Research

How can natural language processing techniques be improved to conduct analyses of more sophisticated psychological constructs? How can psychological theories be used to tackle fairness and bias related issues in NLP? Relatively little natural language processing research has focused on measuring complex psychologically relevant variables in a manner that enables inferential testing. Accordingly, we are interested in devising approaches that allow us to both exploit natural language processing techniques and make inferences about a priori specified phenomena. We are also interested in using theories from psychology to improve issues related to bias and fairness in NLP.

How can spatial variation of psychological constructs be used to investigate regional association between psychological phenomena and real-world outcomes? Much of social science research necessarily focuses on constructs that have been abstracted away from the actual environments in which they occur. Further, considerable efforts are typically made to isolate effects in order to establish parsimonious theoretical and statistical models. Such strategies are essential for building basic knowledge of human phenomena. However, while such estimates can be useful for understanding a phenomenon in general, they are typically not useful for making location-specific predictions, which requires accounting for environmental and spatial variation, not marginalizing over it or treating it as random noise. In this line of work, we use geo-spatial analysis to investigate how various psychological constructs are distributed across space, and how we can use this variance to make predictions about real-world phenomena.