Career Cocoons: Analyzing Occupational Mobility with Graph Embedding Model

The Information cocoon is discovered in social media, which is a phenomenon of biased information acquisition and opinion reinforcement. We find that occupational mobility is analogous to information exchange in social media and the mechanism of career choice may cause an information cocoon-like effect. The study identifies and measures the information cocoon-like effect on occupational mobility, and names this effect the career cocoon. We first construct a complex network to model occupational mobility using the real-world dataset, where the edges of the network represent connection tendencies between occupations and the weights of edges represent selection preferences of individuals for occupations. The information cocoon-like effect is identified based on four characteristics: separation, homogeneity, polarization, and system efficiency. Subsequently, the graph embedding model is used to map the occupational mobility network into a vector space. We design indicators to evaluate the ability of career trajectories to span the occupational space, thus enabling the quantification of career cocoons. This study introduces information cocoons into occupational research for the first time, and utilizes data-driven approaches as well as intelligent techniques to measure career cocoons, providing new insights into occupational mobility research.