Loading Events

« All Events

  • This event has passed.

The Small-World structure of a population-scale social network

13 September 2022 , 15:16 15:36 UTC+1

Lecture by Frank Takes, Eszter Bokànyi, and Eelke Heemskerk at the 6th European Conference on Social Networks (EUSN 2022).

The analysis of social networks at the level of an entire population provides a unique opportunity to revisit what is perhaps the most fundamental universal finding in the field: the small world phenomenon. Popularized as “six degrees of separation”, it refers to the remarkably low average node-to-node distances and typically high amount of clustering observed in real-world social networks.

In this talk we revisit this concept from the perspective of the multilayer population-scale social network of the Netherlands consisting of over 17 million people and five characteristic layers of connectivity based on family, household, school work and neighborhood relations. By means of a comparison with artificially simulated networks originating from the well-known Watts-Strogatz and Newman-Watts-Strogatz models, we show that each type of connectivity (i.e. each layer) in the considered population-scale social network has a different characteristic function in realizing the small-world structure of the network. In particular, we show how highly clustered family relations form the backbone of the network, akin to the initial regular (ring-shaped) graph in the WS model. Then, work and school relations primarily act as “random” bridges between different parts of the clustered (yet, on its own not-so-small-world) family network. Such edges are ultimately crucial for realizing the actual small-world connectivity patterns in the complete population-scale social network. 

Apart from the explicit multilayer aspect of our population-scale social network data, we can also take advantage of node attributes (i.e., people’s demographic characteristics) in understanding how the low average distances of the network are realized. We find distinct patterns of assortativity between node pairs that are unconnected in one layer (e.g., family) and connected in the other (e.g., work). Subsequently, these patterns are used to improve the rewiring step of the Watts-Strogatz model in an attempt to more accurately represent our population-social network data. Specifically, this step takes into account empirical patterns of homophily, utilizing the extent to which connections are more frequently formed between nodes with close spatial and similar demographic attributes, such as level of education.

On the theoretical side, the talk provides insights into the relevance of the WS model, more than 20 years after its inception. Moreover, it highlights the importance of multilayer approaches in explaining the connectivity of multilayer networked systems in the real world, as well as the relation between similarity of node attributes and links spanning larger distances. Substantively and empirically, the talk contributes to an understanding of how distances are bridged in a social network. The work has important implications in processes that may take place on networks, such as diffusion of information, dissemination of resources, or epidemic spreading.