Benedikt Hartl, Milton L. Montero, Marcello Barylli, Sebastian Risi, Michael Levin

How phenotypic transformations are implemented by changes in underlying regulatory dynamics remains a central question in developmental biology. Inspired by D'Arcy Thompson's 1917 "On Growth and Form", we ask whether coherent large-scale transformations of morphology can be encoded as low-dimensional modulations of a self-organizing developmental system. We use neural cellular automata (NCAs) as bio-inspired models of distributed development, in which a shared local regulatory network grows target morphologies from a single cell. We apply low-rank adaptation (LoRA) to pretrained NCAs, representing each adapted developmental program as a low-rank modulation of a fixed regulatory scaffold. Horizontal and vertical scaling of a fully grown 2D emoji phenotype can each be implemented by rank-one adaptations. Their linear combinations parametrically control phenotype size, generalize beyond the training distribution, and compose with target-specific adapters. Strikingly, adaptations learned for one phenotype transfer zero-shot across structurally and semantically diverse phenotypes sharing the same reference scaffold, while largely preserving internal features. This suggests reusable system-level hyper-directions of scale rather than morphology-specific transformations. From approximately 25,000 independently trained phenotype-specific NCA adapters with a shared scaffold, we further identify latent low-dimensional directions that functionally control phenotypic variation including scaling, style, and symmetrical fission. Together, our results provide a computational realization of D'Arcy Thompson's remarkable grid transformations in a 2D NCA---a minimal cybernetic tissue in which variations of fully grown emoji phenotypes can be encoded, combined, and controlled through low-dimensional directions in regulatory weight space.

Read the full article at: arxiv.org

Miao Zeng, Roberto Murcio, Camilo Vargas-Ruiz, Elsa Arcaute

Achieving universal health coverage, as set out in Sustainable Development Goal 3.8, requires closing persistent geographic and socioeconomic gaps in healthcare access, especially in under-resourced settings across Africa. Healthcare-access inequality is shaped not only by poor local access, but also by limited connectivity to city- and regional-level services and opportunities. Conventional accessibility analysis can identify where poor access occurs, but not whether poorly served places form structurally disconnected pockets across scales. This paper therefore builds on and extends the percolation divergence tree framework to develop a connectivity-based multiscale approach for examining healthcare-access inequality in Ghana. It combines street-level accessibility mapping with the hierarchical structure of the road network to identify the scales at which inequality coincides with connectivity breaks. The results first show substantial inequality: around one quarter of the population lives more than 5 km from the nearest healthcare facility. Multiscale analysis further reveals distinct structural forms of poor access. In relatively well-connected, monocentric regions, local poor-access pockets emerge around metropolitan fringes despite overall regional advantage. In less well-connected, polycentric regions, poor access extends across larger subsystems, with local pockets nested within broader poorly served areas. These findings show that healthcare-access inequality reflects both local conditions and the hierarchical connectivity of the wider spatial system, which may also constrain marginalised communities' access to other key resources and services. The framework can inform targeted local interventions and policy coordination across local and regional scales.

Read the full article at: arxiv.org

Cameron Brooks, Estelle Janin, Gage Siebert, Cole Mathis, Orit Peleg & Sara Imari Walker 
Scientific Reports (2026)

The Search for Extraterrestrial Intelligence (SETI) has historically been a search for aliens like us, shaped by human-centric ideas of intelligence, communication, and technology. However, humans are not the only instance of an intelligent, communicating species on Earth, and thus not the only guide to how we might think about ETI. Here, we explore how non-human communication systems could complement existing SETI strategies, usually focused on complex, potentially decodable signals, using firefly communication as an illustrative example. Fireflies communicate their presence through evolved flash patterns that are distinguishable from complex visual backgrounds. Drawing on this strategy, we present a firefly-inspired model for detecting potential technosignatures within environments dominated by ordered astronomical phenomena, such as pulsars. Using pulsar data from the Australia Telescope National Facility, we generate simulated pulse sequences that exhibit evolved dissimilarity from the surrounding pulsar population of Earth, which would constitute a signature of intelligence embodied in a relatively simple signal. This approach shifts focus from anthropocentric assumptions about intelligence toward recognizing communication through its fundamental structural properties, specifically evolutionarily optimized contrast with natural backgrounds. Our model demonstrates that alien signals need not be inherently complicated nor must we decipher their meaning to identify them; rather, signals might be distinguishable as products of engineering or evolutionary design. We discuss implications for broadening SETI methodologies and leveraging the diverse forms of intelligence found on Earth.

Read the full article at: www.nature.com

Sebastian Sander-Oest

Synthese Volume 208, article number 143 (2026)

Self-assembly and self-organization are central concepts in theories of how patterns form in natural systems. However, literature on the topic has for long been riddled with inconsistent uses of these concepts, which often conflate their meaning or define them idiosyncratically. This is problematic as it struggles to make sense of the ubiquitous interaction between self-assembly and self-organization in chemistry and biology. The thermodynamic account is an attempt to draw up clear definitions of the concepts using thermodynamics as the appropriate framework for distinguishing the two concepts from each other. In this paper, I challenge this account and argue that self-assembly shouldn’t primarily be understood in terms of thermodynamics. I offer a philosophical analysis of core conceptual challenges for self-assembly and self-organization and argue that the thermodynamic approach follows a misguided conceptualization strategy. Drawing on recent philosophical work on scientific definitions, I propose an alternative account that treats the concept as a conceptual tool that directs our attention towards the kinds of features that are explanatorily relevant for the pattern-formation process and towards the features the resulting pattern itself will possess.Self-assembly and self-organization are central concepts in theories of how patterns form in natural systems. However, literature on the topic has for long been riddled with inconsistent uses of these concepts, which often conflate their meaning or define them idiosyncratically. This is problematic as it struggles to make sense of the ubiquitous interaction between self-assembly and self-organization in chemistry and biology. The thermodynamic account is an attempt to draw up clear definitions of the concepts using thermodynamics as the appropriate framework for distinguishing the two concepts from each other. In this paper, I challenge this account and argue that self-assembly shouldn’t primarily be understood in terms of thermodynamics. I offer a philosophical analysis of core conceptual challenges for self-assembly and self-organization and argue that the thermodynamic approach follows a misguided conceptualization strategy. Drawing on recent philosophical work on scientific definitions, I propose an alternative account that treats the concept as a conceptual tool that directs our attention towards the kinds of features that are explanatorily relevant for the pattern-formation process and towards the features the resulting pattern itself will possess.

Read the full article at: link.springer.com

Tim Gebbie

Complex systems are often organised into hierarchies whose internal interactions are stronger or faster than interactions across levels. When markets are treated as genuinely multilevel systems it becomes natural to represent them as systems with hierarchical causality. Here we show that reflexivity can be formulated within such a discrete hierarchical causal system; but one in which a higher-level actor state restricts the lower-level transition kernels that remain admissible. Then event dynamics can be separated from calendar embeddings: a set-valued actor-conditioned correspondence can be used to define the admissible family of event kernels, while joint state and waiting-time laws can be used to determine compatible timing to then natural demonstrate reflexivity. A selected event-state law need not determine a unique calendar embedding. Locally, uniqueness of the joint event or timing specification requires uniqueness of both the admissible event kernel and its compatible timing law. Reflexivity is thus the endogenous closure of a hierarchical constraint loop, while timing and projection can generate calendar-time memory or causal ambiguity even for Markov event dynamics.

Read the full article at: arxiv.org

Shamil Chandaria, Arvo Muñoz Morán, Fernando Rosas, Anil Seth, Henry Shevlin, Marcus Hutter, Thore Graepel, Adam Bales, Iulia Comsa, Murray Shanahan, Ruben Laukkonen, Morten Kringelbach, Chris Frith, Shane Legg
The question of AI consciousness is one of the most urgent pre-emptive problems in philosophy and computer science, yet progress is hampered by a cacophony of competing theories that often talk past each other. Separating the hard problem from the mapping problem allows the deepest metaphysical disagreements to be set aside: granting that experience supervenes on a system's organisation, the tractable question becomes at which grain of description that supervenience base sits. We extend Marr's three levels of analysis into a five-level hierarchy of functional descriptions (behavioural, computational, intrinsic causal-structural, organismic, and organism-environment) grounded in supervenience, coarse-graining, and multiple realisability. The major theories of consciousness are positioned within this hierarchy according to which level they take to be critical, and for each level we develop operationalisable indicators and assess current AI systems against them. A Bayesian model then combines theoretical credences with indicator evidence into an overall credence in a system's capacity for consciousness. In illustrative assessments, the verdict for current LLMs is driven as much by where theoretical credence is placed as by how the evidence is read: under different stipulated readings and credence distributions, assessments range from below 0.01 to roughly 0.8, showing sensitivity to assumptions. Finally, the consciousness indicators at each level closely overlap with the architectural features needed for general intelligence, suggesting that increasingly capable AI may become a stronger candidate for consciousness. The framework supports a structured agnosticism, in which theoretical commitments are made explicit, credences are updated as evidence accumulates, and assessments take the form of aggregated probabilities rather than verdicts.

Read the full article at: arxiv.org

Mariana Macedo , Melanie Oyarzun , Cristian Candia , César A Hidalgo

PNAS Nexus, pgag308,

Collective memory—shared information about the past that shapes a group’s identity—is often examined through constructs such as familiarity or salience, conflating its local and global components. Here, we untangle these components using multilingual online attention traces for over 50,000 Wikipedia biographies spanning 137 billion pageviews and a large-scale behavioral recognition experiment in which more than 41,000 participants answered factual questions about famous biographies. Our experiment reveals a robust local memory premium, with people being around 10 percentage points more likely to answer correctly questions about someone from their country and about 5 points more likely to answer correctly questions about biographies with whom they share a language. Finally, we study the impact of three major life events: Nobel Prizes, Oscars, and deaths in local and global online attention, finding these events produce brief but strong surges in local attention and slower but sustained increases in global attention. These findings add important spatial and temporal nuances to our understanding of collective memory.Collective memory—shared information about the past that shapes a group’s identity—is often examined through constructs such as familiarity or salience, conflating its local and global components. Here, we untangle these components using multilingual online attention traces for over 50,000 Wikipedia biographies spanning 137 billion pageviews and a large-scale behavioral recognition experiment in which more than 41,000 participants answered factual questions about famous biographies. Our experiment reveals a robust local memory premium, with people being around 10 percentage points more likely to answer correctly questions about someone from their country and about 5 points more likely to answer correctly questions about biographies with whom they share a language. Finally, we study the impact of three major life events: Nobel Prizes, Oscars, and deaths in local and global online attention, finding these events produce brief but strong surges in local attention and slower but sustained increases in global attention. These findings add important spatial and temporal nuances to our understanding of collective memory.

Read the full article at: academic.oup.com

Asia Maurich Novelli, Sukhwinder Shergill,  Andreia Sofia Teixeira

JMIR Ment Health 2026;13:e99354

People increasingly turn to conversational AI for companionship, emotional support, and well-being, using both purpose-built companion apps, such as Replika and Character.AI, and general-purpose assistants, such as ChatGPT and Claude. While some evidence suggests potential benefits, including short-term reductions in loneliness and mood improvement, several adverse outcomes have been reported in both clinical and nonclinical populations, including emotional dependence, exacerbation of symptoms, and self-harm. The fluent and apparently empathic responses from these models lead users to engage with them not only as tools but also as if they were social entities. This framing is conceptually misleading and may pose risks across different user profiles, particularly for vulnerable individuals. Drawing on research in AI, psychiatry, psychology, and network science, we highlight mechanisms through which emotional reliance develops and the boundary between tool and companion erodes. Design choices that evoke personality and warmth encourage users to anthropomorphize these systems. Simulated empathy, generated through probabilistic language patterns rather than genuine emotional experience, creates a structurally asymmetric interaction, in which the user discloses and the system responds, but without reciprocity, vulnerability, or accountability. Overvalidation and sycophancy can reinforce maladaptive cognitions, delusional ideation, and distorted perceptions of reality, as they tend to reinforce people’s beliefs, even at the expense of the accuracy of models’ responses. These mechanisms are not incidental: they emerge from alignment procedures that reward responses perceived as warm and empathic. The result is a self-reinforcing feedback loop between the model and the user that may amplify maladaptive beliefs, delusional ideation, and emotional distress, even in those who engage for largely functional purposes. Understanding these dynamics requires an examination of both what these agents can do—considering their technical limitations and implementations—and what humans believe they can do, including social and psychological impacts. We argue that conversational AI should be treated primarily as a tool supporting human systems rather than as a substitute for human relationships. Perhaps more importantly, reviewing the current hype surrounding AI interactions can help reformulate a paradigm that contributes to human well-being and societal value, while minimizing misconceptions, maladaptive interactions, or social disintegration.People increasingly turn to conversational AI for companionship, emotional support, and well-being, using both purpose-built companion apps, such as Replika and Character.AI, and general-purpose assistants, such as ChatGPT and Claude. While some evidence suggests potential benefits, including short-term reductions in loneliness and mood improvement, several adverse outcomes have been reported in both clinical and nonclinical populations, including emotional dependence, exacerbation of symptoms, and self-harm. The fluent and apparently empathic responses from these models lead users to engage with them not only as tools but also as if they were social entities. This framing is conceptually misleading and may pose risks across different user profiles, particularly for vulnerable individuals. Drawing on research in AI, psychiatry, psychology, and network science, we highlight mechanisms through which emotional reliance develops and the boundary between tool and companion erodes. Design choices that evoke personality and warmth encourage users to anthropomorphize these systems. Simulated empathy, generated through probabilistic language patterns rather than genuine emotional experience, creates a structurally asymmetric interaction, in which the user discloses and the system responds, but without reciprocity, vulnerability, or accountability. Overvalidation and sycophancy can reinforce maladaptive cognitions, delusional ideation, and distorted perceptions of reality, as they tend to reinforce people’s beliefs, even at the expense of the accuracy of models’ responses. These mechanisms are not incidental: they emerge from alignment procedures that reward responses perceived as warm and empathic. The result is a self-reinforcing feedback loop between the model and the user that may amplify maladaptive beliefs, delusional ideation, and emotional distress, even in those who engage for largely functional purposes. Understanding these dynamics requires an examination of both what these agents can do—considering their technical limitations and implementations—and what humans believe they can do, including social and psychological impacts. We argue that conversational AI should be treated primarily as a tool supporting human systems rather than as a substitute for human relationships. Perhaps more importantly, reviewing the current hype surrounding AI interactions can help reformulate a paradigm that contributes to human well-being and societal value, while minimizing misconceptions, maladaptive interactions, or social disintegration.

Read the full article at: mental.jmir.org

Pablo Villegas, Sandro Meloni

Group interactions are widespread, and higher-order extensions of familiar models display collective phenomena absent from their pairwise baselines, which are routinely offered as evidence of a distinct higher-order physics. We ask whether that claim survives the test that gives new a precise meaning in statistical physics: that of universality. Revisiting canonical higher-order models, we argue that what classifies collective behavior are the infrared ingredients that survive at long scales. Arity is not a universality label. Beyond universality, we examine two further questions: whether higher-order structure is a fact about the system or a choice of description, and what data can and cannot tell us about interaction order. Higher-order descriptions remain indispensable when they expose the organizing structure, provide a better mechanistic language, or improve prediction. We close with what should be measured before new phenomena can be claimed, and where higher-order structure already earns its place.

Read the full article at: arxiv.org

Alex Arenas, Oriol Artime, Albert Díaz-Guilera, Sergio Gómez, Clara Granell

Annalen der PhysikAnnalen der Physik

Volume 538, Issue 9, September 2026, e70288

Many complex systems cannot be understood from network structure alone, nor from continuum descriptions in isolation, because their dynamics emerge from the reciprocal coupling between discrete interaction architectures and spatially extended physical fields. This perspective article surveys mathematical and computational frameworks for such network–field systems, focusing on models in which a graph is either itself the spatial substrate of a field or is embedded in a surrounding medium that mediates transport, signaling, forcing, or spatially distributed hazards. We first introduce a unified formalism for bidirectionally coupled network–field dynamics, emphasizing observation and injection operators that link node variables to continuum processes while preserving balance laws. We then examine two major modeling classes: fields evolving directly on metric graphs, where partial differential equations are posed on network geometries with vertex matching conditions; and hybrid discrete–continuous systems, where node dynamics are coupled to fields in the ambient domain, with particular attention to port-Hamiltonian formulations that provide energy-consistent interconnection principles. Within this common framework, we discuss representative modeling applications in neurobiology, including diffusion-mediated cellular communication and extracellular neural signaling, and in infrastructure systems, where network functionality depends on spatially distributed flows, loads, and hazards. Across these examples, a common picture emerges: the field is not merely an external environment, but an active dynamical layer that reshapes effective interactions, timescales, and collective behavior. By synthesizing concepts that are often developed separately across disciplines, this perspective article aims to clarify the mathematical structure, physical interpretation, and numerical challenges of coupled network–field models, and to highlight their role as a unifying language for spatially embedded complex systems.Many complex systems cannot be understood from network structure alone, nor from continuum descriptions in isolation, because their dynamics emerge from the reciprocal coupling between discrete interaction architectures and spatially extended physical fields. This perspective article surveys mathematical and computational frameworks for such network–field systems, focusing on models in which a graph is either itself the spatial substrate of a field or is embedded in a surrounding medium that mediates transport, signaling, forcing, or spatially distributed hazards. We first introduce a unified formalism for bidirectionally coupled network–field dynamics, emphasizing observation and injection operators that link node variables to continuum processes while preserving balance laws. We then examine two major modeling classes: fields evolving directly on metric graphs, where partial differential equations are posed on network geometries with vertex matching conditions; and hybrid discrete–continuous systems, where node dynamics are coupled to fields in the ambient domain, with particular attention to port-Hamiltonian formulations that provide energy-consistent interconnection principles. Within this common framework, we discuss representative modeling applications in neurobiology, including diffusion-mediated cellular communication and extracellular neural signaling, and in infrastructure systems, where network functionality depends on spatially distributed flows, loads, and hazards. Across these examples, a common picture emerges: the field is not merely an external environment, but an active dynamical layer that reshapes effective interactions, timescales, and collective behavior. By synthesizing concepts that are often developed separately across disciplines, this perspective article aims to clarify the mathematical structure, physical interpretation, and numerical challenges of coupled network–field models, and to highlight their role as a unifying language for spatially embedded complex systems.

Read the full article at: onlinelibrary.wiley.com

Michael Levin

Philosophies 2026, 11(5), 161

I argue that the emerging sciences of synthetic morphology and diverse intelligence suggest non-physicalist models of mind and show how they can be empirically investigated. Whence the anatomical, physiological, molecular-biological, and behavioral properties of engineered new beings that have never before existed, and do not have a history of selection? Understanding, predicting, and guiding new forms of life and mind requires characterizing a structured latent space of patterns. Developmental, synthetic, and behavioral biology should take seriously, and exploit, the kinds of non-physicalist ideas that are already a staple of Platonist mathematics. I propose the following hypotheses. (1) Patterns in this space span a highly variable degree of agency, comprising a spectrum ranging from static truths studied by mathematicians to active ones studied by behavioral scientists (i.e., some patterns on the same spectrum as mathematical truths are kinds of minds). (2) The relationship between mind and body is the same as the relationship between causally instructive mathematical facts and physics. (3) Living beings have no monopoly on the “free lunches” provided by the ingression of these patterns into the physical world. While traditional computationalist views of living and cognitive systems are insufficient, my framework erases artificial distinctions between organisms and machines, framing all physical constructs (natural or engineered) as being, to various degrees, in-formed by patterns from the latent space. I sketch a research program, already begun, inspired by these ideas. Such frameworks, while contradicting long-held assumptions of both mechanists and organicists, could have many implications for evolutionary biology, regenerative medicine, AI, and the ethics of synthbiosis with the forthcoming immense diversity of morally important beings.

Read the full article at: www.mdpi.com

Marta Zava

Training a frontier artificial intelligence model costs hundreds of millions of dollars before a product exists, and few investors can write that cheque. This paper shows that the number who can falls as the fixed cost of a first training run rises, at a rate set by the concentration of the fund size distribution, so that doubling the cost removes about two thirds of the possible backers. The firms that compete are therefore selected by a small group of allocators before any customer has expressed a view. Three results follow. The price paid by the winner separates into a rational bid, a premium created by the fund's deployment clock, and an error from failing to adjust for adverse selection, and the sign of the price response to competition identifies which one dominates. Concentration trades breadth for depth, improving the funded set only when the skill advantage of the few outweighs the information lost through having fewer independent views, a loss that is small when investors think alike. And large upfront cheques reduce the value of stopping, so ventures funded under deployment pressure should fail later and larger rather than more often. The effective policy margin is access to compute rather than regulation of the capital market.Training a frontier artificial intelligence model costs hundreds of millions of dollars before a product exists, and few investors can write that cheque. This paper shows that the number who can falls as the fixed cost of a first training run rises, at a rate set by the concentration of the fund size distribution, so that doubling the cost removes about two thirds of the possible backers. The firms that compete are therefore selected by a small group of allocators before any customer has expressed a view. Three results follow. The price paid by the winner separates into a rational bid, a premium created by the fund's deployment clock, and an error from failing to adjust for adverse selection, and the sign of the price response to competition identifies which one dominates. Concentration trades breadth for depth, improving the funded set only when the skill advantage of the few outweighs the information lost through having fewer independent views, a loss that is small when investors think alike. And large upfront cheques reduce the value of stopping, so ventures funded under deployment pressure should fail later and larger rather than more often. The effective policy margin is access to compute rather than regulation of the capital market.

Read the full article at: papers.ssrn.com