Collective Dynamics from Heterogeneous Interactions in Socioeconomic Networks

Abstract

Understanding the drivers of human behavior is central to uncovering insights into how our social worlds are shaped. Modeling these social dynamics has widespread applications in computational social science such as driving information diffusion, targeting network interventions, economic forecasting, and designing public policies. This thesis centers on understanding influence, behavior, and social dynamics in socioeconomic networks to craft strategies that affect the evolution of these networks. Social influence is modeled using heterogeneous, group-level frameworks with large-scale behavioral data to uncover the temporal evolution of social dynamics and target network interventions. This enables analyzing inequality that only becomes emergent at the network-level to design networks that promote fairness and innovation. Hidden social groups have a dynamic influence on decision-making and behavior on networks. Using a community-based grounding for social influence in the context of decision prediction, we uncover latent influence dynamics behind social inequality while predicting future decision behavior. Further work studies how latent social groups evolve to predict a more comprehensive view of human behavior with observational datasets including features such as mobility and communication habits. Using a psychological account for social influence and latent groups, it was found that globally latent groups are most predictive of future behavior. These group-level influence dynamics reveal significant macro patterns driving broader collective phenomena and behavior, demonstrative of far-reaching applications in behavior change and enacting social policy. These aggregate influence frameworks have also been applied to salient real-world contexts, where regional social dynamics and interventions were estimated in regard to the interdependence of mobility behavior, social connectedness, and adherence to social policies in the US. Lastly, this thesis focuses on the design of socioeconomic network interventions to address issues of fairness and inequality, specifically in financial systems through network-level impact estimation. In the rising era of data availability, this dissertation demonstrates the utility of interpretable, heterogeneous frameworks for emergent collective influence dynamics on socioeconomic networks.