Michael Levin

orcid 0000-0001-7292-8084 openalex A5085228887 name_hash ce320a138526cc9e31662d4f…orcid 0000-0003-2767-6919 openalex A5100662025 openalex A5135994514 openalex A5135901917 orcid 0000-0002-5765-6591 openalex A5064159188

Authored

Introduces

Studies

Affiliations

Cited by

Authored papers (32)

Brains and where else? Mapping theories of consciousness to unconventional embodiments2026ⓒ 1
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 theories of consciousness (ToCs) are not substrate-specific to neural tissue—brains are privileged by convention, not by theoretical necessity. Surveying major current ToCs and mapping their mechanistic requirements against findings from evolutionary biology, developmental bioelectricity, and synthetic bioengineering, the paper shows that algorithms and mechanisms characteristic of neural cognition have ancient pre-neural roots: minds, on this view, preceded brains. The analytical instrument introduced is a systematic substrate-mapping procedure that asks, for each ToC, whether its load-bearing features specifically pick out brains or whether those features are realizable in unconventional embodiments including cells, tissues, organoids, and life-technology chimeras. Notably, several contemporary theorists—responding to advances in AI and organoid bioengineering—have already begun extending their frameworks to synthetic systems, lending empirical traction to the mapping. Published in a theme issue on world models in natural and artificial intelligence (DOI: 10.1098/rsta.2025.0082), the paper argues this implies that the science of consciousness must remain open to inner perspective in unconventional substrates, and that a ToC which cannot account for distributed, pre-neural, or synthetic minds is underdetermined by the biological evidence it purports to explain.
Cognitive glues are shared models of relative scarcities: the economics of collective intelligence2026
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 affordance that allows autonomous, heterogeneous agents operating at multiple scales to form mutually compatible plans without requiring centralized direction. Deploying the TAME (Technological Approach to Mind Everywhere) framework, published in *Philosophical Transactions of the Royal Society A* 384(2320), the paper formalizes cognitive glue as a shared model of relative scarcities, and claims that any system achieving genuine collective intelligence must solve the Hayekian distributed-knowledge problem in broadly the same structural way that price signals do: by compressing dispersed local information into a substrate-neutral, action-guiding representation that propagates across the system. The paper introduces the price system as a *generic template* — an abstract model against which candidate cognitive glues in biological, social, and artificial multi-scale systems can be evaluated. This implies that identifying the cognitive glue of any collective intelligence (a neural assembly, an immune system, a venture-capital network) requires locating the mechanism that encodes and transmits relative scarcities across subunits, and that designing or enhancing collective intelligence is, at root, a problem of designing better shared scarcity models.
Open questions about time and self-reference in living systems2026
The Causally Emergent Alignment Hypothesis: Causal Emergence Aligns with and Predicts Final Reward in Reinforcement Learning Agents2026
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 finding Pigozzi and Levin formalize as the Causally Emergent Alignment Hypothesis. Measured via ΦID (Integrated Information Decomposition) applied to neural-network agents' latent states over their full training lifetimes, causal emergence scores captured early in training predicted end-of-training reward across six environments spanning a complexity spectrum and across multiple RL algorithms and agent architectures. The instrument introduced is a ΦID-based causal emergence estimator applied to agent latent-space dynamics, enabling trajectory-level comparison between representational reorganization and reward signals. Crucially, this alignment parallels a known biological phenomenon: minimal biological agents demonstrably increase their causal emergence after acquiring new memories, and the same axis of representational reorganization appears operative in artificial agents. This paper argues that causal emergence constitutes a previously undisclosed dimension of how neural representations reorganize during RL training, implying that interventions targeting causal emergence directly — rather than reward signals — may yield mechanistically grounded routes to more capable and interpretable RL agents, while simultaneously identifying a principled structural axis along which biological and artificial cognition converge.
Topological constraints on self-organization in locally interacting systems2026
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 goal. By analyzing free-energy scaling under domain-wall formation across three model systems—the Potts model, autoregressive models, and hierarchical networks—the paper establishes necessary conditions on graph topology for an ordered phase to exist. The core method introduced is a domain-wall free-energy scaling analysis applied comparatively across these three substrate classes, tracking how the combinatorics of graph connectivity either suppress or permit spontaneous ordering. On planar graphs with the interaction topology characteristic of flat autoregressive language models, domain-wall formation is combinatorially cheap relative to system size, so long-range coherence degrades as sequence length grows; this is a structural, not a training, limitation. Hierarchical and multiscale networks—the topology prevalent in biological systems such as morphogenetic and neural substrates—make domain-wall formation sufficiently costly that ordered phases persist across scales. Published in the Phil. Trans. R. Soc. A theme issue on world models (384, issue 2320, 2026), the paper operationalizes the claim that all intelligence is collective intelligence via free-energy geometry, arguing that architectural hierarchy is not merely a performance heuristic but a topological prerequisite for sustained coherent self-organization, with direct implications for why biological morphogenesis succeeds at pattern maintenance where flat LLMs structurally cannot.
World models, artificial general intelligence and the hard problems of life–mind continuity: toward a unified understanding of natural and artificial intelligence2026
Ingressing Minds: Causal Patterns Beyond Genetics and Environment in Natural, Synthetic, and Hybrid Embodiments2025ⓒ 2
Topological constraints on self-organisation in locally interacting systems2025
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 for arbitrarily long output sequences. By generalizing the Landau–Lifshitz scaling argument and Peierls' domain-wall counting to a broad universality class via a Topological Equivalence Theorem (Theorem 1), the paper shows that any local Hamiltonian on a graph shares asymptotically equivalent free energy with a nearest-neighbour Ising model on the same combinatorial structure—meaning the existence or non-existence of a phase transition reduces entirely to graph topology. Three model systems are analyzed: the one-dimensional windowed Potts model, AR(ω) autoregressive models (Corollary 2), and hierarchical clique networks. For the Potts chain, domain-wall entropy scales as log(L−1) while energy is bounded by ωE^max, forcing ΔF negative for sufficiently large sequence length L at any nonzero temperature. Transformer attention with a finite context window ω maps directly onto the AR(ω) framework (Proposition 2 via Theorem 3), inheriting the same no-go result. Conversely, biological systems organized as nested cliques—cells forming tissues, tissues forming organs—admit a non-empty critical temperature range of hierarchical order (Proposition 3), achievable when the clique count ℓ and size nmax satisfy ℓ/r > nmax^nmax/e. The paper argues this constitutes a principled thermodynamic explanation for why autoregressive LLMs exhibit coherence failures on long tasks while multicellular organisms maintain large-scale morphogenetic order, and proposes that stigmergy and embodiment function as evolutionary responses to this topological no-go constraint.
AI: a Bridge toward Diverse Intelligence and Humanity’s Future2024ⓒ 1
Current AI debates are importantly incomplete because they fixate on large language models while ignoring the broader space of impending minds — including cyborgs, hybrots, genetically augmented humans, and other chimeric beings — that will demand ethical frameworks far beyond anything LLMs require. Levin's central claim is that virtually every concern raised about AI (alignment, confabulation, persona instability, trust, objectophilia, replacement anxiety) maps onto perennial, still-unsolved problems in developmental biology, child-rearing, and human identity that predate AI by millennia. The paper introduces the concept of *synthbiosis* — a term coined collaboratively with GPT-4, derived from Greek σύνθεσις and βίος, to denote the flourishing co-existence of evolved and engineered material in novel chimeric configurations such as cyborgs and hybrots — and deploys the Diverse Intelligence (DI) framework developed across prior work including Levin 2019 (Frontiers in Psychology) and Clawson & Levin 2022 (Biological Journal of the Linnean Society) to argue that intelligence, agency, and moral worth form a continuum rather than a binary. The cognitive light cone construct — which demarcates the spatial and temporal scale of goals an agent can effectively pursue, scaling from individual cells to multicellular organisms — is used to argue that cancer represents a shrinkage of this cone back to microbial scale, while embryogenesis expands it, demonstrating that the Self/World boundary is plastic within a single lifetime. Levin argues that humanity's survival of the coming wave of unconventional beings requires abandoning origin-based and morphology-based moral heuristics entirely and replacing them with principled, science-driven continuum ethics, because the failure mode of excessive xenophobia has historically vastly exceeded the failure mode of misplaced compassion.
Collective intelligence: A unifying concept for integrating biology across scales and substrates2024ⓒ 94
Collective intelligence, understood as William James' capacity to reach the same goal by different means, operates not only in beehives and ant colonies but as a scale-free organizing principle across all biological substrates — from gene-regulatory networks capable of Pavlovian conditioning, to Xenopus melanocytes executing all-or-none neoplastic conversion, to planarian fragments stochastically regenerating 1-head and 2-head worms at a stable ~1:2 ratio. The paper introduces the multiscale competency architecture (MCA) as its unifying conceptual instrument, which formalizes how each hierarchical level — molecular, cellular, tissue, organismal, and swarm — navigates distinct problem spaces (metabolic, physiological, morphological, behavioral) and how higher levels deform the energy landscape for subunits without micromanaging them. Specific mechanistic evidence includes: bioelectric disruption of GlyCl-expressing instructor cells in Xenopus tadpoles driving 70% of cohort animals into a fully-converted melanoma-like phenotype with no partially-converted individuals until an AI-parameterized model predicted a drug combination that finally produced them; keratocyte fragments electrotaxing to the anode while intact keratocytes migrate to the cathode, demonstrating that collective behavior can directly contradict the summed tendency of components; and mouse neural crest cells grafted into chick embryos successfully navigating the foreign embryonic face to form teeth, while collectives of rhombomere cells resist neighbor re-induction that overrides individual cells. The SCHEEPDOG electrotactic platform is named as a cross-disciplinary tool for steering keratinocyte collectives with patterned dynamic fields, operationalizing the distinction between individual and collective cell behaviors. The paper argues these examples compel developmental biology, regenerative medicine, and cancer research to adopt behavioral-science formalisms — including active inference, perceptual bistability modeling, and causal information theory — to predict and control large-scale morphogenetic outcomes that molecular pathway mapping alone cannot address.
Self-Improvising memory: a perspective on memories as agential, dynamically reinterpreting cognitive glue2024
merged
Stress sharing as cognitive glue for collective intelligences: a computational model of stress as a coordinator for morphogenesis2024
merged
Cellular signaling pathways as plastic, proto-cognitive systems: Implications for biomedicine2023ⓒ 47
candidate
Darwin's agential materials: evolutionary implications of multiscale competency in developmental biology2023
Cellular collectives operating between the genotype and anatomical phenotype constitute an agential substrate that fundamentally reshapes the evolutionary search process—this is the central claim of Levin's review, which introduces the multiscale competency architecture (MCA) as the organizing framework. Cells, tissues, and organs exhibit regulative plasticity across metabolic, transcriptional, physiological, and anatomical problem spaces because metazoan cells descend from unicellular ancestors with rich behavioral repertoires; evolution therefore searches not the astronomically rugged space of genomic microstates but the smoother space of behavior-shaping signals that exploit these pre-existing competencies. Concrete examples anchor the argument: Xenopus laevis frog skin cells liberated from developmental context spontaneously form self-motile Xenobots capable of kinematic self-replication, a mode unknown elsewhere in the tree of life; polyploid newt kidney tubules achieve normal diameter through a single giant cell wrapping around itself rather than the usual eight-to-ten-cell arrangement, demonstrating real-time downward causation without genomic change; and a 2-day bioelectric intervention targeting planarian ion channels permanently resets head-number patterning, with gap junctional blockade shown to recapitulate 100–150 million years of morphospace divergence across flatworm species. A computational model by Shreesha and Levin (2023) in Entropy directly demonstrates that higher cellular competency levels accelerate evolutionary search and initiate a self-reinforcing ratchet in which improved competency makes structural genomes harder to read by selection, further driving investment in problem-solving capacity. The paper argues this implies that biological evolvability is not a property of genetic architecture alone but emerges from the computational intelligence of the morphogenetic layer, explaining the speed and robustness of evolutionary change and motivating top-down intervention strategies for regenerative medicine and synthetic bioengineering.
Darwin's agential materials: evolutionary implications of multiscale competency in developmental biology2023
merged
Developmental Bioelectricity: the cognitive glue enabling evolutionary scaling from physiology to mind2023
Developmental bioelectricity—the network of ion channels, gap junctions, and neurotransmitter-mediated Vmem (membrane potential) dynamics operating across all body cells, not just neurons—functions as an evolutionarily ancient cognitive glue that scaled physiological competencies of single cells into collective intelligences capable of navigating morphospace, long before brains and muscles evolved to navigate 3D behavioral space. Published in Animal Cognition (2023) 26:1865–1891, Levin's framework introduces the Multiscale Competency Architecture as the organizing concept, arguing that biological systems are nested problem-solvers in which each hierarchical level—from ion channels through cells, tissues, organs, and organisms—deploys adaptive behavior within its own problem space. Concrete evidence for morphogenetic cognition includes: planarian flatworms regenerating two-headed morphologies after transient pharmacological Vmem perturbation, with the two-headed state persisting through subsequent rounds of amputation without further manipulation; Xenopus tadpoles engineered with ectopic tail eyes that nonetheless support functional color-vision learning; and Xenobots, assembled from dissociated frog skin cells, that achieve kinematic self-replication within 48 hours despite having no evolutionary history as self-replicating entities. The deep symmetry between neural and non-neural bioelectricity is operationalized through the Evolutionary Pivot hypothesis: the same ion-channel and gap-junction hardware that originally coordinated anatomical morphospace navigation was exapted for behavioral 3D-space navigation when nerve and muscle appeared, differing chiefly in timescale (milliseconds vs. hours). Levin argues this implies that the conceptual and technical toolkit of behavioral neuroscience—memory, representation, perceptual bistability, false-memory induction, neural decoding—is fully portable to developmental biology and regenerative medicine, and that training paradigms for cells and tissues will outperform bottom-up genetic micromanagement for controlling complex morphological outcomes.
Endless forms most beautiful 2.0: teleonomy and the bioengineering of chimaeric and synthetic organisms2023
merged
Teleonomy — goal-directed behavior measurable across any substrate — is proposed as the single deep invariant that unifies evolved organisms, engineered machines, and every hybrid configuration between them, replacing phylogenetic classification as the operative framework for synthetic biology and diverse intelligence research. Xenobots, derived from dissociated Xenopus laevis epidermal cells with zero genomic editing, self-assemble within 48 hours into motile spherical constructs that repair damage and, critically, discover kinematic self-replication — herding loose cells into daughter constructs — a reproductive mode unprecedented in any known organism (Kriegman et al., 2021). The Willett et al. (2021) brain-computer interface achieving real-time handwriting decoding from motor cortex microelectrode arrays, the MEART hybrot coupling rat cortical cultures on multi-electrode arrays to robotic drawing arms, and the Ophiocordyceps unilateralis zombie-ant system in which fungal networks invade adductor muscle fibers while leaving the ant brain entirely intact (Fredericksen et al., 2017) all instantiate the same principle: functional chimaerism does not require understanding the host system's internal wiring, only identifying the teleonomic lever. The paper introduces the multi-scale competency architecture as both a descriptive framework and an experimental instrument — the claim that each hierarchical subsystem maintains homeostatic goal-directedness in its own problem space, enabling upper levels to treat lower levels as reliable black boxes and thereby making evolutionary search, regenerative robustness, and rational bioengineering mutually intelligible. This implies that the genome-to-anatomy relationship is permanently underdetermined from sequence data alone, that meaningful prediction and control of large-scale morphology requires reading and writing goal states at the appropriate scale rather than micromanaging molecular pathways, and that ethical frameworks must be rebuilt around teleonomic capacity rather than substrate or phylogenetic origin.
Endless forms most beautiful 2.0: teleonomy and the bioengineering of chimaeric and synthetic organisms2023
Cellular collectives exhibit goal-directed competency that is substrate-independent, composition-independent, and origin-independent — a property Clawson and Levin term teleonomy — and this invariant, not genomic or phylogenetic identity, is proposed as the organizing principle for understanding, engineering, and ethically evaluating the full spectrum of possible living agents. The argument is grounded in concrete demonstrations: Xenopus tadpoles with ectopic eyes grafted onto their tails still achieve functional vision, with optic nerves routing to the spinal cord rather than the brain (Blackiston & Levin, 2013); Fankhauser's polyploid salamander embryos maintain normal tubule cross-sections by having a single enlarged cell wrap around itself when cell number is insufficient; and frog skin cells dissociated and cultured in vitro self-organize within 48 hours into Xenobots — spherical proto-organisms that swim via cilia, repair damage, and execute kinematic self-replication never observed in any other organism (Kriegman et al., 2020, 2021) — without any transgenic modification. The multi-scale competency architecture (MSCA) is introduced as the explanatory framework: every subsystem from molecular networks to organs pursues goals in its relevant problem space, enabling the collective to navigate novel morphological and behavioral spaces that selection never directly visited. Planarian lines permanently converted to two-headed morphologies by bioelectric circuit manipulation, and brain-computer interfaces achieving handwriting decoding from paralyzed patients at speeds exceeding prior approaches (Willett et al., 2021), are marshaled as evidence that teleonomic robustness operates across evolved and engineered configurations alike. The paper argues this implies that regenerative medicine, robotics, and ethics must abandon binary categories — organism vs. machine, evolved vs. designed — and replace them with a continuum parameterized by the spatiotemporal scale and competency of an agent's goal-directed behavior.
The collective intelligence of evolution and development2023
merged
Toward an ethics of autopoietic technology: stress, care, and intelligence2023
merged
Biology, Buddhism, and AI: Care as the Driver of Intelligence2022
Care—defined as concern for the alleviation of stress (the delta between current and optimal states)—is proposed as the substrate-independent invariant that unifies biology, Buddhist philosophy, and artificial intelligence in explaining how intelligence scales. The paper introduces the Care Light Cone (CLC) formalism, a spatiotemporal representation adapted from relativistic light cone geometry, which maps the boundary of states any agent can represent, pursue, and work to modify, distinguishing it from the Physical Light Cone (PLC) that maps merely achievable physical states. Across embodiments ranging from bacterial biofilms navigating metabolic space to Xenobots (protoorganisms made of frog skin cells) exhibiting coherent behavior without evolutionary backstory, to hybrots and cyborgs integrating living brain tissue with artificial bodies, the size of an agent's CLC directly indexes its cognitive sophistication. Cancer is reframed as a pathological contraction of the CLC caused by inappropriate reduction of gap junctional connectivity, reverting cells to ancient unicellular stress-reduction loops. The Bodhisattva vow—'for the sake of all sentient life, I shall achieve awakening'—is operationalized as a design principle that formally extends a system's CLC to infinite spatiotemporal scope, triggering a positive feedback loop between expanding Care and expanding intelligence analogous to major evolutionary transitions such as the archaea-eubacteria fusion producing eukaryotic cells. The paper argues this implies that outward-directed Care is not merely ethically desirable but mechanistically necessary for the development of artificial general intelligence, and that moral obligation toward any being—chimeric, synthetic, or evolved—should be calibrated to the scope of Care that being can exhibit, rather than to its material composition or phylogenetic origin.
Design for an Individual: Connectionist Approaches to the Evolutionary Transitions in Individuality2022
Evolutionary transitions in individuality (ETIs) require interaction structures among lower-level units that compute non-linearly separable functions — the same class of functions that single-layer Perceptrons cannot represent and that necessitate depth in connectionist models. Watson, Levin, and Buckley formalize this claim through the framework of evolutionary connectionism, which demonstrates a functional equivalence (not merely analogy) between the action of natural selection on heritable variation in relationships and unsupervised associative learning in neural networks. The core argument distinguishes non-decomposable collective characters — where the sign of the effect of one particle's character on collective fitness reverses depending on context, as in an XOR or division-of-labour game — from merely non-aggregative but monotonic interactions, which remain explanatorily redundant at the collective level. Two formal hypotheses follow: H1, that individuality requires a developmental process computing a non-linearly separable function of embryonic particle characters to coordinate reproduction; and H2, that the conditions for deep model induction, familiar from multi-layer Perceptrons and LeCun et al. (2015)-style deep learning, are predictive of the conditions under which bottom-up natural selection can produce an ETI. Power's Sudoku-based ecological model and Tudge et al.'s two-player division-of-labour simulations provide partial empirical scaffolding, but neither delivers a unified evolutionary model with deep, asymmetric interaction structures and no system-level selection. The paper argues this implies that ETIs are inseparable from the evolution of developmental (basal cognitive) processes and that four conditions — heritable relational variation, asymmetric interaction structures capable of depth, repeated environmental perturbations, and parsimony pressure — are jointly necessary and potentially sufficient for a transition in individuality to occur under bottom-up selection.
Technological Approach to Mind Everywhere: An Experimentally-Grounded Framework for Understanding Diverse Bodies and Minds2022ⓒ 174
TAME—Technological Approach to Mind Everywhere—formalizes a non-binary, empirically grounded framework for recognizing, comparing, and manipulating cognition across radically diverse substrates, from single cells and gene regulatory networks to chimeric bioengineered organisms and hybrots. Central to the framework is an axis of persuadability ranging from brute-force hardware rewiring (e.g., mechanical clocks) through homeostatic circuits and trainable animals to rational-argument-responsive humans, which serves as a semi-quantitative tool for determining optimal intervention strategy for any given system. Empirical anchors include: planarian flatworms regenerating barium-insensitive heads by efficiently traversing transcriptional space to regulate a small subset of genes; gap-junctional blockade producing planaria with heads morphologically matching other extant species despite wild-type genetics; and tadpoles with ectopically tail-placed eyes successfully performing visual learning tasks via spinal cord re-routing. The framework introduces a 'cognitive light cone' diagram plotting spatio-temporal scale of goal-directed activity to place microbes, rats, and humans on a common axis without appealing to substrate or evolutionary origin. Developmental bioelectricity—implemented through pre-neural ion channels and gap junctions scaling cell-level feedback into anatomical homeostasis—is identified as evolution's primary medium for enlarging cognitive boundaries, and the same gap-junction closure that produces cancer is argued to represent a shrinking of the multicellular Self back to unicellular-scale goals. TAME implies that morphogenesis is a tractable model of basal cognition, that multi-scale competency architecture smooths fitness landscapes and accelerates evolution, and that synthetic bioengineering will soon produce minds for which neither phylogeny nor genetics provides an adequate cognitive framework.
Living things are not (20th century) machines: updating mechanism metaphors in light of the modern science of machine behavior2021
referenced-only
Living Things Are Not (20th Century) Machines: Updating Mechanism Metaphors in Light of the Modern Science of Machine Behavior2021ⓒ 99
Bongard and Levin argue that the longstanding debate over whether living things are machines has been conducted against a 20th-century, static definition of 'machine' that modern engineering has already surpassed, making the debate largely obsolete. Seven classical properties routinely invoked to distinguish machines from life—independence, predictability, human design, linear modularity, cognitive absence, reductionist tractability, and clear hardware/software distinction—each fail when tested against contemporary systems: evolutionary algorithms have produced jet engines (Yu et al., 2019), metamaterials (Zhang et al., 2020), and computer-designed Xenopus-cell organisms (Kriegman et al., 2020) without direct human specification of outcomes; backpropagation-trained deep networks resist exactly the reductionist decomposition Nicholson (2019) listed as a necessary machine feature; and planarian flatworms harbor re-writable bioelectric voltage patterns in non-neural cells that function as latent morphogenetic memory editable without touching the genome (Durant et al., 2017). The framework the paper introduces is the multi-axis continuum of 'machine behavior'—a 2D option space spanning degree of design vs. evolution and degree of autonomy, applicable independently at each level of biological organization (cell, organism, swarm)—drawn from the emerging interdisciplinary field Rahwan et al. (2019) named 'machine behavior.' Bongard and Levin argue this implies that the correct response is not to abandon the machine metaphor but to update it: biology and computer science are branches of a single information science, sharp boundaries between evolved and designed systems will not persist, and a conceptual framework that treats agency, programmability, and autonomy as continuous variables across all substrates is both necessary and sufficient to guide synthetic bioengineering, regenerative medicine, and machine design in the coming decades.
The Cognitive Lens: a primer on conceptual tools for analysing information processing in developmental and regenerative morphogenesis2019ⓒ 84
candidate
The computational boundary of a 'self': developmental bioelectricity drives multicellularity and scale-free cognition2019ⓒ 274
Scale-Free Cognition, the framework introduced here, proposes that any coherent Self is demarcated by a 'cognitive light cone'—a spatio-temporal boundary of events a system can measure, model, and attempt to regulate—and that this boundary expands through evolutionarily conserved bioelectric mechanisms rather than requiring nervous systems. The core claim is that developmental bioelectricity, implemented primarily through gap junctions and voltage-gated ion channels, provides the proximate mechanism by which single-cell homeostatic loops scale into multicellular cognitive agents: when cells couple via gap junctions they share a unified Umwelt, transforming individually local set points into organ-level morphogenetic goals. Empirical support is drawn from three substrate types: in Xenopus tadpoles, craniofacial organs in abnormal positions still converge on a 'correct frog face configuration' (Vandenberg et al., 2012), demonstrating invariant anatomical goal-pursuit; in genetically normal tadpoles, depolarization of a specific melanocyte population is sufficient to induce metastatic transformation (Lobikin et al., 2012); and conversely, human oncogene-driven tumorigenesis can be blocked by optogenetic or constitutive hyperpolarization (Chernet and Levin, 2013b, 2014). Cancer is reframed not as genomic chaos but as a reversible shrinkage of the computational boundary—gap-junction uncoupling collapses a cell's cognitive horizon from whole-body to single-cell scale, recapitulating unicellular behavioral modes including maximal proliferation and migration. A 24-hour progesterone stimulus via wearable bioreactor was sufficient to initiate 11 months of autonomous limb-regeneration activity in adult Xenopus (Herrera-Rincon et al., 2018), illustrating how brief intervention at the correct level of organization can trigger a pre-encoded morphogenetic module. Levin argues this implies that biomedicine, AI design, and exobiology should prioritize identifying and communicating with agents at the level of their actual goal-directed organization rather than exclusively targeting molecular mechanisms.
The computational boundary of a "self": developmental bioelectricity drives multicellularity and scale-free cognition2019
merged
Re-membering the body: applications of computational neuroscience to the top-down control of regeneration of limbs and other complex organs2015ⓒ 177
candidate
Generalizing frameworks for sentience beyond natural speciesⓒ 28
Criteria anchored to vertebrate neuroanatomy and verbal behavior — exemplified by the Smith & Boyd (1991) framework and the Turing Test — are structurally inadequate for the full space of possible sentient agents, and the Crump et al. (2022) 8-criterion decapod framework, while a genuine advance, must now be generalized far beyond natural phylogenetic lineages. Levin argues that associative learning, Crump et al.'s criterion #7, already occurs in gene regulatory networks and non-neural morphogenetic agents, meaning the pivot from 'neurons' to 'electrically active cell' unlocks most of the 8 criteria for substrates including organoids, xenobots (Kriegman et al., 2020, PNAS 117:1853), cultured-neuron 'hybrots' that learned in simulated game-worlds (Kagan et al., 2021), and synthetic living machines built on frog-cell platforms (Blackiston et al., 2021, Sci Robot 6:eabf1571). The instrument Levin implicitly introduces is a substrate-neutral invariant search: replacing phylogenetic or anatomical heuristics with deep functional invariants — such as 'competency in navigating arbitrary problem spaces' (Fields & Levin, 2022) — that apply across biological, chimeric, AI, and exobiological agents alike. The LaMDA debate (Thopilian et al., 2022) illustrates the cost of this gap: strong opinions circulate with no defensible criteria. Levin argues this implies that developing principled, quantitative, substrate-independent sentience frameworks is an existential requirement for humankind, not merely an academic refinement, because the coming decades will embed genuinely novel agents — cyborgs, biorobots, AI systems, synthetic organisms — into society before ethics has caught up.
Multiple ways to implement and infer sentienceⓒ 4
Rouleau and Levin argue that the evidential standard applied to inferring sentience is inconsistently applied: behavioural signatures used to attribute subjective experience to non-human animals — goal-directedness, anticipatory reorientation, classical conditioning, game-theoretic risk evaluation, and mimicry — are present in plants yet routinely dismissed without principled justification. The core theoretical move is invoking multiple realizability (Bickle, 2006) and substrate independence (Bostrom, 2003) to argue that felt states, like computation, need not depend on any single physical medium, including the neural circuitry of vertebrates. Plants synthesize glutamate — the most abundant excitatory neurotransmitter in the human central nervous system — and propagate action-potential-like depolarizations along distributed vascular networks, providing at least partial electrochemical homology with animal nervous systems despite radically different anatomical organization. Rouleau and Levin extend the argument beyond plants to metaplastic nanowire networks (Loeffler et al., 2023), robotic systems (Clawson and Levin, 2022), and artificial intelligences, proposing a system-agnostic inferential framework they call a behaviour-based, substrate-independent approach to sentience attribution. Drawing on biological degeneracy — the principle that structurally dissimilar neural regions can achieve identical functional outcomes, as in blindsight responses mediated by subcortical nuclei — the paper argues that if diverse brain architectures can instantiate sentience within animals, non-neural tissues and non-biological substrates cannot be excluded a priori. This implies that current taxonomies of sentience are anthropocentric artefacts rather than principled scientific boundaries, and that a unified, system-agnostic framework spanning cybernetics, bioengineering, materials science, and biomedicine is both possible and necessary.
Topological constraints on self-organisation in locally interacting systems (Sacco, Sakthivadivel, Levin 2025)
referenced-only