Humans are social beings and the relationships that we form with each other affect our health, happiness, and financial outcomes in life. However, even given the large amount of evidence that social networks are important in many aspects of life, a systematic large-scale study of the role of social networks, and the magnitude of their impact has not been carried out. In this PhD project, we aim to take the first steps towards performing such a study. Historically, one of the main barriers that have hindered this line of research is that network datasets of adequate size and quality have proven
too challenging to establish. We will breach this barrier by using extensive (and globally unmatched) registry data for the entire population of Denmark to construct nation-scale social networks. Trough collaborations, we will combine the subject-specific knowledge of researchers from sociology, economy, and politology with the technical and computational expertise of researchers from section the of Cognitive Systems at DTU.
There are many potential ways to investigate the effects of network position on in-dividuals lives. One promising way is to bring the network view into the deep learning paradigm. This approach will then rely on network embeddings that incorporate nodal information. Such nodal information could be based on embeddings of life-events using techniques developed in natural language processing and currently being developed on for life-trajectory embeddings. The network embeddings can then hopefully be generalized to include time-dependencies and higher order structures to enhance the performance of these models.