Abstracts Seminar 04-02-21

Vincent Buskens (Utrecht University)

Interrelated dynamics of social networks and infectious disease spread

Recent research shows an increasing interest in the interplay of social networks and infectious diseases. Many studies either neglect explicit changes in health behavior or consider networks to be static, despite empirical evidence that people seek to distance themselves from diseases in social networks. We propose an adaptable steppingstone model that integrates theories about social network formation, risk perception, and infectious diseases. We argue that networking behavior in the context of infectious diseases can be described as a trade-off between the benefits, efforts, and potential harm a connection creates. Agent-based simulations using a generic model implementation show that (i) high (perceived) health risks create strong social distancing, thus resulting in low epidemic sizes, (ii) large numbers of ties provide more opportunities for disease spread, thus creating larger final sizes and shorter epidemics, and small changes in health behavior can be decisive for whether the outbreak of a disease turns into an epidemic or not, (iii) high benefits for social connections create more ties per agent, providing large amounts of potential transmission routes and opportunities for the disease to travel faster, and (iv) higher costs of maintaining ties with infected others reduce final size only when benefits of indirect ties are relatively low. These findings suggest a complex interplay between social network, health behavior, and infectious disease dynamics. Furthermore, they contribute to solving the issue that neglect of explicit health behavior in models of disease spread may create mismatches between observed transmissibility and epidemic sizes of model predictions.

(this is joint work with Hendrik Nunner and Mirjam Kretzschmar)

Frank P. Pijpers (CBS and University of Amsterdam)

Network reconstruction and privacy preservation for large networks

CBS is the National Statistical Institute of the Netherlands. The primary function of National Statistical Institutes is to compile descriptive statistics (tabular material and time series) on a very broad range of social and socio-economic phenomena. The rationale behind this is that for good governance, planning, and also for taxation, it is important to have accurate and up-to-date information on a wide variety of aspects of society. In the first half of the last century, it could be said that, on the whole, the life course of people followed a fairly linear course, in a number of ways. For instance, one and the same economic activity remained their main source of income. Similarly, economic activity of companies tended to be confined to one sector. In such a linear system, descriptive statistics can be argued to be sufficient in order to enable the study of its behaviour.

In more recent decades, increased automation and digitalization of manufacturing processes has had a number of consequences. Formerly separate sectors of economic activity are increasingly intermeshed, and, in the Netherlands, all have gained strong services (financial, logistics, etc.) components. In the wake of this, the life course of people has also become much more branched out and diverse than before, both in the personal and professional spheres. It is therefore much more appropriate now, to view economic activity (the population of businesses) as well as social activities (between people) as a complex network of relationships. This has far-reaching consequences for governance and planning, which implies that CBS needs to use insights from dynamical processes on (and of) networks to redesign what it monitors. The publications of CBS need to be a set of representative indicators of the system as a whole, rather than merely of its constituent parts.

While CBS already has access to a very large amount of data on citizens and businesses in the Netherlands, such data can never completely cover all aspects of their interactions. This matters, because a complex system is less defined by the characteristics of the constituent parts (people, business) and much more by the interactions between these parts and how these interactions vary with their characteristics. A good understanding of how the system functions, e.g. whether it generates its own crises or perhaps conversely tends to some form of equilibrium, requires at least a representative if not complete view of all (types of) interactions. However, there is a tension here as well with privacy concerns. While the specific law governing the mandate of CBS certainly allows mapping out such networks for statistical and scientific purposes, in its publications CBS must safeguard against the disclosure of sensitive personal or business information.

In this talk I will describe the efforts to perform Bayesian reconstructions of incompletely known networks, as well as aspects of techniques that CBS would need to employ in order to ensure that any publication of statistical aspects of the network properties does not inadvertently reveal too much information which can be classified as sensitive