Social Identity Detection
We have developed a toolkit – ASIA (Automated Social Identity Assessment) – that allows researchers to create a classification model which uses linguistic style to assess which social identity is salient during writing.
What does ASIA measure?
Self-categorisation theory suggests that our cognitions, emotions and behaviours are affected by the social identity that is salient in a given moment. Assessing salience is difficult due to reactivity and lack of introspection, and therefore largely confined to experimental paradigms. ASIA allows researchers to understand which out of two social identities is salient – it is a relative measure of social identity salience.
How does ASIA work?
ASIA uses linguistic style indicators based on the idea that the group prototype informs behaviour, including linguistic behaviour. Put differently, when switching from one identity to another, we follow different norms that are reflected in the style in which we write. This creates homogeneity of style within a group, and heterogeneity between identities. Using machine learning, a model that can distinguish between the salience of two identities can be trained and cross-validated. We use naturally occurring data from online forums to train our ASIA models, and test their validity across platforms and in experiments.
Which ASIA tools have already been trained?
We currently have trained models to assess the salience of the following identities:
Parent v feminist identity: Koschate, M., Dickens, L., Stuart, A., Naserian, E., Russo, A., & Levine, M. (2019, September 6). Predicting a Salient Social Identity from Linguistic Style. https://doi.org/10.31234/osf.io/zkunh
Entrepreneur v libertarian identity: Cork, A., Everson, R., Levine, M., & Koschate, M. (accepted). Using Computational Techniques to Study Social Influence Online. Group Processes and Intergroup Relations (Special Issue: Rethinking the Group – Group Processes in the Digital Age)
- EPSRC Innovation Fellowship “Psychological Identity in a Digital World” (EP/S001409/1; Fellow: Dr Miriam Koschate-Reis)
- EPSRC studentship “Detecting and preventing criminal network involvement from digital footprints” (1929614; Alicia Cork)
- EPSRC “Privacy Dynamics: Learning from the Wisdom of the Group” (EP/K033425/1; PIs: Prof Mark Levine, Prof Alessandra Russo)
- EPSRC “Identi-scope” (EP/J005053/1; PI Prof Mark Levine)