Johan Bollen
University of Amsterdam, Netherlands
Computational science approaches to study the complex systems that shape mental health.
Millions of individuals have now published extensive longitudinal records of their emotions, thoughts, and activities on social media. We leverage machine learning and artificial intelligence methods to study the complex dynamics of the bidirectional relationship between cognitive biases as they are expressed in large-scale records of online language and internalizing disorders such as depression and anxiety. Our work suggests an interesting relation between the online expressions of cognitive biases and societal polarization, even at the timescale of years and decades. Unraveling the connections between online language, cognitive biases, and social phenomena such as polarization may point towards a better quantitative understanding of the dynamics of the cognitive, behavioral, and social drivers of mental health disorders at population scale.
Short bio
Tiziana Di Matteo
King’s College London, UK
Network based filtering tools: a machine learning framework
Data are everywhere and carry valuable information, making their understanding, analysis, andfiltering central to modern science, industry, and society. Developing tools that can analyzesuch information while it is generated, while reducing complexity and dimensionality withoutcompromising data integrity, has become crucial. Network theory has emerged as a powerfulframework for this purpose. In this talk, I will explore two complementary perspectives onnetwork-based filtering and dimensionality reduction: one focusing on edges and the other onnodes within a network representation. I will first introduce correlation-based informationfiltering network tools [1], which have proven effective for analyzing complex datasets. Thesetools are particularly valuable for risk management and portfolio optimization, enabling theconstruction of probabilistic sparse models for financial systems that support forecasting, stresstesting, and risk allocation [2–4]. Next, I will present a newly developed method, the Best-PathAlgorithm Sparse Graphical Model (BPASGM) [5]. BPASGM extends the original Best-PathAlgorithm by transforming dependency discovery into a structured, economically motivatedasset-selection procedure. This machine-learning framework for portfolio constructioncombines sparse graphical modeling with portfolio theory, offering a statistically grounded andcomputationally efficient approach for dependence-aware asset selection. BPASGM isdesigned to improve realized portfolio performance in finite samples, addressing the highsensitivity to estimation error seen in classical sample-based Markowitz implementations.Monte Carlo simulations indicate that BPASGM-based portfolios exhibit more stable risk-return characteristics, lower realized volatility, and enhanced risk-adjusted performancecompared to standard mean–variance portfolios. Applications to real financial datasetsdemonstrate the method’s practical utility.
1. M. Raddant, T. Di Matteo, Journal of Economic Interaction and Coordination, 2023https://doi.org/10.1007/s11403-023-00389-6.
2. F. Pozzi, T. Di Matteo and T. Aste, Scientific Reports
3 (2013) 1665.
3. N. Musmeci, T. Aste and T. Di Matteo, Scientific Reports 6, 36320; doi:1038/srep36320(2016).
4. W. Barfuss, G. Previde Massara, T. Di Matteo, T. Aste, Phys.Rev. E 94 (2016) 06230.
5. T. Di Matteo L. Riso, M. Zoia A Novel approach to portfolio construction, arXivpreprint arXiv:2602.03325, submitted 2026.
Short bio
Chiara Poletto
University of Padova, Italy
Unravelling the complexity of multi-pathogen interaction
The simultaneous circulation of multiple pathogens fundamentally alters epidemic dynamics and the effectiveness of public health interventions. While advances in genomic sequencing now enable detailed detection of pathogen diversity, interpreting this diversity remains challenging, as it is shaped by complex human behaviour and interaction patterns. These complexities have profound implications for epidemic control.
In this talk, I will explore how human and environmental factors shape the dynamics of co-circulating viruses. By combining data analysis, stochastic modelling, and network theory, I will demonstrate how complex networks operating at different scales, together with dynamically changing transmission conditions, influence viral co-existence and dominance patterns. I will illustrate these concepts using case studies on acute respiratory infections and show how theoretical insights can provide ground knowledge to inform public health response.
Short bio