Schedule

Acreditación al evento

Jesús Gómez Gardeñes

"Ni contigo ni sin ti": Contagion at the Rhythm of Real Groups 

Complex contagion relies on collective reinforcement: adoption is driven by simultaneous exposure to several adopters within a group. Higher-order models capture this mechanism, but almost always assume interaction structures that are either frozen or infinitely fast-mixing. Neither with them nor without them: static groups cannot sustain adoption at all, while infinitely fast reshuffling dissolves the very reinforcement that spreads it. Here we show that the dynamics of group composition is not a correction to these limits but a primary mechanism shaping higher-order contagion. Using an approximate master-equation framework that continuously interpolates between the quenched and annealed regimes, we find that group turnover renders the invasion threshold non-monotonic, revealing an optimal mixing rate, and can also change the order of the phase transition and widen the bistable region. Once realistic group-size heterogeneity is included, temporality, nonlinear reinforcement and mesoscopic localization combine to yield multiple coexisting active states that differ in where activity resides. Extending the framework to adaptive turnover shows that not only how fast groups reorganize, but how and why individuals switch, qualitatively alters critical behaviour. 

Silvio Ferreira

Quantifying information gaps and the coupled dynamics of media and public agendas in social networks 

Media and public agendas co-evolve through interactions between institutional information sources and networked users, shaping collective attention and public discourse in online environments. Despite extensive research on information diffusion and agenda-setting, quantitatively disentangling topics promoted by mass media from those emerging through public interactions remains challenging. Here, we introduce a framework to jointly reconstruct and characterize media and public agendas in social networks. Moving beyond conventional analyses centered on individual hashtags, our approach combines topic co-occurrence networks with user interactions to identify shared and agenda-specific topics, quantify information gaps and topic dominance, and characterize temporal delays between media and public attention. We apply the framework to large-scale Twitter (now X) datasets collected during the 2019 Argentine and 2022 Brazilian presidential elections. We find that media-related topics tend to occupy the structural core of the information ecosystem, whereas topics specific to the public agenda are typically more fragmented and peripheral. Moreover, the coupling between the two agendas is strongly heterogeneous: some topics display synchronized media and public attention, while others exhibit substantial temporal delays or remain largely confined to one agenda. These results indicate that agenda-building in social media is not a homogeneous process but emerges from the competition and interaction of multiple information dynamics. Our framework provides a quantitative approach for investigating information gaps and the coupled evolution of institutional media and public attention, with potential applications to a broad range of digital communication systems. 

Albert Díaz Guilera  

Topology, Attractors and Collective Dynamics in Directed Modular Neuronal Networks 

Directed modular neuronal networks exhibit a rich repertoire of collective dynamical states whose relationship with the underlying network topology remains an open problem. We explore this relationship from the perspective of the attractor landscape associated with state-transition dynamics. We first show that the deterministic attractors extracted from the network topology provide a useful framework for interpreting the dynamics of both stochastic state-transition models and biophysically realistic spiking neural network simulations. Rather than focusing solely on individual trajectories, we analyze the probability distributions over the relevant dynamical states and compare different dynamical descriptions of the same network. This perspective highlights the role of attractors and their basins as an intermediate level of description linking network structure and collective dynamics. Finally, we discuss how structural properties of directed networks may shape the organization of these probability distributions and how this framework can help explain global dynamical observables in modular neuronal networks. 

Coffee Break

Hernán Makse

From cells to brains: symmetry fibrations for biological and artificial neural networks 

Symmetry is the cornerstone of theoretical physics, yet it remains silent to explain (artificial) life, despite life emerging from physical law. Here, we attempt to bridge the gap between physics, biology and AI using symmetries, but with a twist. The traditional symmetry groups of physics are global and too rigid to describe biological networks. Instead, the notion of symmetry fibration derived from Grothendieck’s fibrations in category theory, is local, flexible, and adaptable to evolutionary pressures, providing a framework for understanding biological networks but also how AI works through symmetry and symmetry breaking in inference and learning. On the application front, fibrations compress LLMs up to 10x in parameter size with no loss of performance, providing a plausible solution to the impending collapse of the scaling era. 

Sebastián Goncalves

Exploratory Characterization and Seasonal Analysis of the Guaíba River Level Time Series 

We present an exploratory analysis of the Guaíba River level time series in Porto Alegre, Brazil, aimed at systematically characterizing its dynamics and investigating the possible separation of components operating at different time scales. The dataset, consisting of high-frequency observations from 2018 to 2026, is first examined in terms of data quality, level distributions, and missing-data intervals. Smoothed versions of the time series are also analyzed using moving averages over different temporal windows.

To investigate the presence of seasonal structure, the observations are subjected to an annual folding procedure, in which measurements from different years are represented along a common calendar axis. This representation reveals a consistent annual pattern, with higher water levels occurring predominantly during the austral winter. An empirical seasonal profile is then constructed from years considered representative of normal conditions and subsequently described using a periodic annual harmonic expansion.

A particularly relevant result is obtained by applying the seasonal profile, without recalibration, to the year 2024, which was characterized by an exceptional hydrological event. The model provides a reasonable description of the background evolution of the river level, while the extraordinary flood appears as a clear deviation from the seasonal component. These preliminary results suggest that the Guaíba dynamics can be investigated through a decomposition of the form h(t)≃S(t)+R(t), where S(t) represents the seasonal component and R(t) accounts for higher-frequency fluctuations, trends, and extreme events. 

Celia Anteneodo

Class emergence and inequality in an agent based model 

We analyze inequality through the dynamics of an agent-based model of capitalist economy, known as Social Architecture of Capitalism, introduced by Ian Wright [1]. The model contemplates two main types of agents, workers and capitalists, which may also be unemployed. Starting from an egalitarian initial condition, in which all agents are unemployed and possess the same initial wealth, the system self-organizes into two distinct economic classes: capitalists and workers/unemployed. After a short transient, the dynamics attains a state that reproduces several statistical properties of real economies worldwide, most notably the two-regime distributions of wealth and income.

Through extensive simulations, we investigate how the model parameters (including number of agents, total wealth, and salary range) affect the resulting distribution of wealth and income, the social distribution of agents, social stratification and other socioeconomic properties [2-3]. One of our main findings is that, under wealth conservation and fixed average wages,

increasing wealth per capita amplifies socioeconomic inequality.

Partial financial support by Capes, CNPq and FAPERJ is acknowledged.

[1] I. Wright, The social architecture of capitalism, Physica A 346, 589 (2005).

[2] J. S. Borba, S. Gonçalves, and C. Anteneodo, Inequality in a model of capitalist economy, Physica A 664, 130457

(2025).

[3] J. S. Borba, C. Anteneodo and S. Gonçalves, and Can rising consumption deepen inequality? The European Physical Journal Special Topics, 1-11 (2026). 

Almuerzo 

Marcelo Kuperman

Collective Spatial Organization with Cultural Convergence 

We present a novel theoretical framework that unifies Axelrod's model of cultural dissemination with the spatial self-organization of swarmalators. Traditional models of social influence operate on static networks or fixed spatial lattices, ignoring how cultural alignment dictates physical movement. Conversely, classical swarmalator systems couple spatial mobility with continuous oscillatory phases, failing to capture discrete, non-periodic cultural traits. Our model resolves this divide by embedding agents with discrete cultural feature vectors into a continuous two-dimensional domain. Spatial forces are governed by cultural homophily, where similarity increases physical attraction, while local cultural updates occur probabilistically among spatially proximate agents under a neighborhood majority rule. Numerical simulations reveal that this bidirectional feedback gives rise to rich emergent spatio-cultural phase regimes, spanning compact monocultural clusters, spatially segregated multicultural enclaves, and dynamic fragmented swarms. 

Orlando Billoni

Chess as a complex system

Chess is one of the most widely played and challenging games in the world. This ancient game is also interesting from a scientific perspective, as it has historically served as a testbed for the development of computer science and artificial intelligence.

From an interdisciplinary perspective, chess is relevant to the social and cognitive sciences, where it has been used to study decision-making,

expertise, population-level learning, and mechanisms of human memory, among other topics.

The large number of games recorded on online platforms and in chess databases has stimulated the statistical analysis of chess,

revealing behaviors characteristic of complex systems. For example, Zipf’s and Heaps’ laws—both ubiquitous in human language—have

also been observed in the distribution of chess positions. Memory effects have been identified in chronological sequences of games

played by groups of players, while bursty dynamics have been detected in the activity patterns of individual players.

More recently, we have studied the statistical properties of chess in connection with decision-making processes.

In this talk, I will present and discuss results concerning the statistical properties of chess, with particular

emphasis on innovation, memory effects, and decision-making processes. 

Pablo Balenzuela

Quantifying Collective Dynamics in Social Media Across Three U.S. Elections 

Understanding how collective attention emerges and organizes itself in digital environments is a central problem in the study of complex social systems. This talk presents a computational framework for quantifying the coupling between political actors' discursive agendas and the collective dynamics of public engagement on social media. Using approximately one billion Facebook interactions across the 2016, 2020, and 2024 U.S. presidential elections, we combine large language models to classify topics, subtopics, and ideological stance in posts and campaign speeches, multilayer network analysis to capture zero-lag and lagged temporal correlations between candidate discourse and public reaction, and a hierarchical Bayesian model to infer the geographic distribution of engagement from advertising exposure data. This layered approach reveals a stable core of recurring issues around which collective attention organizes across cycles, together with structural asymmetries in how tightly different actors' agendas couple to that collective dynamic asymmetries that sharpen considerably by 2024. Beyond the specific electoral findings, the talk focuses on the methodological choices underlying this framework and discusses how the same architecture could be extended to study collective dynamics in other platforms, domains, and real-time settings. 

Coffee Break + Poster Session

Ana Amador

Emergent dynamics and neural-behavioral coordination in birdsong 

Complex behaviors emerge from the coordinated dynamics of neural populations and the body across multiple timescales. A central challenge is to identify the organizing principles underlying this coordination and to determine how population-level neural dynamics relate to ongoing behavior.

In this talk, I will present birdsong as a model system for addressing these questions. Canary song is organized into rhythmic sequences spanning a broad range of timescales, providing a unique opportunity to investigate the relationship between neural population activity and vocal behavior. Using simultaneous recordings of neural activity, respiratory dynamics, and vocal output during natural song production, I will discuss two complementary approaches to uncovering structure in these complex signals.

First, using unsupervised dimensionality reduction, we find that neural population activity can be represented by a small number of collective dynamical variables. These latent dynamics exhibit robust oscillatory organization, with frequencies closely matching both the rhythmic structure of song and the underlying respiratory motor gestures. Second, moving from population-level organization to the moment-to-moment coordination between neural activity and behavior, we use a phase-resolved cross-correlation framework to identify transient neural–vocal interactions. This analysis reveals preferred temporal relationships, including neural activity preceding, accompanying, or following vocal output, suggesting that neural–behavioral coordination is organized around distinct temporal regimes rather than a single characteristic delay.

Together, these results reveal complementary levels of dynamical organization: low-dimensional collective dynamics capture the global temporal structure of behavior, while local phase relationships uncover how neural and behavioral signals coordinate in time. More broadly, they illustrate how combining data-driven dimensionality reduction with minimally processed time-series analyses can reveal organizing principles in complex biological systems. 

Jordi Soriano Fradera

Neuronal Cultures to Explore Complex Systems in Action 

Neuronal cultures consist of dissociated neurons that self-organize into de novo networks within a few days. Experimentally, they can be interrogated and manipulated in a wide variety of ways, effectively shaping a true laboratory for probing and modeling living neuronal networks. In this talk, I will provide an overview of neuronal cultures and discuss experimental approaches that are especially relevant to the complex systems community, including the engineering of connectivity, resilience to damage, and plasticity. I will also address the emerging frontier of biological artificial intelligence and its associated challenges. 

Enzo Tagliazucchi

Computational models of whole-brain activity across health and disease 

The human brain is a complex system wherec rich and unpredictable dynamics emerge constrained by the backbone of anatomical connectivity. In this talk, we show that computational models can contribute to the understanding of how these emergent  patterns of neural activity underlie human consciousness and cognition, and how their alterations can lead to pathological states. We focus on the application of whole-brain models to different but related domains: global states of consciousness, neurodegenerative diseases, and human aging trajectories, unconvering their shared dynamical and biophysical mechanisms. 

Coffee Break

Mario Chávez

Representations of Brain Networks

Whole-brain neuroimaging and neurophysiological data provide descriptions of brain structure and function, which can be transformed into discrete mathematical models such as brain networks. In turn, both neuroimaging data and brain networks can be embedded into low-dimensional feature or geometric spaces that capture different properties of the system’s state. In this talk, I’ll briefly present existing methods for generating low-dimensional brain representations and discuss their applications in cognitive and neurological research.

Gabriel Mindlin

Learning to Sing in the Dark: Neural Networks, Vocal Dreams, and the Role of Sleep

How does a neural network learn to generate the coordinated motor patterns required for a complex behavior? Birdsong provides a powerful model for addressing this question. During sleep, neural activity related to song is reactivated and reaches the respiratory and vocal muscles, although the bird remains silent. By combining these motor signals with dynamical models of avian phonation, we can reconstruct the sounds associated with this nocturnal activity.

In birds that learn their songs, these “vocal dreams” are far more variable than the stereotyped patterns observed in species with largely innate vocalizations. This suggests that sleep may provide an exploratory space in which the nervous system generates and reorganizes candidate motor patterns. I will discuss how synthetic birdsong can be used to study this process and to test whether modifying sensory feedback during sleep can alter vocal learning. 

Claudio Mirasso

From Schouten’s Residue to Ghost Resonance and Temporal Interference: Same Trick, Different Names 

How can a system respond at a frequency that is not present in its input? This question connects phenomena historically described as the missing fundamental illusion, ghost resonance, and, more recently, temporal interference (TI) stimulation. In this talk, we will briefly revisit these links before focusing on the biophysical origin of frequency selectivity in TI. Using morphologically detailed models of hippocampal CA1 pyramidal neurons and interneurons, we show that the TI response depends critically on intrinsic membrane resonance, proximity to spike threshold, and cell-specific active conductances. Together, these properties determine both the preferred envelope frequency and the conditions under which temporal interference can most effectively modulate neuronal activity. 

Almuerzo 

Sara Solla

Neural manifolds for the control of movement

The analysis of neural dynamics in several brain cortices has consistently uncovered low-dimensional manifolds that capture a significant fraction of neural variability. These neural manifolds are spanned by specific patterns of correlated neural activity, the "neural modes." We discuss a model for neural control of movement in which the time-dependent activation of these neural modes is the generator of motor behavior. This manifold-based view of motor cortex may lead to a better understanding of how the brain controls movement. 

Pablo Mininni

TBA

TBA

Predrag Cvitanovic

A field theory of turbulence

Gutzwiller semi-classical quantization, Boven-Sinai-Ruelle dynamical zeta functions for chaotic dynamical systems, statistical mechanics partition functions, and path integrals of quantum field theory are often presented in ways that make them appear as disjoint, unrelated theories. However, recent advances in describing fluid turbulence by its dynamical, deterministic Navier-Stokes underpinning, without any statistical assumptions, have led to a common field-theoretic description of both (low-dimension) chaotic dynamical systems, and (infinite-dimension) spatiotemporally turbulent flows.

In this talk I will use a lattice discretized field theory in 1 and 1+1 dimensions to explain how temporal `chaos', `spatiotemporal chaos' and `quantum chaos' are profitably cast into the same field-theoretic framework.

For in-depth dive, see https://ChaosBook.org/overheads/spatiotemporal/ 

Ariel Chernomoretz

Network organization of protein binding-site landscapes across scales

Protein binding sites provide a local description of molecular recognition that can remain conserved even when the proteins hosting

them are globally dissimilar. Here, we combine protein language model embeddings with complex-network representations to organize

experimentally resolved binding sites into a quantitative similarity landscape. Binding-site embeddings derived from ProtT5 were used to

construct a mutual nearest-neighbor network in which nodes represent non-redundant protein–ligand binding sites and edges connect nearby

sites in embedding space. Comparison with structure-based alignments showed that proximity in this space is associated with binding-site

structural similarity.

The resulting network provides a multiscale description of protein–ligand interaction space. At the global and mesoscopic levels, its topology captures ligand-related organization and partitions the binding-site landscape into communities that are not fully accounted for by conventional sequence or domain classifications. At the microscopic scale, analysis of a kinase subnetwork reveals communities associated with binding sites occupying distinct regions of the enzymatic cavity. Conversely, by retaining similarity edges only between binding sites hosted by proteins with very low sequence similarity, we identify local binding-site relationships that persist across otherwise distant proteins. These results illustrate how network topology can connect local molecular recognition with protein-space organization across scales. 

Coffee Break + Poster Session