Collective intelligence: A unifying concept for integrating biology across scales and substrates

What to take away

1. 1. Bioelectric disruption of a small number of GlyCl-expressing instructor cells in Xenopus tadpoles converts up to 70% of a cohort into a fully melanoma-like hyperpigmented phenotype, and this conversion is all-or-none at the individual level — every melanocyte in a given animal converts or none do.
2. 2. An AI-parameterized computational model of the serotonergic-bioelectric signaling pathway (Lobikin et al. 2015, Sci. Signal 8:ra99) predicted two drugs plus a dominant-negative construct that, when tested, produced partially-converted animals for the first time in nearly a decade of experiments in this system (Lobo et al. 2017, Sci. Rep. 7:41339).
3. 3. Keratocyte cell fragments electrotax to the anode while intact keratocytes (a collective of those same fragments) migrate to the cathode, demonstrating that the directional preference of a collective can be the direct opposite of the summed preference of its individual components.
4. 4. Planarian fragments with bioelectrically destabilized states ('Cryptic Worms') regenerate as 1-head or 2-head forms at a stable ~1:2 ratio, yet any single fragment commits entirely to one outcome — head/tail identity is randomized at the population level, not at the level of individual cells within a fragment.
5. 5. Modification of gap-junction-mediated cell:cell communication during planarian regeneration can cause genetically wild-type fragments to produce head shapes characteristic of a different planarian species, visiting attractors in morphospace that are normally restricted to other genetic lineages.
6. 6. The paper introduces the multiscale competency architecture (MCA) as a framework asserting that each hierarchical level of biological organization — from molecular networks to swarms — solves problems in its own distinct problem space, and that higher levels harness subunit competencies by deforming their energy landscape without micromanaging individual components.
7. 7. Mouse neural crest cells grafted into chick embryos successfully navigate the foreign embryonic face and form teeth (Mitsiadis et al. 2003, PNAS 100:6541), while collectives of rhombomere cells transplanted across anterior-posterior Hox domains maintain their original gene expression against neighbor re-induction that readily overrides individual transplanted cells.
8. 8. The SCHEEPDOG electrotactic platform (Zajdel et al. 2020, Cell Syst. 10:506) is identified as an operational technology that uses patterned dynamic electric fields to precisely steer large-scale keratinocyte collective migration in a way that empirically distinguishes collective from individual cell-level behaviors.
9. 9. A segmentation-clock analog has been identified in bacterial biofilms responding to nitrogen stress via a negative-feedback oscillatory loop (Chou et al. 2022, Cell 185:145), suggesting that collective intelligence mechanisms operating in morphospace are recapitulated in physiological space by unicellular organisms, implying deeper evolutionary conservation than currently recognized.
10. 10. The paper raises the open hypothesis that the remarkable ability of neurons to unify into a centralized self — an emergent agent with memories and goals not assignable to any individual neuron — is an evolutionary repurposing of cell-communication strategies that originally solved problems in anatomical morphospace, implying that neuroscience and developmental biology share a common ancestral computational substrate.

Peer brief — for seminar discussion

McMillen and Levin's 2024 perspective in Communications Biology argues that collective intelligence is not a property unique to nervous systems or animal societies but a scale-free organizing principle running through all of biology, from gene-regulatory circuits exhibiting Pavlovian-like conditioning to Xenopus melanocytes executing population-level neoplastic decisions. The vehicle for this argument is what they call the multiscale competency architecture (MCA), a framework asserting that every hierarchical level of biological organization — molecular, cellular, tissue, organismal, swarm — navigates its own distinct problem space and that higher levels harness the autonomous competencies of subunits by reshaping their energy landscape rather than micromanaging them. The paper could alternatively have been organized around active inference or perceptual control theory frameworks, which the authors acknowledge but treat as complementary rather than foundational. The load-bearing empirical evidence comes from several systems. In Xenopus tadpoles, bioelectric disruption of instructor cells drives 70% of cohort animals into a fully-converted melanoma-like state with no partial phenotypes — a collective all-or-none decision. An AI-parameterized model of the serotonergic signaling circuit (Lobikin et al., Sci. Signal 8:ra99, 2015) then predicted a drug combination that produced partially-converted animals for the first time in nearly a decade of work (Lobo et al., Sci. Rep. 7:41339, 2017), demonstrating that the collective decision-making framework generates tractable predictive interventions. Planarian 'Cryptic Worms' regenerate as 1-head or 2-head forms at a ~1:2 stochastic ratio, yet each fragment commits entirely to one fate — randomization operates above the individual-cell level. Keratocyte fragments electrotax to the anode while intact keratocytes migrate to the cathode, showing collective behavior can directly invert component-level tendencies. Mouse neural crest cells grafted into chick embryos form teeth in a foreign embryonic environment, while neural crest cell collectives resist neighbor-induced Hox domain re-specification that readily overrides individual transplanted cells. The SCHEEPDOG electrotactic system is named as a cross-disciplinary tool that operationalizes the distinction between individual and collective cell behavior using patterned electric fields. The implication the authors press hardest is that developmental biology, regenerative medicine, and cancer research need behavioral-science formalisms — perceptual bistability modeling, active inference, causal information theory — to predict and control morphogenetic outcomes that molecular pathway analysis cannot, because collective decisions about anatomical targets are irreducibly supracellular. Cancer is reframed as a dissociative identity disorder of somatic collective intelligence. The prediction follows that tools from cognitive science explaining failures of learning, Bayesian updating, and attention will translate into explanations of birth defects and malformation. The most significant thing a critical reader would push back on is the evidentiary standard for attributing 'intelligence' and 'decision-making' to subcellular and cellular systems. The paper explicitly argues that all claims of intelligence must be grounded in the empirical utility of the framing for prediction and control, but it does not provide a systematic benchmark showing that MCA-derived interventions outperform, say, standard morphogen-gradient or gene-regulatory-network models across a defined set of regenerative or cancer outcomes. The Xenopus melanoma case is compelling but singular; whether the framework scales to routine predictive utility in other systems, or whether the cognitive vocabulary is doing conceptual work beyond organizing existing data under a unifying metaphor, remains to be demonstrated by prospective experimental tests in additional model organisms and disease contexts.

Methods (5)

• Chemical Genetics
  Use of small molecules to perturb specific ion channels or pathways and study resulting morphological outcomes.
• Gastruloids (trunk-like organoids)
  Stem-cell-derived 3D structures that recapitulate segmentation and axis formation, used to test morphogenetic goal-directedness.
• Grafting and Ablation
  Classical techniques to interrogate regulative capacity of embryos and neural crest by tissue removal or transplantation.
• Paraxial mesoderm explants in 2D culture
  In vitro system to study the segmentation clock in a flat geometry, revealing robustness and collective dynamics.
• SCHEEPDOG system
  Electrotactic platform using dynamic electric fields to steer collectives of keratinocytes, distinguishing collective vs individual cell behavior.

Frameworks (1)

• Perceptual Field Model
  Model introduced in Figure 2 explaining how collective intelligence expands the spatiotemporal perceptual field of a group beyond any individual member's capacity.

Findings (24)

• Modification of cell-cell communication during planarian regeneration causes genetically-normal fragments to produce heads appropriate to other species (Emmons-Bell et al. 2015, Sullivan et al. 2016)

  Shows that morphological attractors can be switched via physiological cues, revealing the navigation of morphospace by collectives.

• Bacterial biofilms exhibit bioelectrically-coordinated oscillatory growth patterns, with a negative feedback loop similar to the vertebrate segmentation clock (Liu et al. 2015, Chou et al. 2022)

  Shows that collective physiological oscillations in bacterial communities resemble mechanisms in animal development.

• Brief chloride ion channel activator drug exposure converts normal melanocytes to melanoma-like phenotype in an all-or-none population-level fashion (Blackiston et al. 2011, Lobikin et al. 2015)

  Reveals that melanocytes across the whole animal make a coordinated, stochastic decision to convert or remain normal.

• Individual neural crest or rhombomere cells lose memory of past inductive cues and adopt neighbors' expression; grafted collectives maintain original identity (Trainor and Krumlauf 2000)

  Collectives have extended temporal perceptual fields, maintaining positional memory that single cells lose.

• Gastruloids derived from embryonic stem cells arrive at a segmented target morphology despite very different ontogenic history (Veenvliet et al. 2020, van den Brink et al. 2020)

  Shows that the segmentation goal can be reached via alternative developmental pathways, fitting James' definition of intelligence.

• Model-predicted intervention (two drugs and a dominant negative construct) broke concordance among melanocytes, producing first partially-pigmented animals (Lobo et al. 2017)

  Demonstrates that breaking collective decision-making can be achieved, separating the decision from its coordination across cells.

• In planaria, the most depolarized region becomes the head; altering bioelectric pattern changes head location and number (Beane et al. 2011, Durant et al. 2017)

  The collective interprets relative voltage differences, not absolute values, to decide anterior identity.

• Grafted neural crest cell collectives maintain original Hox gene expression, while individual cells adopt neighbors' expression (Trainor and Krumlauf 2000)

  Collectives have increased positional memory and resist environmental induction, demonstrating expanded perceptual field in time dimension.

• Cells hetero-grafted from tissue in one phase of the segmentation clock into a different phase synchronize to the host phase (Horikawa et al. 2006)

  Demonstrates that the clock phase is collectively determined; individual cells entrain to the local collective rhythm.

• A computational model of melanocyte regulatory pathway revealed state-space decision points explaining all-or-none conversion (Lobikin et al. 2015)

  Mathematical modeling showed how cells navigate biochemical state space and face collective decision points.

Claims (17)

• Collective intelligence is not only the province of groups of animals; an important symmetry exists between the behavioral science of swarms and the competencies of cells and other biological systems at different scales.

  Foundational claim extending collective intelligence to all biological scales.

• Cognitive capacities exist on a spectrum and tools from behavioral sciences can be productively applied to understand and control systems far beyond familiar animals with central nervous systems.

  Core tenet of diverse intelligence; justifies the methodological borrowing across fields.

• The randomization of bioelectric state and downstream morphogen gradients is interpreted with respect to anterior-posterior organ identity by collectives, not by individual cells.

  Argues that stochastic outcomes in regeneration are still collective decisions, not cellular chaos.

• In axial patterning, collective decision-making enables large numbers of cells in a compartment to agree on an organ-scale anatomical fate despite stochastic influences upstream.

  Generalization of the all-or-none and cryptic planaria phenomena to axis formation.

• Higher-order entities distort the energy landscape for their subunits, benefiting from their competencies to navigate spaces of which the subunits are unaware.

  Explains how top-down causation leverages lower-level problem-solving for large-scale goals.

• The ability to recruit participants to complete tasks may be a central competency of collective intelligence that works across scales, from cells to swarms.

  Identifies recruitment as a cross-scale hallmark of collective intelligence.

• Biology's robustness, open-endedness, evolvability, and unique complexity likely depend on the fact that evolution works with an agential material.

  Central claim linking life's properties to the inherent competencies of its material substrate.

• The number of embryos emerging from an embryonic blastoderm is not genetically fixed but determined in real-time by physiological processes.

  Highlights dynamical determination of individuality through collective autopoiesis.

• The neural crest acts as an intelligently migrating collective that can alter individual cell behavior to achieve target morphology.

  Specific claim illustrating collective intelligence in a well-studied embryonic system.

• Some visual illusions that plague vertebrate nervous systems are recapitulated in collective intelligences such as ants.

  Suggests deep commonalities between neural and swarm cognitive processing.

Hypotheses (4)

• If a specific framing, which uses tools normally reserved for brains, results in fruitful new research programs on unconventional systems, then the scientific approach requires that we consider those systems to be bona fide subjects of that corner of the natural world.

  Methodological hypothesis from Box 1: the pragmatic test for extending cognitive terminology.

• We explore the hypothesis that collective intelligence is not only the province of groups of animals, and that an important symmetry exists between the behavioral science of swarms and the competencies of cells and other biological systems at different scales.

  The paper's guiding hypothesis, explicitly stated in the abstract and introduction.

• The remarkable ability of neurons to unify toward a centralized self is an evolutionary pivot of far earlier cell communication strategies that first solved problems in navigating anatomical morphospace.

  Proposes an evolutionary trajectory linking morphogenesis to neural cognition.

• Neurons achieve remarkable centralization through evolutionary reuse of earlier cell communication strategies for anatomical problem-solving

  Brain cognition may be evolutionary extension of collective intelligence mechanisms originally solving anatomical morphospace problems in multicellular organisms.

Questions (7)

• How cells build specific head shapes or how they know when to stop mitosis and morphogenesis when the correct head shape has been achieved.

  Identifies the gap in understanding the setpoints and stop signals in anatomical homeostasis.

• How do higher-order entities (organisms, organs, tissues, etc.) distort the energy landscape for their subunits?

  Questions the mechanism of top-down causation in collective systems.

• What shape head would regenerate (and would it ever reach the stop criterion and cease remodeling)?

  Thought experiment from Figure 5c asking what happens when half the neoblasts are from a different species.

• What precisely determines the borders between Self and outside world in multicellular collectives?

  Foundational question about emergence of individuality in biological systems; central to understanding how collectives self-organize into discrete agents.

• How many human Selves can be sustained by the excitable medium of a human brain?

  Parallels the embryonic blastoderm question to the human brain, linking developmental and cognitive science.

• How many individuals are present in an embryonic blastoderm?

  Questions the fixed-ness of individuality and points to the dynamical determination of embryo count.

• What precisely are we counting when we say it is 1 embryo?

  Raises the problem of defining individuality in a collective of cells.

Related work— refs + corpus + external arXiv

Cited / in-corpus / arXiv badges show which signals surfaced each row. Multi-source rows weighted higher.

• The computational boundary of a 'self': developmental bioelectricity drives multicellularity and scale-free cognition

  cited

  in corpus

  2019
  ≈ 88%
• Technological Approach to Mind Everywhere: An Experimentally-Grounded Framework for Understanding Diverse Bodies and Minds

  cited

  in corpus

  2022
  ≈ 87%
• Endless forms most beautiful 2.0: teleonomy and the bioengineering of chimaeric and synthetic organisms

  in corpus

  2023
  ≈ 91%
• Darwin's agential materials: evolutionary implications of multiscale competency in developmental biology

  in corpus

  2023
  ≈ 90%
• Artificial Collective Intelligence Engineering: a Survey of Concepts and Perspectives

  Roberto Casadei

  2023
  ≈ 88%
• The scaling of goals from cellular to anatomical homeostasis: an evolutionary simulation, experiment and analysis

  cited

  2023
  ≈ 87%
• The collective intelligence of evolution and development

  in corpus

  2023
  ≈ 87%
• The scaling of goals via homeostasis: an evolutionary simulation, experiment and analysis

  Johanna Bischof, Jennifer V. LaPalme, and Michael Levin Leo Pio-Lopez

  2022
  ≈ 87%
• AI: a Bridge toward Diverse Intelligence and Humanity’s Future

  in corpus

  2024
  ≈ 87%
• Collective Intelligence for Deep Learning: A Survey of Recent Developments

  Yujin Tang David Ha

  2022
  ≈ 86%
• Machine Collective Intelligence for Explainable Scientific Discovery

  Gyoung S. Na and Chanyoung Park

  2026
  ≈ 86%
• Future medicine: from molecular pathways to the collective intelligence of the body

  cited

  2023
  ≈ 86%
• Swarms, Phase Transitions, and Collective Intelligence

  LANL and Santa Fe Institute) Mark M. Millonas (Center for Nonlinear Studies and Theoretical Division

  2008
  ≈ 86%
• The biogenic approach to cognition

  in corpus

  2005
  ≈ 86%
• AI-enhanced Collective Intelligence

  Hao Cui and Taha Yasseri

  2024
  ≈ 85%
• Collective Intelligence Quanitifed for Computer-Mediated Group Problem Solving

  Craig Kaplan, Marko Rodriguez, Juana Diaz, Newton Der, Suzanne Garcia Dan Steinbock

  2007
  ≈ 85%
• Collective Intelligence in Citizen Science -- A Study of Performers and Talkers

  Elena Simperl, Markus Luczak-Roesch, Max Van Kleek, Nigel Shadbolt Ramine Tinati

  2014
  ≈ 85%
• Developmental Bioelectricity: the cognitive glue enabling evolutionary scaling from physiology to mind

  in corpus

  2023
  ≈ 85%
• Bootstrapping Life-Inspired Machine Intelligence: The Biological Route from Chemistry to Cognition and Creativity

  Michael Levin Giovanni Pezzulo

  2026
  ≈ 85%
• Collective intelligence: aggregation of information from neighbors in a guessing game

  Toni P\'erez and Jordi Zamora and V\'ictor M. Egu\'iluz

  2016
  ≈ 85%
• Cognitive glues are shared models of relative scarcities: the economics of collective intelligence

  in corpus

  2026
  ≈ 85%
• An AI Theory of Mind Will Enhance Our Collective Intelligence

  Catherine Drysdale, Jaime Ruiz-Serra Michael S. Harr\'e

  2025
  ≈ 85%
• (Im)possibility of Collective Intelligence

  Krikamol Muandet

  2025
  ≈ 85%
• Remapping and navigation of an embedding space via error minimization: a fundamental organizational principle of cognition in natural and artificial systems

  L\'eo Pio-Lopez, Chris Fields, Michael Levin Benedikt Hartl

  2026
  ≈ 85%
• Measurements of collective machine intelligence

  Michel Halmes

  2013
  ≈ 85%
• The Society of HiveMind: Multi-Agent Optimization of Foundation Model Swarms to Unlock the Potential of Collective Intelligence

  Susie Xi Rao Noah Mamie

  2025
  ≈ 85%
• Rules of collective migration: from the wildebeest to the neural crest

  cited

  2020
  ≈ 84%
• Toward an ethics of autopoietic technology: Dukkha, care, and Intelligence

  in corpus

  2023
  ≈ 84%
• Generalizing frameworks for sentience beyond natural species

  in corpus
  ≈ 84%
• Multiple ways to implement and infer sentience

  in corpus
  ≈ 84%

+27 more

Similar preprints — Semantic Scholar

Cited by (5)

• Topological constraints on self-organisation in locally interacting systems

  Topology of local interactions is the decisive factor determining whether a system can sustain long-range order, and decoder-only transformer architectures are provably unable to maintain such order f

• Brains and where else? Mapping theories of consciousness to unconventional embodiments

  Rouleau and Levin's 2026 analysis in *Philosophical Transactions of the Royal Society A* (384, issue 2320) demonstrates that the functional and operational principles underlying most contemporary theo

• The Causally Emergent Alignment Hypothesis: Causal Emergence Aligns with and Predicts Final Reward in Reinforcement Learning Agents

  Causal emergence in the latent-space representations of reinforcement learning agents is consistently predictive of final reward and aligns dynamically with reward improvement across training — a find

• Topological constraints on self-organization in locally interacting systems

  Topological constraints on interaction graphs determine whether a locally interacting system can sustain long-range ordered phases, and therefore whether it can self-organize toward a system-level goa

• Cognitive glues are shared models of relative scarcities: the economics of collective intelligence

  Levin and Lyons argue that the price system in market economies is not merely a resource-allocation device but the canonical instantiation of what they term a *cognitive glue* — a coordinating afforda