A fascinating debate in the psychometric literature happened in the 1970s about whether we should focus on "open-response" vs. "closed-response" questions when trying to understand the impact of behavioral interventions. Ultimately, that debate hinged on the limitations of statistical technologies at the time; we had means for doing rigorous causal inference with closed-response questions, but we were at a loss when using free-text. But empirical tools have changed dramatically, including the development of natural language processing (NLP) technologies. We introduce a new framework for causal inference with words, using natural language processing, motivated by our work on sexual assault prevention. We then discuss results from a demonstration study, to show how this new framework can be operationalized.
Speakers:
Mike Baiocchi, PhD, Assistant Professor of Epidemiology and Population Health, Stanford University
Jordan Rodu, PhD, Assistant Professor of Statistics, University of Virginia
Register: http://tiny.cc/EpiBioEmailList