Invited Speakers
Title: Insights and lessons from data-driven modeling of COVID-19
Abstract: The COVID-19 pandemic forced an unprecedented response from health authorities worldwide. It also elicited a huge effort from the scientific community that rushed to improve and make more accurate already existent tools in epidemiological modeling. Research has produced many key results that combine theoretical models with data-driven simulations. Here we show results that correspond to different stages of the COVID-19 pandemic in order to illustrate the trade-off between what models can tell and where more detailed data-driven modeling is useful, even if data is incomplete. Our models are tailored to mobility and population data from Europe and the U.S. and allow estimating the effectiveness of customary public interventions on the spread of COVID-19. We conclude by identifying the most pressing needs for better preparedness for future similar pandemic scenarios. |
Title: Trustworthy Network Science
Abstract: As the use of machine learning (ML) algorithms in network science increases, so do the problems related to explainability, transparency, fairness, privacy, and robustness, to name a few. In this talk, I will give a brief overview of the field and present recent work from my lab on the (in)stability and explainability of node embeddings. Time permitting, I will also present work on adversarial ML algorithms on complex networks and on equality of information access in complex networks. |
Title: Modeling COVID-19 Transmission: From Toy Models to Real-World Applications
Abstract: The field of epidemic spreading has seen significant progress in recent years, with various frameworks available for modeling the diffusion of epidemics on networks. The Microscopic Markov Chain Approach, a probabilistic formulation developed over a decade ago, has been used to model compartmental epidemic spreading models on networks with good accuracy. Efforts have been made to improve this framework to better represent real-life scenarios, including adapting it to networks with metapopulation structure. The outbreak of COVID-19 in 2020 prompted questions regarding the suitability of these models for predicting the spread of the virus in real-world scenarios. In this talk, we will review the Microscopic Markov Chain Approach from its inception and present a model that has successfully described, predicted, and controlled the spread of COVID-19 in Spain. |
Title: The role of higher-order interactions in the dynamics of social contagion and norm change
Abstract: Complex networks have become the main paradigm for modelling the dynamics of interacting systems. However, networks are intrinsically limited to describing pairwise interactions, whereas real-world systems are often characterised by interactions involving groups of three or more units. In this talk, I will discuss the effects of considering group interactions on the dynamics of social systems, a natural testing ground for higher-order approaches. In particular, I will generalise two models of social contagion and norm evolution, traditionally studied on graphs, now defined as dynamical processes on hypergraphs. Leveraging real-world interaction data and analytical insights, I will show the emergence of novel phenomena such as discontinuous transitions and critical mass effects induced by the higher-order interactions. Finally, I will propose a measure of hyper-coreness to characterise the centrality of nodes in hypergraphs, and show that tailored seeding strategies can play a crucial role in dynamical processes driven by group interactions. |
Title: Complex networks in high school: structure and dynamics
Abstract: In this talk I will present the results of a study on networks of relationships we have been conducting with high schools since 2018. Our rich dataset allows us to answer both structural and dynamical (longitudinal) questions. From a more structural viewpoint, I will discuss how the features of the network reflect and influence the characteristics of the school, and I will show how the existence of a relationship can be predicted with great accuracy both from a microscopic and a microscopic viewpoint. On the side of the dynamics, I will provide evidence of the stability of the results along several years and I will show how the Dunbar’s circle structure can be seen in the dynamics of the relationships. Leveraging on the information on negative relationships, I will discuss the theory of social balance in light of our observations and our structural knowledge. Our results provide important insights in our quest for a theory that shows how the individual “social atoms” merge onto the network of society. |
Title: Structural marginalization in society and algorithms
Abstract: Structural marginality refers to structural conditions that push certain groups towards a network’s margins, limiting their access to resources. Despite its importance, there is a minimal quantitative understanding of its manifestation in networks, and therefore we are unequipped to answer several urgent societal questions. For example, what underlying structural mechanisms drive marginalization? How can marginalized groups improve their social positions? Do hard (mathematical) limits exist on actions that would alleviate structural inequalities? In this talk, I argue that tools in complexity science are instrumental in addressing and mitigating structural marginality in society and algorithms. |
Title: Uncovering the multiscale dynamics of temporal networks
Abstract: Many systems exhibit complex temporal dynamics due to the presence of different processes taking place simultaneously. Dynamic community detection consists in extracting a simplified view of the time-dependent network of interactions describing those systems. Attempts to generalize static-network methodologies based on an underlying dynamic to temporal networks usually rely on aggregation over time windows and face the fundamental difficulty that a stationary state of the dynamics does not always exist. Usually, these methods consider a process decoupled from the intrinsic time of the system under study to guarantee its stationarity. In this talk, I will present an approach for community detection based on a dynamical process evolving on the temporal network that allows dynamics that do not reach a steady state. By varying the rate of the dynamical process this method provides a natural way to disentangle the different dynamical scales present in a system. Examples of applications include the characterization of the multiscale dynamics of a wild mice population and the identification of network scientists’ communities of influence in a collaboration network. This framework opens the door for the definition of new concepts for temporal networks in terms of dynamical processes and flows that may help to disentangle the complex processes simultaneously occurring in systems described as temporal networks. |