• Parallel Session 2 - Monday 23th, 14:45-16:25

    Biology / Urban Systems

Social systems

Auditorium

Chair: Laura Hernandez

Author: Jean-Claude Dreher

What are the mechanisms underlying how we integrate disparate and redundant information spreading in social networks? According to the DeGroot (DG) model, we adjust our beliefs based on the weighted average opinions of our social connections. In contrast, reinforcement learning (RL) models assume that learning occurs sequentially, driven by error-driven beliefs updating. Here, we directly compare these models based on two datasets of information propagating in social networks and determine which learning rule best accounts for the integration of information spreading in social networks. We found that variants of the DG model provided a better overall fit across participants and were more effective in building consensus and generating accurate estimates of the true state of the world. Finally, simulated models showed how social network topology affect choice accuracy and consensus depending on the learning rule. Together, our findings shed light on the computational mechanisms underlying information propagation in social networks.

Lucille Calmon, Elisabetta Colosi, Giulia Bassignana, Alain Barrat and Vittoria Colizza

High-resolution temporal data on contacts between hosts provide crucial information on the mixing patterns underlying infectious disease transmission. Publicly available data sets of contact data are however typically recorded over short time windows with respect to the duration of an epidemic. To inform models of disease transmission, data are thus often repeated several times, yielding synthetic data covering long enough timescales. Looping over short term data to approximate contact patterns on longer timescales can lead to unrealistic transmission chains because of the deterministic repetition of all contacts, without any renewal of the contact partners of each individual between successive periods. Real contacts indeed include a combination of regularly repeated contacts (e.g., due to friendship relations) and of more casual ones. We propose an algorithm to longitudinally extend contact data recorded in a school setting, taking into account this dual aspect of contacts and in particular the presence of repeated contacts due to friendships.

To illustrate the interest of such an algorithm, we simulate the spread of SARS-CoV-2 on our synthetic contacts using an agent-based model specific to the school setting. We compare the results with simulations performed on synthetic data extended with simpler algorithms to determine the impact of preserving friendships in the data extension method. Notably, the preservation of friendships does not strongly affect transmission routes between classes in the school but leads to different infection pathways between individual students. Our results moreover indicate that gathering contact data during two days in a population is sufficient to generate realistic synthetic contact sequences between individuals in that population on longer timescales. The proposed tool will allow modellers to leverage existing contact data, and contributes to the design of optimal future field data collection.

authors: Amina Azaiez and Antoine Mandel

We use transparency data published by the European Commission (EC) to perform a
quantitative analysis of the structure and dynamics of stakeholder consultation in EU policy- making process. We analyze the dataset through the prism of network theory by constructing a hypergraph whose nodes are EC officials and stakeholders, and hyperedges connect entities that participate in the same meetings. Our analysis highlights the presence of a hierarchical core-periphery structure, with a few well-connected entities that occupy the center of the network and enjoy a stable integration in the EC policy-making process. Examination of the core composition reveals that companies and trade associations maintain closer relationships with the EC. A regression analysis shows that lobbying efforts and company size are significant predictors of company centrality, independent of other objective characteristics. Our findings provide quantitative evidence supporting the perception of lobbying as a tool dominated by well-connected actors, while also revealing heterogeneous lobbying strategies across stakeholder groups.

Authors: Zakariae Benchrif, Floriana Gargiulo and Dominique Guillo

Understanding emotional responses in crisis events is crucial, as they shape public discourse, influence policy decisions, and affect community resilience [1]. Social media serve both as repositories of collective emotions and as vehicles for emotional contagion, providing therefore valuable insights into how societies process and react to disasters [2]. This study investigates the evolution of collective emotions following the 8th September
2023 Haouz earthquake, in Morocco, through sentiment analysis of user-generated content on two distinct online platforms: YouTube and Hespress, a prominent Moroccan news website. By analyzing comments before and after the disaster, we identify patterns of emotional expression and their temporal dynamics. Using a baseline established from comments posted in the month preceding the earth quake, our findings indicate that YouTube content generally exhibits a higher emotional intensity compared to Hespress. However, immediately following the earthquake, sentiment levels on both platforms converge, suggesting a shared emotional response across different online communities. Our analysis reveals the presence of distinct emotional waves. In the immediate aftermath (September 8–12), sentiment shifts towards a more positive tone, dominated by themes of solidarity, prayer, and communal support. This trend is further amplified, particularly on YouTube, on September 12, coinciding with the Moroccan king’s visit to the disaster
zone, which redirected discourse towards national unity and leadership support. However, from September 17 onward, sentiment turns negative on both platforms, correlating with growing criticisms of the government’s crisis management. A temporary recovery occurs on September 20 following the announcement of a new relief fund, but sentiment declines again from September 25, in parallel with a decreasing volume of comments and a discourse
increasingly focused on governmental accountability.
Our findings illustrate how natural disasters trigger significant shifts in communication patterns, leading to a temporary convergence of sentiment across platforms. Moreover, they highlight the presence of intrinsic emotional rhythms—alternating between solidarity and frustration—alongside exogenous fluctuations driven by external events, such as political actions and relief measures. To further investigate these emotional waves, and in particular to isolate the endogenous emotional responses, we are extending our analysis by comparing the Moroccan case to 62 other natural disasters that occurred between 2016 and 2019, using a large-scale database of Twitter content [3]. This comparative approach aims to uncover commonalities in emotional
responses to crises across different socio-political contexts, offering deeper insights into the mechanisms of collective sentiment formation in times of distress.

authors: Marc Delepouve, Ada Diaconescu and Emmanuel Ferrand

Nous appelons Système hyper-complexe un système complexe où des émergences inédites et imprévisibles prolifèrent et impactent l’état d’ensemble du système, conférant à ce dernier une évolution constante. Dans le contexte d’un Système hyper-complexe, nous définissons le Reste causal associé à une représentation de l’évolution future d’un phénomène donné, comme étant l’ensemble des phénomènes qui influenceront ou pourraient influencer cette évolution mais qui ne sont pas pris en compte dans cette représentation. Nous appliquons cet outil aux projections données par les modèles climatiques et aux scénarios publiés par le GIEC. Nous montrons que des phénomènes du Reste causal associé à ces projections et à ces scénarios pourraient avoir une action conséquente sur le climat. Pourtant, du 2e au 5e rapport d’évaluation du GIEC, les Résumés à l’intention des décideurs (RID) du GT1 ont présenté des scénarios tout en ne citant aucun des phénomènes du Reste causal associé. Nous concluons sur la nécessité de publier, avec les projections et les scénarios climatiques, le Reste causal associé, tant pour des raisons scientifiques que politiques. Puis nous élargissons cette conclusion au-delà du domaine du climat.

Urban Systems

Conference room

Chair: Julien Randon-Furling

Vehicular traffic in cities causes negative externalities in the real world, but it also represents a theoretical conundrum, insofar as it mingles multiple layers of complexity: the complexity of human behaviour, that of the street network, and the collective effects that emerge from the interactions between cars. Our contribution aims to underline that, despite this complexity, some salient features of traffic can be quantitatively rationalised, by introducing adequate stochastic processes to render uncertain real behaviours.

We will first consider the emergence of traffic oscillations (i.e., stop-and-go waves) when traffic gets dense, even without external perturbation. A widespread idea has it that this emergence is related to a deterministic linear instability, caused e.g. by response delays. In contrast, we will show that inaccuracy in drivers’ perceptions and responses, modelled by stochastic noise and termed `human error’ in (Laval et al., 2014), gives rise to realistic traffic oscillations; a physics-based reasoning will provide quantitative insight into their emergence.

In the second part, we will turn to parking search, which is a central issue for metropolitan transport authorities and for individual drivers: it plays a key role in mode choice and cars cruising for on-street parking may represent more than 10\% of the traffic
in many large cities (Hampshire et al., 2018).
Vast swaths of literature in Transport Engineering have thus been dedicated to parking and, more recently, a couple of works have shown that, in extremely stylised settings, parking search can be addressed analytically \cite{krapivsky2019simple}.
Our goal, here, is to evince that even in realistic settings — with a large-scale street network and heterogeneous drivers’ behaviours — the machinery of statistical physics can give quantitative insight into parking search (Dutta et al., 2023) and help explore smart-parking solutions. (See attached PDF)

Recently, we have started extending this approach to assess the potential of parking-related measures, such as a reduction of the parking supply (see inset) or smart parking applications, guiding users towards probably vacant parking areas.

Pedestrians demonstrate remarkable navigational capabilities in highly constrained environments, such as dense crowds or cluttered urban settings. When multiple individuals interact, complex collective behaviors emerge, including lane formation and stop-and-go waves. While many modeling approaches aim to replicate these familiar behaviors, few account for the pedestrians’ ability to mutually anticipate their non-linear moves over finite time horizon. Here we fill this gap by introducing a new crowd modeling branch inspired by game theory. Our work integrates methods and insights from diverse research fields: condensed-matter physics, through an analogy between space-time trajectories and polymers; biological studies, particularly regarding energy expenditure; and robotics.

Mass gatherings like Lyon’s Fête des Lumières, which draw millions annually, highlight the need for better crowd dynamics modeling to ensure safety and optimise pedestrian flow. Traditional models struggle in medium-density scenarios, oversimplifying pedestrian behavior and physical interactions. I propose an enhanced framework addressing these limitations by integrating anisotropic pedestrian shapes, realistic mechanical interactions, and advanced decision-making processes.
Conventional models often use isotropic circular shapes, failing to account for sneaking behaviour. They also inadequately replicate high-density scenarios observed at events like Fête des Lumières. Existing approaches typically separate mechanical and decisional behaviors—mechanical models handle collisions but lack sophisticated avoidance strategies, while decision-making models underestimate collision frequencies.
The proposed framework combines a decision-making layer (governing translational velocity and rotational speed based on constraints) with a mechanical layer (modeling physical interactions inspired by granular dynamics). It uses anisotropic shapes based on anthropometric data to reflect individual heterogeneity. Constraints include destination goals, biomechanical limits, personal space preservation, and time-to-collision strategies.
Validated through simulations of real-world scenarios, the model improves accuracy across density ranges. Its advancements have critical implications for enhancing safety and optimising pedestrian flow during large-scale events.

The ever-increasing urbanization of the world meets us with pressing socio-environmental challenges. The sprawl of human settlements all over the planet leads to losses of arable land and biodiversity, and increases flood risks. Furthermore, this expansion is concerning with regard to climate change. In this context and considering the developing will of limiting urban sprawl (see for example the No Net Land Take objective [5]), we are faced with the task of understanding the fundamental structure and dynamics of cities.
Cities are more than just points on a map. They have an internal structure which unfolds radially, from center to periphery, revealing patterns that shape urban dynamics. To understand this spatial organization, we analyze how the share of built-up land evolves as we move outward. Since cities present a wide variety of sizes, scaling laws provide a powerful framework for modeling such behavior, capturing how a system’s properties shift with its size. Viewing cities as systems and population as their defining scale, we study how cities sprawl as population grows, at the global scale.

In order to do so, we establish a robust radial scaling law which quantifies the connection between the distance r to the city center and the amount of built-up land share sN (r), and how this relation scales with city size N.
We extend the homothetic scaling obtained in previous work [6,7] to a global sample of cities and at different dates to study the evolution over time. We focus our work on the 1860 cities of the world whose population is greater than 300,000 inhabitants in 2020. This sample presents a large diversity in terms of population size, topology, land use, urbanization policies and more. Despite such a wide variety, the scaling law still applies with surprising regularity. Furthermore, looking at the data at different points in time — from 1975 to 2020, with a 5 year step — allows us to analyze the evolution of this internal urban structure and scaling law of built-up land.
The data used in the study come from the Global Human Settlement Layer (GHSL), produced by the European Commission. It provides high-resolution and high-quality, globally consistent distributions of built-up areas, which we combine with the World Urbanization Prospect database from the United Nations for trustworthy population statistics. For each city of choice and each date, we analyze this GHS BUILT-S raster layer at 100 meters resolution and compute the average built-up land share in concentric rings of 200 meters width around the city center. To ensure the viable comparability between cities, we use for each city, with population N , a rescaled distance to the center r′ = r (N_Tokyo/N)^(1/2) which makes all cities comparable to the largest one, Tokyo (with population N_Tokyo ≃ 3.7 × 10^7 in 2020).

We analyze the evolution of the mean rescaled profile on Figure 1, and observe that built-up land increases over time all along the center-periphery profile, even when the size effect is controlled by the homothetic scaling law. In linear scale, the change is especially visible near the center, while it appears more clearly in the periphery on a semilog graph (Figure 1). This result means that the built-up surface per capita increases over time globally. We link this urban sprawl phenomenon with economic development and further analyze its geographical variations at national scale on the planet. This clearly questions the sustainability of urban expansion.