Embracing Complexity: Abstracts

Below, you will find the abstracts submitted for the workshop "Embracing Complexity: Principled and Practical Approaches to Emergence".

Beyond Reduction and Emergence: A Distributed View of Minds

Antonella Tramacere (Università Roma Tre, Italy)

This talk presents a distributed view of mental systems, challenging both reductionist and overly broad emergentist frameworks. Drawing on key cases from cognitive neuroscience, I propose a diachronic and integrative perspective where causal control over system behavior is dynamically distributed across internal and external components. Mental phenomena arise through continuous interactions across components and processes, dissolving strict ontic boundaries between levels and challenging distinctions between internal and external influences. Methodologically, this view encourages interdisciplinary approaches that incorporate temporal, environmental, and developmental dimensions of living systems, while also calling for criteria for defining what constitutes a mental system.

Antonella Tramacere is a philosopher of mind and biology specializing in the evolution of cognition, agency, and consciousness. Her research spans multiple levels of explanation, integrating insights from evolutionary biology, cultural evolutionary theories, human and comparative psychology, molecular biology, and phenomenology.
She critically analyzes internalist and externalist frameworks of cognition, where cognitive processes are often attributed either to internal factors (such as genes and neurons) or external ones (such as the environment and the body). Challenging this dichotomy, she advocates for a more nuanced, non-binary approach that remains analytically rigorous while recognizing the organizational role of action in structuring low and higher-level features of cognition.
For more information, visit: https://www.researchgate.net/profile/Antonella-Tramacere?ev=hdr_xprf
Social media: [https://x.com/AntonellaTrama3].

Advanced Methods in Nonlinear Time Series Analysis

Julia Mindlin (University of Leipzig, Germany)

This course provides an in-depth exploration of advanced techniques in nonlinear time series analysis. A core component of the course covers time delay embeddings, based on Takens’ theorem, for reconstructing phase space dynamics from scalar observations. We introduce methods for selecting optimal embedding parameters, and imparticular work with an implementation of the false neighbors method, ensuring minimal self-intersections in reconstructed attractors. To validate dynamical models that can reproduce the behaviour of the experimental time series data, we explore topological analysis of periodic orbits through linking and self-linking numbers (LN, SLN), a method for comparing model-generated and observed attractor structures. We then transition to complexity measures, introducing Shannon entropy, disequilibrium metrics, and the Bandt-Pompe permutation entropy framework to quantify information content and system disorder.

The course includes hands-on computational exercises in Python, emphasizing practical implementation of these methods using real-world application from the climate system. We will work examining the Niño-3.4 region's sea surface temperature (SST) variability using daily data reconstruction and spectral filtering methods. We will discuss how to infer appropiate sampling rates and delays for the time delay embedding based on this particular application. By the end, participants will be equipped with the theoretical foundations and coding skills to analyze nonlinear time series in climate and other complex systems. This will allow participants to approach more complex methods (i.e. machine learning) recently proposed to reconstruct phase spaces or find dynamical models from data. Finally, we will provide a short overview of relevant literature to further explore these analysis tools.

Dr. Julia Mindlin is an Argentinian geophysicist specializing in climate dynamics and large-scale atmospheric circulation. She earned her Ph.D. from the University of Buenos Aires (UBA), where she investigated the Southern Hemisphere’s extratropical circulation response to anthropogenic forcing using storyline approaches. Currently on leave from her position as an Assistant Professor at UBA, she is a postdoctoral researcher at the Leipzig Institute for Meteorology, where she integrates causal inference frameworks with storyline methods to study large-scale teleconnections and their influence on extreme events. In parallel, she is interested in dynamical systems modelling and has worked on the characterization of El Nino Southern Oscillation as a chaotic system. She has worked at the University of Reading (UoR), the Barcelona Supercomputing Center (BSC) and the International Center for Theoretical Physics (ICTP) as a visiting scientist and received numerous accolades for her research, including the prestigious Microsoft PhD Research Fellowship (2021), the WMO Professor Mariolopoulos Trust Fund Award (2024), and a Marie Curie Postdoctoral Fellowship (2025). She is actively involved in the climate research community, co-chairing the Fresh Eyes for CMIP steering group and serving on the Executive Committee of the Young Earth System Scientists (YESS) community. Her research aims to enhance understanding of climate variability and improve predictive models by combining innovative statistical techniques with physical climate science. She's a trumpet enthusiast and loves punk concerts.

Together, but not the same: self-organisation towards emergent collectives

Madalina Sas (Imperial College London, UK)

We are living in a time of social alienation and political division, where many people feel disconnected from the leadership structures who fail to represent their interests; but in the natural world of eusocial insects and social animals, such inefficient social structures would not survive.

In my talk, I will try to show how the kind of self-organising behaviour seen in animals can be tremendously beneficial to humans. I will draw from studies of swarm intelligence in the natural world, starting with how birds flock, and ants vote, and slime moulds solve mazes, and discuss how these behaviours relate to behaviour in humans, by comparison with different frameworks in philosophy, anthropology, and sociology. Then I will present quantitative studies of self-organised behaviour in humans, and relate it to important collective activities such as rituals, protests and marches, art and sport, but also important modes of interaction to foster such collective activities, such as joint action and improvisation.

Finally, I present Synch.Live, a novel participatory experimental technology and collective artistic experience inspired from swarm intelligence. Its goal is to induce collective behaviour in human groups, study the conditions required for self-organisation to emerge, as well as the balance between individual and collective.

Madalina Sas an interdisciplinary scientist and artist from Romania, currently based in London, UK. She has a computer engineering master’s from Imperial College, and have recently completed a PhD in complex systems at Imperial’s Centre for Complexity Science. Her main scientific interest is in the phenomenon of emergence, where multiple components or agents in a complex system create something new and unexpected from their local interactions: "the whole is greater than the sum of its parts".

Using an emergentist framework, She study self-organising collective behaviour in mathematical simulations, swarm intelligence, and complex human activities. Her research combines two main aspects: the quantitative and scientific focuses on data-agnostic means to measure collective behaviour in any system, at any scale, and the relationships between individuals and the group; while the qualitative and artistic focuses on human collective experience, cooperation and improvisation. The union of these two aspects allows she to work with scientists and artists and participate in devising experiments which blend scientific inquiry, technological invention and artistic experience. Contact: website only, https://mis.pm

Large-scale integration of perceptual and predictive information is encoded by non-oscillatory neural dynamics

Andres Canales (University of Cambridge, UK)

 The brain is characterized by extensive recurrent connectivity within and between areas. This recurrent connectivity enables various patterns of arrhythmic (non-oscillatory) and rhythmic (oscillatory) neural activity that are temporally coordinated between regions. What role do these distinct dynamics play in the large-scale integration of perceptual and predictive information? In this talk, I will discuss how information theory combined with EEG , ECoG, and computational modelling can help us uncover large-scale neural patterns of non-oscillatory activity during perception and prediction. In the first series of studies, I will show how non-oscillatory rather than oscillatory dynamics encode perceptual and predictive information across sensory modalities in different species. In the second part, I will discuss how non-oscillatory dynamics encode synergistic (complementary) rather than redundant (common) information between brain areas during visual and auditory predictive processing. These empirical and theoretical observations will provide new insights into the functional role of non-oscillatory dynamics during the large-scale integration of perceptual and predictive information.

Andres Canales an Affiliated Lecturer at Cambridge's Psychology Department and one of the principal investigators in the Consciousness and Cognition Lab. He received a Biology degree from Universidad de Valparaiso (Chile), and a PhD in Biological Sciences from the University of Cambridge (United Kingdom). His part of the lab combines non-invasive (EEG) and invasive (ECoG) electrophysiological recordings with cutting-edge tools from information theory, spectral analyses, and computational modeling to investigate consciousness and cognition in human and animal models. His research is highly collaborative and interdisciplinary, currently involving collaborations with European and US laboratories regarding the study of perception and consciousness and South American labs about the neurocognitive effects of poverty and social vulnerability.

Are human coordination dynamics embedded in social network structures?

Ivana Kovanlinka (Technical University of Denmark, Denmark)

Ivana Konvalinka is an Associate Professor at the Section for Cognitive Systems, Technical University of Denmark (DTU Compute), and PI of the Social Interaction and Neuroscience (SINe) lab. She has a background in electrical/biomedical engineering (BASc/MSc) and neuroscience (PhD). She is interested in the intra- and interpersonal neural and behavioural mechanisms that underlie and shape social cognition and interaction, and how these are modulated by higher level social factors such as social network properties and social ties. Specifically, she works on developing experimental and computational tools for quantifying multi-person interaction dynamics across behavioural, physiological, and neural signals, in controlled lab-based and ecological real-world settings.

Emergence in the Life Sciences

Lauren Ross (University of California, Irvine, USA)

This talk examines emergence in the life sciences, with a focus on examples of paradigmatic cases, how these cases are explained in scientific contexts, and what makes them distinct from non-emergent examples. This work suggests that clear conceptions of emergence are supported by keeping considerations of (i) physicalism distinct from (ii) scientific explanation, including the principles, reasoning, and guidelines that support such explanations. In this analysis, emergence is related to systems science approaches, complexity considerations, and it is contrasted with common reductionist frameworks in science.

Lauren N. Ross PhD, MD is an Associate Professor in Logic and Philosophy of Science at the University of California, Irvine. Her research concerns causal reasoning and explanation in the life sciences, primarily neuroscience and biology. One main area of her research explores causal varieties---different types of causes, causal relationships, and causal systems in the life sciences. Her work identifies the features characteristic of these causal varieties and their implications for how these systems are studied, how they figure in scientific explanations, and how they behave. A second main area of work focuses on types of explanation in neuroscience and biology, including distinct forms of causal and noncausal explanation. Ross’ research has received a National Science Foundation (NSF) CAREER award, a Humboldt Experienced Researcher Fellowship, a John Templeton Foundation Grant, and an Editor’s Choice Award at The British Journal for the Philosophy of Science. Recent publications include a paper on “Causation in Neuroscience: Keeping Mechanism Meaningful” with Dani Bassett in Nature Reviews Neuroscience and book on Explanation in Biology (Cambridge University Press: Elements Series).

Emergence through Dynamical Closure

Borjan Milinkovic (NeuroPSI, France)

Emergence manifests in a variety of forms. Contemporary operational approaches reflect this diversity by employing a range of distinct formalisms. One such approach focuses on identifying closure within subsystems of a larger dynamical system composed of many interacting components. This session introduces the concept of closure and examines how it has been formalised across various domains. We will explore operational closure in biological systems, information closure in information-theoretic systems, and dynamical closure in both dynamical and statistical systems. Emphasis will be placed on drawing conceptual connections between these often disparate theoretical frameworks. The session will further speculate on how regularities between components enable closure to coarse-grain the sub-processes that define a system, thereby offering a dynamics-first perspective of emergent phenomena. We conclude by considering how information theory may provide a promising foundation for a general theory of emergence via dynamical closure.

Tipping points in the assembly and disassembly of life across scales of biological organization

Jordi Bascompte (University of Zurich, Switzerland)

I will start my talk by presenting a brief account of the history of life on earth highlighting two major evolutionary transitions: the origin of the complex cell, which paved the path towards the posterior evolution of multicellular organisms, and the rise of complex ecological communities. I will argue that the origin of the eukaryotic cell can be understood in terms of an algorithmic evolutionary phase transition. Protein-based gene regulation faced a limit as finding longer proteins become computationally unfeasible. Evolution escaped such a limit to complexity---while maintaining a conserved mechanism of gene growth---by shifting to a new type of genetic regulation based on non-coding DNA sequences. I will then move to a larger scale of biological organization, that of ecological communities, by focusing on their resilience in the face of anthropogenic influences. I will describe ecosystem shifts to alternative states, mechanisms that can delay such shifts by increasing community resilience, and to what degree such shifts can be predicted by generic early-warning signals. Using these two examples, I will advocate that both the assembly and the disassembly of life on earth follow sudden transitions with common statistical properties.

Jordi Bascompte is Professor of Ecology at the University of Zurich and Director of its Specialized Master on Quantitative Environmental Sciences. He is mostly well-known for having brought the interactions of mutual benefit between plants and animals into community ecology, at the time largely dominated by predation and competition. His application of network theory to the study of mutualism has identified general laws that determine the way in which species interactions shape biodiversity. Jordi is one of the most highly cited scientists according to Thompson Reuters. Among his distinctions are the European Young Investigator Award (2004), the Ecological Society of America’s George Mercer Award (2007), the Spanish National Research Award (2011), the British Ecological Society’s Marsh Book of the Year Award (2016), and the Ramon Margalef Prize in Ecology (2021). Recipient of an ERC’s Advanced Grant, Jordi has served in the Board of Reviewing Editors of Science and has been the Ideas and Perspectives Editor at Ecology Letters. Among his books are Self-Organization in Complex Ecosystems (with R.V. Solé) and Mutualistic Networks (with P. Jordano), both published by Princeton University Press.

On the Emergence of a Hierarchy of Timescales in the Brain: Reconciling Criticality with Temporal Processing

Leonardo Gollo (IFISC, Spain)

The brain exhibits a hierarchy of timescales, from rapid neural responses to slower cognitive processes, which we propose arises from regions operating at varying distances from criticality. Slower regions are positioned closer to the critical point, where critical slowing down prolongs recovery times, while faster regions function in subcritical regimes, allowing rapid activity dissipation. This structured distribution of critical and subcritical dynamics enables the brain to balance sensitivity and stability, supporting both adaptive responsiveness and robust internal processing. By integrating nonlinear dynamics and criticality theory, our framework provides a mechanistic explanation for the emergence of hierarchical timescales, offering new insights into neural computation, cognition, and the brain's functional organization, with implications for understanding neurological disorders linked to disrupted temporal processing.

Dr. Gollo is a leading expert in modeling neuronal dynamics, with a distinguished track record in uncovering multiscale phenomena spanning subcellular, neuronal, circuit, large-scale, and whole-brain levels. As a Ramón y Cajal Fellow, he applies advanced quantitative methods from physics, complex systems, dynamical systems, and computational neuroscience to unravel the intricate patterns of brain activity. His research combines theoretical modeling and non-invasive neuroimaging to advance our understanding of neural function, criticality, and large-scale brain dynamics.

Emergence in language: Theoretical models and data analysis

David Sánchez (IFISC, Spain)

Language is a complex adaptive system where individual interactions, cognitive constraints, and social structures collectively give rise to new linguistic patterns. In this talk I will argue that this emergence in language can be examined from a twofold perspective. Whereas intrinsic emergence explains how language internally evolves, extrinsic emergence explains how language shapes and is shaped by speech communities. The former is useful to understand patterns of word usage, the emergence of syntactic rules through grammaticalization and semantinc drifts due to sociocultural changes. The latter is key to the development of speech communities that share linguistic norms or show departures from standard rules (dialects, sociolects) and the vitality of language depending of its prestige the speakers' linguistic ideology. I will emphasize how modern tools such as the analysis of social media large datasets sheds light on the mutual influence between micro-level changes and macro-level variation.

Emergence of collective learning in coupled neural networks

Lucas Lacasa (IFISC, Spain)

In this talk we aim to discuss whether more is different in AI. We will discuss how collective learning can emerge in systems of coupled neural networks, or swarms of interacting “brains”, based on a recent work with Lluís Arola-Fernández [1]. To that aim, we will introduce a minimal assembly model of coupled brains -where each brain is a deep artificial neural network model- and the brains are coupled in a parameter-to-parameter way. The dynamics of the assembly incorporates the competition between two terms: the local learning dynamics in the parameters of each neural network unit, and a diffusive coupling that tends to homogenize the parameters of the assembly.  Each individual neural network is trained solely on their private dataset (e.g. only images of a specific digit within the MNIST dataset), and performance of each unit of the assembly is then assessed for a test set that includes all data (e.g. images from all digits in the MNIST example). Above a certain coupling threshold, we show how each brain undergoes a transition between a phase where they can only predict images similar to the ones they were trained on, to a phase where brains can accurately predict any kind of image. We argue that this universal generalisation is underpinned by collective learning.

We then derive the low-dimensional behavior of our assembly via an effective theory for deep linear networks which turns out to be formally equivalent to the paradigmatic (zero temperature) Ginzburg-Landau model subject to quenched disorder. Our theory thus predicts disorder-order phase transitions in the parameters’ solutions and the emergence of collective learning –where each individual neural network unit is able to learn from each other, leading to high performance of every individual unit when tested on the full test set– at a critical value of the coupling strength. We further unveil the effects of data privacy, regularization strength and network depth in the phase transitions and validate our predictions with numerical experiments, training an assembly of coupled nonlinear network models in the MNIST and CIFAR-10 datasets.  In some sense, our results advance the theory of deep learning in decentralized settings and bridge a research gap between machine learning and the physics of complex systems.

We will also briefly discuss implications of these results for the AI alignment problem.

Lucas Lacasa is a tenured scientist (Investigador Científico) at CSIC and works at the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC, CSIC-UIB). Before, he has had permanent positions at the School of Mathematical Sciences, Queen Mary University of London (UK, 2013-2021) and at the Technical University of Madrid (Spain, 2010-2013). He has been a visiting scientist at Oxford University and UCLA, among others. His research spans methodological aspects of complex systems, involving nonlinear dynamics, time series, machine learning and network science. He is well known for proposing visibility graphs as a graph-theoretical representation of time series. Lately he has been interested in using complexity science to understand AI, e.g. by using the framework of temporal network theory to describe the training process of deep neural networks, or by assessing whether more is different when several machine learning models are put in interaction.

His research interfaces several disciplines, having published over 80 papers in generalistic venues (PNAS, Nature Communications), physics (PRX, PRL), computer science (IEEE TPAMI) or applied mathematics (Nonlinearity). He has received several prizes, including the International USERN Prize in Formal Sciences (2019) or QMUL Research Excellence prize (2020).

Pedro Mediano

Long ago, when the temperature of the universe was around 108 K, I did a Physics degree at the University of Valencia, Spain, truncating what could have otherwise been a promising career as a jazz melodica player. When I decided that neutrinos were interesting but not quite for me, I did a PhD at Imperial College and a postdoc at the University of Cambridge, where I (rather unsuccessfully, in all fairness) tried to solve the mysteries of consciousness and complexity. At the moment I continue to fail to solve consciousness and artificial intelligence as a lecturer at the Department of Computing, Imperial College London.

At Imperial, I lead the Machine Intelligence and Neural Dynamics (MIND) lab. Broadly speaking, our research lies at the intersection between consciousness science, information theory, complexity science, and machine learning. Our goal is to understand how a complex system composed of interacting parts can give rise to cognition and consciousness – with implications for both artificial intelligence and mental health.

Signatures of criticality in the collective decisions of social animal groups

Carmen Miguel (Universitat de Barcelona, Spain)

Collective decision-making is a self-organized process where a group of individuals, each with their own decision mechanisms, reaches a common agreement. This phenomenon occurs across a wide range of biological and artificial systems, from human elections to schooling fish, social insect colonies, and robot swarms. In animal behavior, the study of collective motion and consensus has been greatly influenced by the diversity of animal signals, which play a key role in understanding how animals communicate and make decisions as a group. Social insects have become prime examples of collective behavior. Despite their simple individual actions, they demonstrate complex, emergent decision-making as a group. I will present a decision-making model inspired by honeybee house-hunting behavior, which incorporates individual exploration, the capability to assess and advertise options' qualities, and several forms of social interactions. Our model adheres to Weber's Law of relative stimulus perception and exhibits critical transitions under certain conditions, where consensus reaches its highest values. The connection between criticality and optimal decision-making is common in biological systems and offers valuable insights into how artificial adaptive systems could make effective decisions when facing similar options.

Full Professor of Condensed Matter Physics at the Universitat de Barcelona since 2020, where I also graduated in Physics in 1991, and received a PhD in Statistical and Condensed Matter Physics in 1995. Postdoctoral researcher at the Massachusets Institute of Technology in Cambridge, USA, and at the Abdus Salam International Center for Theoretical Physics in Trieste, Italy. Appointed visiting or invited professor at the Università La Sapienza in Rome (Italy), the Université d´Orsay in Paris (France), the Kavli Institute for Theoretical Physics in Santa Barbara (USA), the Helsinki University of Technology (now Aalto University, Finland), the ISI Foundation in Turin (Italy), Indiana University, Bloomington (USA), and Northeastern University, Boston (USA). Previous positions: Associate Professor (2006-2020), Ramón y Cajal research associate (2001-2006) and Lecturer (2000-2001) at the University of Barcelona. Awarded the Distinció of the Generalitat de Catalunya for “young scientist” in 2004, and the I3 Program of the Spanish MICINN in 2008.