Processes of dynamic opinion change govern how information spreads and consensus forms in democratic societies. The emergence of social media has changed these processes in two important ways. First, it has made them more transparent, as data from social media are often available and amenable to analysis. Second, however, it has also exposed social interactions and processes of opinion formation to opaque ways of manipulation, often involving multiple competing parties that exert, or allocate, influence strategically to further their respective goals. This is particularly evident in the current debates about foreign influence on election results, the spread of fake news or the emergence of paid influencers to advertise products on social media platforms.
A good understanding of this strategic manipulation by multiple competing parties is needed to: (i) gain better awareness of strategic influence on social media, (ii) learn about ways to counter or prevent it, and (iii) use it effectively for public policy. To date, a large body of literature has investigated processes of opinion spread and strategic influence allocation on social networks. Broadly, one can distinguish two streams of research: (i) models similar to the independent cascade model [1], a model in which opinion formation is treated as a one-off process, i.e. one in which an agent commits to an opinion once and then holds it for the rest of the process (we refer to this type of model as static below) and (ii) models that treat opinion change as a dynamic process in which agents can flip between states according to certain stochastic rules (we refer to these models as dynamic below).
Research in class (i) has a long tradition in economics and computer science and strategic influence allocation is well understood. Models of class (ii) have typically been treated in the statistical physics literature [2], with a focus being on emergent phenomena and transitions between fundamental modes of behaviour rather than issues of optimality or considerations of strategic influence allocation. However, as (some) people frequently change their opinions in electoral debates or when commenting on news, models of the second class are far better suited to the type of problem outlined above. Hence the focus of the work proposed in this pilot is on theoretical advances in influence maximization for dynamic models of opinion formation, with the aim to pave the way for a fuller grant proposal on the understanding of influence and opinion dynamics in real-world settings.
Principal Investigator: Dr Markus Brede (Southampton)
Co Investigators: Dr Long Tran-Thanh (Southampton) and Dr Sebastien Stein (Southampton)