Measuring Stereotypes With Statistical Models Built On Social Theory

Social identities, the labels we use to describe ourselves and others, carry with them stereotypes that have significant impacts on our social lives. Our stereotypes, sometimes without us knowing, guide our decisions on whom to talk to and whom to stay away from, whom to befriend and whom to bully, whom to treat with reverence and whom to view with disgust. Unfortunately, methods for measuring these stereotypes that are socio-cognitively plausible, parsimonious and rapid to collect remain elusive. At a minimum, this complicates our understanding of the effectiveness of different approaches to reducing negative stereotypes.

This talk focuses on two recent projects that provide a new set of conceptual and methodological tools for measuring stereotypes. I will first discuss a new method to measure how identities are portrayed in newspaper data, with applications to religious identities during the Arab Spring. I will then discuss another new method to extract stereotypes from Twitter data, with applications to the Eric Garner and Michael Brown tragedies. In both projects, as opposed to testing social theory using a generic statistical model, I develop (parametric Bayesian) methods crafted around the theories themselves. This work has resulted in new socio-theoretic insights into stereotypes as well as new methods that push the state-of-the-art in machine learning.

Presentation by:

Kenneth Joseph