Scaling-Up Content Analysis Projects Using Crowdsourcing: American Cities’ Policing of Social Movements

• In Event: 420 - Research Poster Session. Communicating Sociology

Abstract

My research uses the comparative leverage provided by the Occupy movement’s activities in over 190 U.S. cities to investigate when, in the course of a protest campaign, and under what circumstances, political context, police characteristics, movement characteristics and on-the-ground interactional dynamics matter more or less in determining police and protester interactions -- questions unanswerable by small-N studies.

Answering these questions requires the categorization and quantification of dozens of variables across multiple units of analysis from over 9,000 news texts. The poster I would present will describe the tools and workflow necessary to complete work at this scale. Specifically, the poster will introduce the substantive question my research asks; explain how that question is only answerable with a large-N comparative study; show our classificatory (coding) scheme; show how we have used trained expert coders to classify text at one unit of analysis (the event/situation); explain how the other classificatory tasks required by the scheme have been reduced down to discrete judgments and converted into an interface for completion by crowds (using CrowdFlower); explain how those crowdsourcing platforms work generally and how we have tailored them for our work; show the data the work has generated in our database; report some of the exciting findings (from multi-level event history and sequence analysis models) that the data have made possible; explain how the tagged and classified text will be used as a training set for machine learning algorithms; and sketch out how the same process can be used for others’ text analysis research.

Author

• Nicholas Adams, University of California-Berkeley