Network inference for measuring opinions on social media with partial information
by Alessandro Galeazzi and Max Falkenberg

Online social networks increasingly mediate how billions of people worldwide consume and share content. This content can often be political in nature, playing a key role in driving political polarization, the creation and spread of misinformation, and the formation of online echo chambers. However, the opinions of online social media users are often difficult to measure directly, therefore their opinions must often be inferred indirectly. Arguably, the greatest impediment to accurately measuring online opinions is the lack of user level data, or data which has been pre-filtered and does not fairly reflect an individual’s social media use. With new restrictions on academic data access having come into place in 2023, a key open question is “what is the minimum amount of data needed about an individual to infer their opinions, and to what extent can social networks be used to reconstruct this information?” To address this question, our project will infer users’ opinions on social media using partial data and network reconstruction methods. Using existing social media data, our study will start from opinion inference techniques which use complete interaction data and aims to extend these methods for use with partial data and indirect interactions. Our project will be primarily computational, but also offers opportunities for users to develop and implement opinion dynamics models if the students are interested. With this approach, we hope to develop measures which are able to predict a user’s ideological leaning in different contexts using minimal datasets.


The tutors

Max and Alessandro are computational social scientists working on social networks and online polarization.

Alessandro is currently Assistant Professor at the University of Padova in Italy. He completed his Ph.D. at the University of Brescia, followed by a Postdoc at Ca’ Foscari University of Venice, on topics related to complex systems, social networks, the formation of echo chambers on social media.

Max is currently a Postdoc at City, University of London, UK. He completed his PhD in the physics of complex systems working on theoretical models of social network growth, and on data analysis for complex systems. Max and Alessandro have been collaborating on a number of projects in the last three years and look forward to welcoming new people to the team.

Studying gender inequalities in networks of global (and local) film industry
by Alessio Cardillo

Studies of gender inequality in film industry have highlighted the persistence of gender imbalances in creative roles (e.g., director or writer). Network science offers an alternative way of identifying and proposing change mechanisms, as it puts emphasis on relationships instead of individuals. In this project I propose to study the network of collaborations in global (and eventually, local) film industry by leveraging the information available on the Internet Movie Data Base (IMDB). We will characterize the structure of the collaboration network of films’ producers, directors, and writers and, in particular, of the sub-networks of male-male, female-female, and male-female (mixed) relationships, as well as how these properties evolve over time. If possible, we will narrow down our attention to either local film industries (e.g., Latin-American, Indian, etc.), or to specific periods (e.g., the first half of twentieth century).


The tutor

Alessio is a senior postdoctoral fellow at the Internet Interdisciplinary Institute at the Universitat Oberta de Catalunya of Barcelona (Spain). He got his Ph.D in Physics in 2014 at the University of Zaragoza (Spain). He has worked as postdoctoral fellow at the Swiss Federal Polytechnic University (EPFL) in Lausanne, the University of Bristol (United Kingdom), and the University Rovira i Virgili (Spain). Alessio’s research concerns the analysis of complex systems mappable as networks via a blend of methodologies based on nonlinear dynamics, statistical physics, and data science. Across his career, he has worked with urban, transportation, social, technological, lexical, and literary systems.

Impacts of COVID-19 restrictions on mobility networks and the spread of endemic respiratory viruses
by Amanda Perofsky

In early 2020, there was widespread adoption of public health measures to slow the spread of SARS-CoV-2, including stay-at-home orders, gathering restrictions, school and business closures, and travel bans. These interventions were effective at reducing not only SARS-CoV-2 transmission but also the spread of other respiratory pathogens, with many endemic respiratory viruses not returning to widespread circulation until the end of 2020 or 2021. Aggregated location data from cell phones have been used extensively to model SARS-CoV-2 dynamics, but few studies have explored relationships between human mobility and the transmission of endemic respiratory pathogens during the COVID-19 pandemic. In this project, we will use uniquely detailed data from a citywide respiratory pathogen surveillance study, in combination with mobile device location data, to investigate the impacts of mobility behavior on endemic pathogen rebound during the pandemic in Seattle, Washington. First, we will perform an in-depth analysis of temporal changes in Seattle’s mobility network during the pandemic, using weekly data on between-neighborhood movement. Next, we will use mechanistic mathematical models of transmission to explore how changes in case importations and within-city connectivity affected the timing and size of rebound for endemic viruses with different biological characteristics (e.g., rhinovirus, respiratory syncytial virus, seasonal coronaviruses). Contact rates and seeding will be informed by mobility data, and epidemic trajectories produced by models will be compared to observed pathogen incidences from Seattle Flu Alliance surveillance data. These two exercises will introduce participants to network analysis and infectious disease compartmental models.


The tutor

Amanda is a Research Scientist in the Brotman Baty Institute for Precision Medicine at the University of Washington and a guest researcher at the Fogarty International Center, US National Institutes of Health. Prior to joining UW, she completed her PhD in Ecology, Evolution, and Behavior at The University of Texas at Austin and a postdoctoral fellowship at the Fogarty International Center, NIH. Amanda’s research focuses on the epidemiology and transmission dynamics of respiratory viruses, and she is particularly interested in the roles of population behavior and viral evolution in driving epidemics.

Urban mobility modeling: A data-driven approach for trajectory reconstruction and prediction
by Davide Colombi

This project delves into the complex domain of the human intra-urban mobility modeling, in particular in the trajectories reconstruction from GPS data with potentially large and varied sampling rate along with spatial and temporal gaps. The main goal of the project is to create a mobility model capable of reconstructing and predicting the intra-urban mobility patterns of a designated device, especially in scenarios with compromised data quality. To address this challenge, two anonymised mobility data samples will be released for a specific number of users, a specific time period and a well-defined urban area. The first sample, serving as a high-quality ground truth, will distinctly delineate the trajectories of individual users. The second sample, on the other hand, will be a modified version of it, designed by introducing noise and resampling to simulate real-world spatial and sampling irregularities. These datasets will be complemented by the corresponding road network extracted from OpenStreetMap. The team’s mission is to develop a methodology capable of reconstructing accurate trajectories from the corrupted subsample. At the same time, the team must provide metrics that facilitate the evaluation of the precision and accuracy of the results against the established ground truth. The entire process will take place within Cuebiq’s Spectus platform, which provides access to the data and computational capabilities required for effective development and analysis. In conclusion, the project enables the exploration of real-world mobility data and the interesting challenge of mobility trajectory reconstruction, all by exploiting the advanced capabilities of the Spectus platform.


The tutor

Davide is a Senior Data Scientist at Cuebiq, expert in computational epidemiology, network science, human mobility and mathematical modeling. Previously a consultant for a global automotive industry, expert in geospatial analysis and predictive maintenance for connected commercial vehicles. Academic basis in condensed matter physics and complex systems, with emphasis on agent-based models simulating social dynamics and consensus phenomena in opinion dynamics. Ph.D. and postdoctoral research focused on the spread of infectious diseases through theoretical and real-world networks.

What is the optimal railway network?
by Ebrahim Patel

This project aims to construct optimal railway networks by employing the novel modelling tool of tropical mathematics. Straightforward algorithms will be formulated and applied to abstractions of railway networks from cities around the world using freely available data of subway/railway network connections and travel times. For each network, we will formulate an algorithm that requires only a standard understanding of tropical mathematics, following which the optimisation steps are grounded solely in the study of circuits in the network. A fundamental theorem says that, in a strongly connected network such as the railway, the eigenvalue in this system is unique and determined by the circuit(s) with the largest average weight. Thus, critical circuits in the network are those circuits that comprise the longest average round-trip time, which also represents average inter-departure times. Reducing the eigenvalue therefore addresses the aim of maximising the frequency of departures; our algorithm achieves this by identifying such circuits and incrementally adding dummy stations on them. Subsequently, new critical circuits are identified, and the process is repeated until desired. Thus, we strategically add nodes and edges, enlarging circuits along the way so as to reduce the eigenvalue.  Ultimately, this produces a new network layout as optimal. We will subsequently analyse the practicality and commonalities of these optimal railway networks.  If time permits, we will look at an extension of this model, which will allow us to explore fresh insights into network science areas such as threshold dynamics, network backbones, and the structure-dynamics interplay.


The tutor

Ebrahim is a Lecturer in Mathematics and Data Science at the University of Greenwich. He first met tropical mathematics during his PhD at the University of Manchester. Most of his research has subsequently been conducted at the University of Oxford, focusing on discrete dynamical systems and network science, collaborating with engineers, computer scientists, and artists, and applying the work to industry. He is also a co-founder of The Bees, an award-winning mathematical writing group, aiming to promote the beauty of mathematics and its applications to non-experts.  Previously, he was a founding faculty member of The London Interdisciplinary School.

The co-evolution of food insecurity and migration networks
by Elisa Omodei

Food insecurity, defined as the lack of physical or economic access to safe, nutritious and sufficient food, remains one of the main challenges of the 2030 Agenda for Sustainable Development. Food insecurity is a complex phenomenon, resulting from the interplay of environmental, socio-demographic, and political events. Insightful qualitative work on the nexus between climate change, conflict, migration and food security was carried out during the last decade (Crush 2013; Martin-Shields and Stojetz 2019; Morales-Muñoz et al. 2020). However, quantitative analyses of the linkages among all these factors, and more specifically between migration networks and food security are still lacking. How do migration networks impact food access in receiving regions and, in return, how does food access impact the structure of migration networks (e.g. by changing flow sizes or inducing new migration routes)? How do migration networks co-evolve with food insecurity? In this project, we aim to address these questions using publicly available data provided by the UN Department of Economic and Social Affairs and the Food and Agriculture Organization.


The tutor

Elisa is an Assistant Professor at the Department of Network and Data Science at the Central European University in Vienna (Austria). She holds a BSc and a MSc in Physics from the University of Padua and Bologna, respectively, and a PhD in Applied Mathematics for the Social Sciences from the École Normale Supérieure (ENS) of Paris. During her career, she spent over four years at the United Nations, first at UNICEF’s Office of Innovation and then at the UN World Food Programme. In her research, she explores how complexity and data science can help us address the needs of the most vulnerable populations and monitor the UN Sustainable Development Goals. She also served as Vice-President Secretary of the Complex Systems Society.

Drivers of group formation and evolution
by lacopo lacopini

This project focuses on deciphering the drivers of social interactions through a comprehensive analysis of longitudinal data supplemented by metadata. The project builds on top of recent research outputs on the social dynamics of link and group formation extracted from temporally-resolved empirical proximity data. In particular, by linking both individual and external factors (as well as environmental constraints) to the dynamical behavior of a temporal social network, the study aims to study the patterns of initiation, maintenance, evolution, and transformation of both interpersonal bonds and group affiliations. The investigation will extend beyond pairwise interactions to also include group dynamics. The project will rely on two independent data collection experiments that include both human interactions and individual metadata (age, language, declared attractiveness, socio-economic features, family education, etc.), bringing valuable insights for social and network scientists.


The tutor

Iacopo is an Assistant Professor in the Network Science Institute at Northeastern University London. His research interests are in the area of complex networks and computational social science, with a focus on adoption processes, seeding/targeting strategies, and group and team dynamics. In parallel to his core research on human behaviour, he is also working on urban networks and animal social networks.
A physicist by training, Iacopo holds a PhD in Mathematics from QMUL. Prior to joining NUL, he was awarded the JSMF Postdoctoral Fellowship hosted at the Department of Network and Data Science at CEU. He has held research positions at the Centre de Physique Théorique of Aix-Marseille Univ, at the Centre for Advanced Spatial Analysis at UCL, and at The Alan Turing Institute. Previously, he worked at the ISI Foundation and the United Nations-UPU as a Data Scientist.

Putting the “network science” into NetSci conference organisation
by Leto Peel

Academic conferences play a key role in the communication of new results in science. NetSci is the flagship conference of the Network Science Society, which aims to bring together leading researchers and practitioners working in the emerging area of network science. Since its inception in 2006, participation at NetSci has steadily been increasing. With increasing numbers of attendees and contributed talks, each year places a greater strain on the organisation of the technical program. A poorly structured program diminishes the participant experience, where related talks are scheduled in parallel with each other and/or individual sessions are not topically aligned. In this project we will address the challenge of scheduling of parallel sessions: Can we find a good way to assign talks to specific sessions and decide which sessions to run in parallel with each other that causes minimal conflicts and improve the conference participant experience? At NetSci 2017 the IUNI Conference Agenda Manager (CAM) was introduced as an application to allow attendees (“users”) to create custom schedules of the talks they were interested in. The CAM has also provided an opportunity to collect data on user preferences of talks (“events”). In this project we will use network science methods to analyse user preferences. Specifically, the data can be represented as a bipartite network of users and events, with rich metadata describing the events including the talk titles, abstracts authors and the session they are assigned to. We have data from multiple editions of the conference and may try to compare against additional data, such as the collaboration network of presenters. The successful project will have the opportunity to join the program committee and put the method into practice for the NetSci 2025 conference, that will be held in Maastricht, The Netherlands.


The tutor

Leto is an Assistant Professor at the Department of Data Analytics and Digitalisation(School of Business and Economics) of Maastricht University. His main research interests are in machine learning for complex networks, in particular the use of probabilistic generative models and Bayesian inference. He completed his doctorate at the Computer Science Department of University College London while working as a research scientist at BAE Systems Advanced Technology Centre and then later as a research associate at UCL CEGE Department working on the Crime Policing and Citizenship Project. He has held postdoctoral positions at University of Colorado at Boulder and Université catholique de Louvain and has been a visiting researcher at the Université de Namur.

The role of mobile phone ownership biases on human mobility segregation
by Mattia Mazzoli

Mobile phone data are crucial to inform transportation planning and public health research. However, population level aggregated origin-destination (OD) matrices often hide inner differences reflecting socio-economic disparities and shaping health outcomes of epidemics in demographic groups. By crossing publicly available mobility datasets provided by the Spanish Ministry of Transportation with census records, here we approach tech adoption bias by studying the correlation among age, gender and income biases of mobile phone ownership in Spanish municipalities. By adjusting OD matrices by age and gender, we analyse socio-economic segregation levels in age groups and gendered respective networks and their differences with respect to the original dataset. The outcome of this study will shed new light on how the intersection of age, gender and income biases affects mobility flows estimates and provide new insights on the drivers of mobility segregation in the post-pandemic era.


The tutor

Mattia is currently a PostDoctoral Researcher at ISI Foundation in Turin, Italy where he works with digital surveillance systems for respiratory diseases and human mobility data integrated epidemic models.
He holds a PhD in Physics of Complex Systems achieved at IFISC in Palma, Spain, with a thesis on human mobility modeling and applications on epidemic spread, urban structure and migration patterns. Relying on his expertise, he mainly worked with spatial transmission models to describe the invasion of infectious diseases at different scales. His main areas of interests are human mobility, pandemic preparedness, health care accessibility and inequalities.

Unraveling cancer dynamics: From multiscale stochastic models to tissue morphology
by Pilar Guerrero

This project delves into the intricate dynamics of cancer, particularly focusing on glioblastoma, employing refined hybrid stochastic models with multiscale population dynamics. The research explores structured master equations at multiple scales, utilizing advanced techniques to analyze cellular population dynamics. Additionally, hybrid multiscale methods are proposed to efficiently model heterogeneous processes, particularly in tumor growth scenarios. The project also introduces a dynamical vertex model for cancer detection, incorporating a 2D cancer growth model and mechanical forces between cells. The model proves effective in capturing tissue abnormalities, with future plans to include diffusion for a comprehensive understanding of cancer dynamics. The validated models offer valuable insights into the mechanical forces governing tumor progression, advancing our comprehension of complex cancer processes.


The tutor

Pilar is an accomplished mathematician with a strong background in applied mathematics, focusing on partial differential equations in developmental biology and stochastic multi-scale modelling of cancer growth. Her research experience includes modelling cancer behaviour, stochastic multi-scale modelling of biological systems, and mathematical modelling of the cell cycle.
She has also worked in tissue 2D computational mathematical modelling (vertex model) to understand mechanical forces and tissue behaviour in gastric cancer and vertebrate embryos. Currently, she is an Associate Professor in the Mathematics Department at Universidad Carlos III de Madrid and a member of the Interdisciplinary Group of Complex Systems (GISC).