Shifting attachments to social groups are a constant in the modern era. What accounts for variation in the strength of group identification? Whereas prior work has explained variation between individuals, this article develops a network-analytic theory of within-person changes in group identification. The authors hypothesize that identification is positively related to occupying positions characterized by local clustering -- having contacts who are mutually interconnected -- and global integration -- having contacts who are a part of a community different from that to which a focal actor belongs. They use the tools of computational linguistics, combining pretrained GloVe word embedding models with the retrofitting-based finetuning model Mittens, to develop a language-based measure of identification. They validate this model using a widely used survey-based scale of organizational identification, and apply the model to a dataset of pooled internal email communications from three disparate organizations. Analyses using linear regression find general support for the theory.
Bio: Lara is a PhD candidate in Macro Organizational Behavior at Stanford Graduate School of Business. She is broadly interested in using computational methods to better understand the dynamics of identity and culture at work, both at the individual and the organizational level of analysis. Her work employs a broad set of tools, including word embeddings, topic modeling, network analysis, surveys, and field experiments. Prior to Stanford, she received bachelor’s degrees in Computer Science and Psychology from University of California, Berkeley, and a master’s degree in Industrial Organizational Psychology from San Francisco State University.