Social Influence Analysis (SIA) in Online Social Networks

Online social networks have become globally ubiquitous, and therefore are an arena where important social phenomena can be observed: e.g. diffusion of (dis)information, social and political polarization, as well as distribution of hate speech and radical content. To understand their spread and effects, it is important to analyze and model the notion of social influence in online networks. For empirical modeling, it is crucial to study the relational nature of interactions between users of the networks, together with analyzing the content of communications between them. This research focuses on investigating social influence in online social networks as the fundamental principle for information diffusion that needs to be modeled, parameterized, and measured. According to Google Scholar, in the first 5 months of 2022 alone, scholars from multiple disciplines roughly produced 30,000 papers dealing with the concept of influence in online social networks. This indicates that this research is indeed a multidisciplinary challenge. Defining the concept of social influence is not trivial and application-specific. However, understanding the concept will contribute to modeling information propagation in social networks beyond traditional network metrics. We build on the early definition coined in psychology, which states that social influence is “a change in a person’s cognition, attitude, or behavior, which has its origin in another person or group” (Raven 1964). This definition is sufficiently broad to accommodate approaches to social influence from diverse application domains. As a starting point, we consider influence maximization with regard to disinformation campaigns and marketing as core application domains. Customers are highly connected to each other, therefore focal consumers’ (buying) behavior is affected by their peers and not only by controllable firms decisions. Thus, marketing science developed concepts associated with social influence, such as social contagion, word-of-mouth, opinion leadership and influencer marketing. The goal of our research is to study these phenomena from a multidisciplinary perspective, combining methodological developments in social network analysis with applications to social science. This will lead to a suitable simulation environment that can represent realistic network sizes as well as different influencing variables of the relationships and information flow in different (and eventually dynamically changing) network structures. There are several factors precluding a direct practical exploitation of existing methods: The influence propagation model should realistically model interactions between agents, yet be simple enough to be effectively calculated over a large network. The propagation model may in turn require additional complexity in the representation of the network, as, for instance, different types of nodes, edges, diversity or influence levels. Creating an adequate simulation environment will be complemented by addressing the lack of suitable (numerical) indicators for measuring social influence. Only such measures allow further steps towards optimization of influence and network structures as well as the evaluation of real networks. Of course, in real environments, ethical and data privacy issues have to be addressed appropriately. Using measures and theoretical insights into (optimal) influence networks, we can finally judge on the benefits or dangers (e.g. in the context of disinformation campaigning) that specific networks expose.