The structure of segregation in co-authorship networks and its impact on scientific production

Co-authorship networks, where nodes represent authors and edges represent co-authorship relations, are key to understanding the production and diffusion of knowledge in academia. Social constructs, biases (implicit and explicit), and constraints (e.g. spatial, temporal) affect who works with whom and cause co-authorship networks to organise into tight communities with different levels of segregation. We aim to look at aspects of the co-authorship network structure that lead to segregation and its impact on scientific production. We measure segregation using the Spectral Segregation Index (SSI) and find 4 ordered segregation categories: completely segregated, highly segregated, moderately segregated and non-segregated communities. We direct our attention to the non-segregated and highly segregated communities, quantifying and comparing their structural topologies and k-core positions. When considering communities of both categories (controlling for size), our results show no differences in density and clustering but substantial variability in core position. Larger non-segregated communities are more likely to occupy cores near the network nucleus, while the highly segregated ones tend to be closer to the network periphery. Finally, we analyse differences in citations gained by researchers within communities showing different segregation categories. Researchers in highly segregated communities get more citations from their community members in middle cores and gain more citations per publication in middle/periphery cores. Those in non-segregated communities get more citations per publication in the nucleus. To our knowledge, this work is the first to characterise community segregation in co-authorship networks and investigate the relationship between community segregation and author citations.