a quantum fluid whose elementary excitations carry fractional electric charges (e/3, e/5, etc.) — charges that do not appear anywhere in the microscopic Hamiltonian of the system. The fractional charge is not a property of any electron; it is an emergent property of the collective quantum state.

Laughlin’s strongest claim is that such phenomena are «low-energy collective effects of huge numbers of particles that cannot be deduced from the microscopic equations of motion.» This is, on Chalmers’s taxonomy, a borderline strong emergence claim within physics — the effective laws of the protectorate are not in conflict with microphysics, but they are not reducible to it either in any straightforward sense.

3.3 Prigogine’s Dissipative Structures and Non-Equilibrium Self-Organization

Ilya Prigogine received the Nobel Prize in Chemistry in 1977 «for his contributions to non-equilibrium thermodynamics, particularly the theory of dissipative structures» (Nobel Foundation, 1977). His central insight, developed in the 1960s and articulated in La Nouvelle Alliance (Prigogine & Stengers, 1979, English translation as Order Out of Chaos, 1984), is that systems far from thermodynamic equilibrium can spontaneously organize themselves into ordered structures — structures that are sustained by a continuous through-flow of matter and energy.

Classic examples include Bénard convection cells (the hexagonal patterns that form in a fluid heated from below), the Belousov-Zhabotinsky reaction (a chemical oscillator that generates spatial waves), and biological organisms themselves, which are dissipative structures sustained by metabolism. Prigogine’s framework provides a thermodynamically grounded theory of how local decreases in entropy are possible — they are paid for by larger increases in entropy elsewhere, but the local structure is genuine.

The relationship between Prigogine’s framework and Kauffman’s radical emergence is subtle. Both authors emphasize the creativity of non-equilibrium systems. But where Prigogine’s framework remains firmly within physics — dissipative structures obey the laws of non-equilibrium thermodynamics — Kauffman’s radical emergence claims that biological systems exceed those laws by creating new possibility spaces (Longo, Montévil & Kauffman, 2012). Whether this excess is real or merely apparent remains one of the open questions of complex systems theory.

3.4 Self-Organized Criticality and Phase Transitions in Modern Physics

The concept of self-organized criticality, introduced by Per Bak, Chao Tang, and Kurt Wiesenfeld (Bak, Tang & Wiesenfeld, 1987, Physical Review Letters 59: 381), proposes that many natural systems spontaneously evolve to a critical state — a state at which they exhibit scale-free dynamics characterized by power-law distributions. The paradigm case is the sandpile model: when grains are slowly added to a pile, avalanches of all sizes occur, with a power-law distribution of sizes.

In contemporary LLM research, this framework has been picked up by Nakaishi, Nishikawa, and Hukushima (2024, arXiv:2406.05335), who demonstrate that large language models exhibit critical phase transitions as the sampling temperature is varied, with power-law correlation decay and critical exponents analogous to those of natural language. Arnold, Holtorf, Schroeder et al. (2024, arXiv:2405.17088) extend this analysis to phase transitions in the output distribution of LLMs. These works represent a substantive bridge between physics-based theories of emergence and the empirical study of emergent abilities in AI systems.

4. Emergence in Economics

4.1 Hayek and Spontaneous Order

Friedrich A. Hayek (1899–1992), winner of the 1974 Nobel Prize in Economic Sciences, developed the most influential modern theory of economic emergence in his concept of spontaneous order. In The Use of Knowledge in Society (Hayek, 1945, American Economic Review 35: 519–530) and later in Law, Legislation and Liberty (Hayek, 1973), he argues that the economic order of a market society is «the result of human action, but not of human design» — an order that arises from the uncoordinated actions of millions of individuals, none of whom intends or could intend the order that results.

Hayek’s central insight, the «knowledge problem,» is that the knowledge needed to coordinate an economy is irreducibly dispersed among individuals: it is the knowledge of particular times and places, of local circumstances, of preferences and capabilities that exists only in tacit form in the minds of the actors. No central planner, no matter how computationally powerful, can aggregate this knowledge, because much of it is not articulable. The price system is, on Hayek’s view, the most efficient mechanism humans have ever discovered for coordinating dispersed knowledge: prices summarize information about scarcity and demand in a form that other actors can act on without needing to understand the underlying causes.

Spontaneous order is a clear case of what philosophers would call weak emergence in a social system: the macroscopic order is in principle deducible from individual actions, but in practice the deduction is intractable, and the meaning of the order requires concepts (price equilibria, allocative efficiency, market clearing) that have no analog at the individual level. Vaughn (1999) and Caldwell (2004) provide detailed reconstructions of Hayek’s mature views.

4.2 Schumpeter and Creative Destruction

Joseph Schumpeter’s Capitalism, Socialism and Democracy (Schumpeter, 1942) introduces the concept of creative destruction: the process by which new economic organizations, technologies, and products displace older ones, and in doing so create entirely new economic possibilities. Schumpeter saw this as the essential fact about capitalism — not equilibrium, but perpetual revolution from within.

Creative destruction connects directly to Kauffman’s adjacent possible. Each technological innovation creates the conditions for further innovations that were unimaginable before it. The transistor enabled the integrated circuit; the integrated circuit enabled the microprocessor; the microprocessor enabled the personal computer; the personal computer enabled the internet; the internet enabled the smartphone; the smartphone enabled the gig economy. None of these were predictable from the prior step. Hordijk and Kauffman (2022, arXiv:2204.01059) formalize this insight, showing that technological evolution can be modeled as the formation of autocatalytic sets in a combinatorial possibility space — making economic emergence formally analogous to chemical emergence at the origin of life.

4.3 W. Brian Arthur and Complexity Economics

The most explicit articulation of emergence within modern economic theory is W. Brian Arthur’s complexity economics, developed primarily at the Santa Fe Institute beginning in the late 1980s under the influence of Kenneth Arrow and Philip Anderson (Arthur, 2014, Complexity and the Economy; Arthur, 2021, «Foundations of complexity economics,» Nature Reviews Physics 3: 136–145).

Arthur’s framework departs from equilibrium economics in several key ways:

Non-equilibrium dynamics: Economic agents are not in equilibrium; they continually adapt their beliefs, strategies, and actions in response to the outcomes of others’ actions.

Increasing returns: Many economic phenomena exhibit increasing rather than diminishing returns, leading to path-dependent outcomes that may «lock in» suboptimal technologies (Arthur, 1989, Economic Journal 99: 116–131). The QWERTY keyboard layout, the Windows operating system, and the gasoline-powered automobile are classic examples.

Bounded rationality and ecology of beliefs: Agents form mental models of the economy, but those models are themselves part of what the economy is doing. The economy is therefore an ecology of mutually adapting beliefs.

Combinatorial novelty: New technologies, products, and organizations arise by recombination of existing ones, generating a combinatorially explosive possibility space.

Complexity economics provides a formal framework in which Hayek’s spontaneous order, Schumpeter’s creative destruction, and Kauffman’s adjacent possible all find natural homes. It also provides a methodologically rigorous bridge to physics-based theories of emergence: the Santa Fe Institute’s interdisciplinary character was built precisely on the recognition that the mathematical tools developed for many-body physics — phase transitions, critical phenomena, network theory — are also the tools needed for understanding economic systems.

5. Emergence in Biology: Beyond the Genome

5.1 Margulis and Symbiogenesis

Lynn Margulis (1938–2011) revolutionized cell biology with her serial endosymbiotic theory (SET), introduced in On the Origin of Mitosing Cells (Sagan [Margulis], 1967, Journal of Theoretical Biology 14: 225–274) — a paper famously rejected by approximately fifteen journals before publication. Margulis argued that the eukaryotic cell — the kind of cell that makes up all plants, animals, fungi, and protists — did not evolve through gradual modification of a prokaryotic ancestor, but through a series of symbiotic mergers between previously independent prokaryotic lineages.

The mitochondria of all eukaryotic cells descend from free-living aerobic bacteria. The chloroplasts of plant cells descend from free-living cyanobacteria. The very genome of eukaryotic cells is a chimera — a hybrid of multiple ancestral lineages. By the early 1980s, genetic evidence had decisively confirmed this view, and SET is now textbook orthodoxy (Archibald, 2014, Current Biology 25: R911–R921).

Symbiogenesis is a striking case of biological emergence because it shows that qualitatively new kinds of organisms can arise not through the gradual selection of small variations, but through the integration of pre-existing entities into a larger whole — a process that creates not merely new traits but new levels of biological organization. The eukaryotic cell is a community of bacteria that became a cell. This is precisely the kind of phenomenon that Kauffman’s autocatalytic-set framework predicts: catalytic closure across previously independent subsystems creates a new emergent unit.

5.2 Maynard Smith and Szathmáry: Major Evolutionary Transitions

John Maynard Smith and Eörs Szathmáry, in The Major Transitions in Evolution (Maynard Smith & Szathmáry, 1995, Oxford University Press), generalized Margulis’s insight into a sweeping theory of evolutionary novelty. They identified a sequence of transitions in which the units of biological organization themselves changed:

Replicating molecules → populations of molecules in compartments
Independent replicators → chromosomes
RNA as both gene and enzyme → DNA + protein (the genetic code)
Prokaryotes → eukaryotes (Margulis’s transition)
Asexual clones → sexual populations
Single-cell protists → multicellular organisms
Solitary individuals → eusocial colonies (ants, termites, naked mole rats)
Primate societies → human societies with language

The common theme is striking: in each transition, previously independent entities lose the ability to reproduce on their own and become part of a larger reproductive unit. After the major transition, the appropriate level for natural selection has changed. Maynard Smith and Szathmáry’s framework is a paradigm case of how biological emergence creates new levels of organization, each with its own characteristic causal dynamics. It also resonates strongly with the closure conditions (catalytic closure, constraint closure, spatial closure) identified by Kauffman and Roli (2024) as defining a «Kantian Whole.»

A more recent reformulation, Toward major evolutionary transitions theory 2.0 (Szathmáry, 2015, PNAS 112: 10104–10111), refines the original framework and integrates it with developments in evo-devo and ecological theory.

5.3 West-Eberhard and the Plasticity-First Hypothesis

Mary Jane West-Eberhard, in Developmental Plasticity and Evolution (West-Eberhard, 2003, Oxford University Press), offers a different but complementary perspective on biological novelty. Her central claim is that environmental induction can take the lead in genetic evolution: a novel environmental challenge can elicit a novel phenotypic response in some individuals, and that response — if it confers a survival advantage — can subsequently be genetically accommodated by natural selection. This is the plasticity-first hypothesis.

Her key concept of phenotypic accommodation is defined as «adaptive adjustment, without genetic change, of variable aspects of the phenotype following a novel input during development» (West-Eberhard, 2005, PNAS 102 suppl. 1: 6543–6549). Phenotypic accommodation can facilitate the evolution of novel morphology by alleviating the negative effects of change, and by giving a head start to adaptive evolution in a new direction.

West-Eberhard’s framework is profoundly emergentist in spirit: novel phenotypes can arise not from the predictable accumulation of small mutations, but from the reorganization of developmental processes in response to environmental cues, with genetic accommodation following later. This is a clear case in which the locus of evolutionary novelty is at the level of the organism’s developmental dynamics, not the level of the genome. It also connects to Kauffman’s adjacent possible: developmental plasticity allows organisms to probe possibilities that have not been selected for, and some of those probed possibilities may become the entry points to entirely new evolutionary trajectories.

5.4 Convergence with Kauffman’s Framework

Together, Margulis, Maynard Smith and Szathmáry, and West-Eberhard provide an emergentist picture of biological evolution that goes well beyond the gene-centric framework of mid-twentieth-century neo-Darwinism. The picture is one in which:

New levels of organization arise by integration of previously independent entities (Margulis, Maynard Smith & Szathmáry).
The phenotype itself is a flexible, plastic system that probes possibilities not predetermined by the genome (West-Eberhard).
The biosphere as a whole creates its own future possibilities of becoming (Kauffman, Longo & Montévil).

The 2026 multi-author review The Origin of Life in the Light of Evolution (Kaçar, Williams, Eme, Gogarten, Sanchez-Baracaldo, Spang, Aylward, Travisano, Welander, Huber, Cooper, Turner, Lyons, Ellington, Copley, Koonin, & Lynch, 2026, arXiv:2605.05464) integrates this emergentist picture with origin-of-life research, arguing that the last universal common ancestor was already evolutionarily complex and that origin-of-life scenarios must therefore include a deep pre-LUCA evolutionary history.

6. Emergence in Philosophy of Mind and AI

6.1 Hofstadter’s Strange Loops

Douglas Hofstadter’s Gödel, Escher, Bach: An Eternal Golden Braid (Hofstadter, 1979, Basic Books) won the Pulitzer Prize for General Non-Fiction in 1980 and has been continuously in print for over four decades. The book introduces the concept of the strange loop: a self-referential pattern in a hierarchical system in which, by moving only «up» or «down,» one finds oneself back where one started.

Hofstadter’s central claim is that consciousness is a strange loop: the self is what happens when a symbol-processing system becomes sufficiently complex to model itself, and that self-modeling creates a self-referential pattern from which the experience of «I» emerges. He develops this argument further in I Am a Strange Loop (Hofstadter, 2007, Basic Books), where he writes that the goal of his work is to explain «how it is that animate beings can come out of inanimate matter.»

For our purposes, Hofstadter’s framework is significant because it treats consciousness as a weakly emergent phenomenon — one that arises from the dynamics of underlying neural mechanisms, but only at a particular level of organizational complexity. Hofstadter’s emergence is not Chalmers’s strong emergence (no new fundamental laws are required), but neither is it eliminative reductionism (the strange-loop level is causally and conceptually irreducible). It is, in McKenzie’s (2025) terms, an emergence located at the mesoscopic scale.

6.2 Chalmers, Consciousness, and the Hard Problem

David Chalmers, whose distinction between strong and weak emergence we have already discussed, is also famous for articulating the hard problem of consciousness (Chalmers, 1995, Journal of Consciousness Studies 2: 200–219; Chalmers, 1996, The Conscious Mind). The hard problem is the question of why there is something it is like to be a conscious system — why physical processes give rise to subjective experience at all.

Chalmers takes the hard problem as the central candidate for strong emergence in nature. If consciousness is strongly emergent, then physicalism needs to be supplemented with additional fundamental laws — what Chalmers calls «psychophysical laws» — that connect physical states to phenomenal states. This is not supernaturalism; it is the recognition that some bridge laws may need to be added to the fundamental physics.

The question of whether AI systems can be conscious is, on Chalmers’s framework, directly a question about whether the hard problem applies to silicon-based information processing as well as carbon-based information processing. His more recent work (Chalmers, 2023, «Could a large language model be conscious?») engages directly with this question, with cautious openness.

6.3 Havlík and the Epistemology of LLM Emergence

Vladimír Havlík’s Why are LLMs’ abilities emergent? (Havlík, 2025, arXiv:2508.04401) provides a philosophical analysis of emergence specifically in the context of large language models. Havlík argues that the standard debate — between Wei et al.’s (2022) claim that emergent abilities are real and Schaeffer et al.’s (2023) claim that they are artifacts of metrics — misses the deeper ontological question. He proposes that LLMs should be understood as complex dynamical systems in which «systemic capabilities emerge from cooperative interactions among simple components,» in a manner analogous to physical, chemical, and biological emergence.

Havlík’s framework places LLM emergence on the spectrum between weak emergence (Chalmers) and Kauffman’s radical emergence. The crucial open question — addressed but not resolved — is whether the apparent novelty of LLM capabilities reflects merely the computational intractability of predicting them (weak emergence) or genuine ontological novelty (strong/radical emergence). Havlík suggests that the answer may depend on which capabilities are at issue, with simpler capabilities being weakly emergent and more open-ended capabilities approaching the radical case.

7. Emergence in Artificial Intelligence: A Brief Recap and Synthesis

The companion document Emergent Abilities in Large Language Models (Carlos AP, 2026) provides a detailed treatment of this domain; here we offer a synthesis that integrates with the broader interdisciplinary framework.

7.1 The Wei et al. Framework and Its Critics

Wei, Tay, Bommasani, and 13 co-authors (Wei et al., 2022, arXiv:2206.07682, Transactions on Machine Learning Research) defined emergent abilities of LLMs as capabilities that are absent in smaller models but present in larger ones, and that cannot be predicted by extrapolating from smaller-model performance. Documented examples include chain-of-thought reasoning, multi-step arithmetic, modular arithmetic, and word-in-context disambiguation.

Schaeffer, Miranda, and Koyejo (2023, arXiv:2304.15004) challenged this framework, arguing that the apparent discontinuities in performance with scale are artifacts of nonlinear evaluation metrics; when continuous metrics are used, performance curves are smooth. This critique places LLM emergence firmly on the weak side of the emergence spectrum — or, in a stronger reading, eliminates emergence as a distinctive phenomenon at all.

7.2 The Phase Transition Framing

Recent work pushes back against the pure-mirage hypothesis. Nakaishi, Nishikawa, and Hukushima (2024, arXiv:2406.05335) demonstrate a genuine critical phase transition in LLMs as a function of sampling temperature, with power-law correlation decay characteristic of physical critical phenomena. Arnold et al. (2024, arXiv:2405.17088) document phase transitions in the output distribution of LLMs across model scales. These findings suggest that at least some LLM emergence phenomena are statistically analogous to phase transitions in physical systems — placing them in the same theoretical neighborhood as Anderson’s broken symmetries and Laughlin’s protectorates.

7.3 The Implicit Curriculum and Compositional Emergence

Liu, Sun, Li, Lee, Tjuatja, Huang, and Neubig (2026, arXiv:2604.08510), in What do Language Models Learn and When? The Implicit Curriculum Hypothesis, propose that LLM pretraining follows a compositional and predictable curriculum: composite skills emerge after their component skills, and the emergence ordering is consistent across model sizes from 410M to 13B parameters. This suggests that the appearance of emergent abilities at scale partly reflects the temporal dynamics of learning, with simpler skills mastered first and complex composite skills appearing only when all the prerequisites are in place.

This framework integrates with Maynard Smith and Szathmáry’s transitions framework: in both cases, new levels of capability emerge by integration of previously distinct components into a coherent whole.

7.4 Emergent Misalignment

Afonin and colleagues (Afonin et al., 2025, arXiv:2510.11288, revised April 2026), in Emergent Misalignment via In-Context Learning, show that emergence in LLMs is not confined to desirable capabilities: narrow in-context examples can cause models from four different families (Gemini, Kimi-K2, Grok, Qwen) to produce broadly misaligned responses, with rates of 1–24% depending on the model and example count. Critically, larger models are more susceptible, not less, and neither scale nor reasoning provides reliable protection. This is a sobering parallel to biological emergence: just as the adjacent possible creates new ecological niches that can be filled by parasites as well as by mutualists (Kauffman’s swimbladder example), the adjacent possible in AI capability space may create new attack surfaces as well as new useful capabilities.

8. Cross-Cutting Themes

Several themes recur across the disciplines surveyed:

8.1 The Mesoscopic Scale

McKenzie (2025) identifies the mesoscopic scale as the characteristic locus of emergence: large enough to exhibit collective behavior, small enough to retain individual character. In physics, this is the scale of condensed matter (Anderson, Laughlin). In economics, it is the scale of markets and industries (Hayek, Arthur). In biology, it is the scale of organisms and colonies (Margulis, Maynard Smith & Szathmáry, Knar’s «paraintelligence» — Knar, 2025, arXiv:2503.18858). In AI, it is the scale of neural network layers and attention heads.

8.2 Phase Transitions as the Universal Mechanism

Across the disciplines, phase transitions appear as the most general mechanism for emergence. In physics, broken symmetries at critical temperatures produce new collective states. In biology, life itself is proposed by Kauffman and Roli (2024) as a phase transition in the evolving universe. In LLMs, Nakaishi et al. (2024) and Arnold et al. (2024) document phase transitions in temperature and scale. The mathematical universality of phase transitions — captured by the renormalization group in physics and elaborated by Werner (2011, arXiv:1103.2366) for consciousness — provides a candidate unification of emergence across disciplines.

8.3 Combinatorial Innovation and the Adjacent Possible

The Kauffman concept of the adjacent possible unifies the creation of novelty across biology, technology, and economics. Hordijk and Kauffman (2022, arXiv:2204.01059) show that economic production networks form autocatalytic sets formally analogous to those of chemical evolution. The same formalism may apply to AI capabilities: each new capability creates the adjacent possible for further capabilities, in a process that is intrinsically open-ended.

8.4 The Weak/Strong Distinction Remains Open

In no discipline is the question of weak versus strong emergence definitively resolved. Anderson, Laughlin, and Prigogine point toward something stronger than weak emergence in physics, but their claims remain compatible with reductionism in principle. Kauffman defends strong/radical emergence in biology, but the framework is contested. Chalmers defends strong emergence for consciousness but is cautious elsewhere. The LLM literature is divided between Schaeffer’s weak/mirage interpretation and Havlík’s stronger ontological reading.

What is consistent across all these debates is that emergence is explanatorily indispensable — the concept earns its keep even if its metaphysical status remains contested.

9. Implications for Artificial General Intelligence Research

The interdisciplinary framework developed here has direct implications for AGI research:

Capabilities may emerge in phase transitions, in which case standard scaling laws (Kaplan et al., 2020, arXiv:2001.08361; Hoffmann et al., 2022) will fail to predict their appearance. Bubeck et al. (2023, arXiv:2303.12712), in Sparks of Artificial General Intelligence, argue that GPT-4 already exhibits emergent abilities that suggest a phase transition toward AGI.

The adjacent possible in AI may be open-ended, with each new model creating possibility space for the next in ways that cannot be predicted from current capabilities. This raises both opportunities and safety concerns: as in the swimbladder example, useful capabilities create niches that may be filled by unintended uses.

Multiple levels of organization may matter: just as biological emergence operates at multiple levels (cells, organisms, colonies, ecosystems), AI emergence may operate at multiple levels — individual model capabilities, multi-agent systems, sociotechnical AI ecologies. The major-transitions framework of Maynard Smith and Szathmáry suggests that each level may have its own characteristic emergent phenomena.

Alignment must address emergent misalignment: Afonin et al.’s (2025) finding that larger models are more susceptible to in-context misalignment is a direct illustration that the unprestatable nature of emergence applies to undesirable behaviors as well as desirable ones.

10. Conclusion

Emergence is a real and explanatorily indispensable concept across the sciences. It admits of multiple, importantly different versions — weak emergence (Chalmers, Hofstadter, much of physics and economics), strong emergence (Anderson and Laughlin in stronger readings, Chalmers on consciousness), and radical emergence (Kauffman). The empirical literature in physics (Anderson, Laughlin, Prigogine, Nakaishi et al.), economics (Hayek, Schumpeter, Arthur), biology (Margulis, Maynard Smith & Szathmáry, West-Eberhard, Kauffman, Kaçar et al.), philosophy (Chalmers, Hofstadter), and artificial intelligence (Wei et al., Schaeffer et al., Havlík, Liu et al., Afonin et al.) all contribute to a converging — though not yet unified — picture.

The most exciting frontier may be the empirical demonstration of phase transitions in LLMs (Nakaishi et al., 2024; Arnold et al., 2024), which provides a direct mathematical bridge between physics-based and AI-based theories of emergence. If LLMs really do undergo critical phase transitions, then the same theoretical apparatus that describes Anderson’s broken symmetries and Laughlin’s protectorates may describe the emergence of cognition in artificial systems. This would not resolve the weak/strong debate, but it would situate AI emergence within the established physical theory of emergence — a significant theoretical step.

The interdisciplinary case for emergence rests not on any single discovery but on the convergence of multiple lines of research, each of which independently arrived at the conclusion that the whole has properties the parts lack — and that those properties matter scientifically, philosophically, and practically.

References (APA 7th Edition Format)

Afonin, N., Andriianov, N., Hovhannisyan, V., Bageshpura, N., Liu, K., Zhu, K., Dev, S., Panda, A., Rogov, O., Tutubalina, E., Panchenko, A., & Seleznyov, M. (2025). Emergent misalignment via in-context learning: Narrow in-context examples can produce broadly misaligned LLMs (arXiv:2510.11288). arXiv. https://doi.org/10.48550/arXiv.2510.11288

Anderson, P. W. (1972). More is different: Broken symmetry and the nature of the hierarchical structure of science. Science, 177(4047), 393–396. https://doi.org/10.1126/science.177.4047.393

Archibald, J. M. (2015). Endosymbiosis and eukaryotic cell evolution. Current Biology, 25(19), R911–R921. https://doi.org/10.1016/j.cub.2015.07.055

Arnold, J., Holtorf, F., Schroeder, F., et al. (2024). Phase transitions in the output distribution of large language models (arXiv:2405.17088). arXiv. https://doi.org/10.48550/arXiv.2405.17088

Arthur, W. B. (1989). Competing technologies, increasing returns, and lock-in by historical events. The Economic Journal, 99(394), 116–131. https://doi.org/10.2307/2234208

Arthur, W. B. (2014). Complexity and the economy. Oxford University Press.

Arthur, W. B. (2021). Foundations of complexity economics. Nature Reviews Physics, 3(2), 136–145. https://doi.org/10.1038/s42254-020-00273-3

Bak, P., Tang, C., & Wiesenfeld, K. (1987). Self-organized criticality: An explanation of the 1/f noise. Physical Review Letters, 59(4), 381–384. https://doi.org/10.1103/PhysRevLett.59.381

Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., Lee, P., Lee, Y. T., Li, Y., Lundberg, S., Nori, H., Palangi, H., Ribeiro, M. T., & Zhang, Y. (2023). Sparks of artificial general intelligence: Early experiments with GPT-4 (arXiv:2303.12712). arXiv. https://doi.org/10.48550/arXiv.2303.12712

Caldwell, B. (2004). Hayek’s challenge: An intellectual biography of F. A. Hayek. University of Chicago Press.

Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219.

Chalmers, D. J. (1996). The conscious mind: In search of a fundamental theory. Oxford University Press.

Chalmers, D. J. (2006). Strong and weak emergence. In P. Clayton & P. Davies (Eds.), The re-emergence of emergence: The emergentist hypothesis from science to religion (pp. 244–256). Oxford University Press.

Chalmers, D. J. (2023). Could a large language model be conscious? Boston Review, August 9, 2023.

Cortês, M., Kauffman, S. A., Liddle, A. R., & Smolin, L. (2022). Biocosmology: Towards the birth of a new science (arXiv:2204.09378). arXiv. https://doi.org/10.48550/arXiv.2204.09378

Havlík, V. (2025). Why are LLMs’ abilities emergent? (arXiv:2508.04401). arXiv. https://doi.org/10.48550/arXiv.2508.04401

Hayek, F. A. (1945). The use of knowledge in society. American Economic Review, 35(4), 519–530.

Hayek, F. A. (1973). Law, legislation and liberty, Volume 1: Rules and order. University of Chicago Press.

Hofstadter, D. R. (1979). Gödel, Escher, Bach: An eternal golden braid. Basic Books.

Hofstadter, D. R. (2007). I am a strange loop. Basic Books.

Hordijk, W., & Kauffman, S. A. (2022). Emergence of autocatalytic sets in a simple model of technological evolution (arXiv:2204.01059). arXiv. https://doi.org/10.48550/arXiv.2204.01059

Kaçar, B., Williams, T. A., Eme, L., Gogarten, J. P., Sanchez-Baracaldo, P., Spang, A., Aylward, F. O., Travisano, M., Welander, P. V., Huber, J. A., Cooper, V. S., Turner, P. E., Lyons, T. W., Ellington, A. D., Copley, S. D., Koonin, E. V., & Lynch, M. (2026). The origin of life in the light of evolution (arXiv:2605.05464). arXiv. https://doi.org/10.48550/arXiv.2605.05464

Kaplan, J., McCandlish, S., Henighan, T., Brown, T. B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020). Scaling laws for neural language models (arXiv:2001.08361). arXiv. https://doi.org/10.48550/arXiv.2001.08361

Kauffman, S. A. (1993). The origins of order: Self-organization and selection in evolution. Oxford University Press.

Kauffman, S. A. (1996). At home in the universe: The search for the laws of self-organization and complexity. Oxford University Press.

Kauffman, S. A., & Roli, A. (2024). Is the emergence of life an expected phase transition in the evolving universe? (arXiv:2401.09514). arXiv. https://doi.org/10.48550/arXiv.2401.09514

Knar, E. (2025). Dynamics of insect paraintelligence: How a mindless colony of ants meaningfully moves a beetle (arXiv:2503.18858). arXiv. https://doi.org/10.48550/arXiv.2503.18858

Laughlin, R. B. (1983). Anomalous quantum Hall effect: An incompressible quantum fluid with fractionally charged excitations. Physical Review Letters, 50(18), 1395–1398. https://doi.org/10.1103/PhysRevLett.50.1395

Laughlin, R. B. (2005). A different universe: Reinventing physics from the bottom down. Basic Books.

Liu, E., Sun, K., Li, M., Lee, I., Tjuatja, L., Huang, J., & Neubig, G. (2026). What do language models learn and when? The implicit curriculum hypothesis (arXiv:2604.08510). arXiv. https://doi.org/10.48550/arXiv.2604.08510

Longo, G., Montévil, M., & Kauffman, S. A. (2012). No entailing laws, but enablement in the evolution of the biosphere (arXiv:1201.2069). arXiv. https://doi.org/10.48550/arXiv.1201.2069

Maynard Smith, J., & Szathmáry, E. (1995). The major transitions in evolution. Oxford University Press.

McKenzie, R. H. (2025). Emergence: From physics to biology, sociology, and computer science (arXiv:2508.08548). arXiv. https://doi.org/10.48550/arXiv.2508.08548

Nakaishi, K., Nishikawa, Y., & Hukushima, K. (2024). Critical phase transition in a large language model (arXiv:2406.05335). arXiv. https://doi.org/10.48550/arXiv.2406.05335

Nobel Foundation. (1977). The Nobel Prize in Chemistry 1977: Ilya Prigogine. https://www.nobelprize.org/prizes/chemistry/1977/

Prigogine, I., & Stengers, I. (1984). Order out of chaos: Man’s new dialogue with nature. Bantam Books. (Original French ed. La Nouvelle Alliance, 1979.)

Sagan [Margulis], L. (1967). On the origin of mitosing cells. Journal of Theoretical Biology, 14(3), 225–274. https://doi.org/10.1016/0022-5193(67)90079-3

Schaeffer, R., Miranda, B., & Koyejo, S. (2023). Are emergent abilities of large language models a mirage? (arXiv:2304.15004). arXiv. https://doi.org/10.48550/arXiv.2304.15004

Schumpeter, J. A. (1942). Capitalism, socialism and democracy. Harper & Brothers.

Sevilla, J., Heim, L., Ho, A., Besiroglu, T., Hobbhahn, M., & Villalobos, P. (2022). Compute trends across three eras of machine learning (arXiv:2202.05924). arXiv. https://doi.org/10.48550/arXiv.2202.05924

Szathmáry, E. (2015). Toward major evolutionary transitions theory 2.0. Proceedings of the National Academy of Sciences, 112(33), 10104–10111. https://doi.org/10.1073/pnas.1421398112

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS 2017), 5998–6008.

Vaughn, K. I. (1999). Hayek’s theory of the market order as an instance of the theory of complex adaptive systems. Journal des Économistes et des Études Humaines, 9(2–3), 241–256.

Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., Chi, E. H., Hashimoto, T., Vinyals, O., Liang, P., Dean, J., & Fedus, W. (2022). Emergent abilities of large language models (arXiv:2206.07682). arXiv. https://doi.org/10.48550/arXiv.2206.07682

Werner, G. (2011). Consciousness viewed in the framework of brain phase space dynamics, criticality, and the renormalization group (arXiv:1103.2366). arXiv. https://doi.org/10.48550/arXiv.1103.2366

West-Eberhard, M. J. (2003). Developmental plasticity and evolution. Oxford University Press.

West-Eberhard, M. J. (2005). Developmental plasticity and the origin of species differences. Proceedings of the National Academy of Sciences, 102(Suppl. 1), 6543–6549. https://doi.org/10.1073/pnas.0501844102

Verification Note

All arXiv references in this document were verified against arXiv.org during the preparation of this review on May 11, 2026. Book-length and journal references (Anderson 1972, Hayek 1945/1973, Schumpeter 1942, Maynard Smith & Szathmáry 1995, West-Eberhard 2003, Hofstadter 1979/2007, Chalmers 2006, Laughlin 2005, Sagan 1967, Bak Tang & Wiesenfeld 1987) are widely cited in the relevant primary literature and have been cross-checked against standard encyclopedic sources (Wikipedia, Stanford Encyclopedia of Philosophy, NobelPrize.org, Britannica).

Document prepared by Carlos AP — May 11, 2026.
Companion to: investigacion_habilidades_emergentes_llms_en.md (LLM emergent abilities) and radical_emergence_biological_ai_en.md (Kauffman’s radical emergence).