Groups in social networks

Montag, 02. Dezember 2019
11:45 – 12:45 Uhr


Centre for the Advanced Study of Collective Behaviour

Ulrik Brandes, ETH Zurich

Diese Veranstaltung ist Teil der Veranstaltungsreihe „Seminar Series of CASCB“.

Ulrik Brandes is a Professor for Social Networks in the Department of Humanities, Social and Political Sciences at ETH Zurich. He has a background in algorithmics, with special interests in graph algorithms and network visualization. His current focus is social network science, where methodology and method development are complemented by a growing interest in societal challenges arising from the Digital Transformation.

Groups in social networks

We discuss two different concepts of groups in social networks, andnovel methods to determine them. The rationale of these methods is bestunderstood when viewed in the context of positional network analysis,an approach we are devloping to solidify network science more broadly.The first grouping concept is known as community detection, or graphclustering, where a community consists of individuals that are morecohesive among each other than toward the rest of the network. Methodsfor community detection are generally based on a global objectivefunction with no interpretable mechanism underlying it. We outline twoalternative approaches based on structural embeddedness and socialinfluence.The second grouping concept is known as role assignment, where a groupof individuals are considered to enact the same role if they have acommon pattern of relationships with others. In the literature, this isusually formalized via node equivalences, which, however, turn out tobe too restricted and overly sensitive to changes in the data to beempirically useful. We introduce an alternative approach that is moregenerally applicable and at the same time provides a micro-foundationfor the evolution of roles.

Image by ETH Zurich
Image by ETH Zurich