American Debates over the Meaning of Labor Unionism Examined with Digital Humanities Tools

Sunday, January 4, 2015
2nd Floor Promenade (New York Hilton)
Vilja Hulden, University of Colorado Boulder
This poster presents selections from the data and argument of my book manuscript, Origins of Right to Work. The manuscript examines how it came to be that American labor unions are routinely accused of being bad for workers; it does so by focusing on employer, reformer, and union views of the closed or union shop in the period from roughly 1880 to 1930.

The  poster aims both to offer an outline of the debates over the meanings of labor unionism around the turn of the 20th century and to explore the potential of digital humanities approaches.  Thus, the poster focuses on those portions of the manuscript that have been shaped by my application of digital humanities tools in my research. The poster presentation mainly takes the form of graphs and charts, while supplementary demonstrations and visualizations will be available on a laptop.

The source material for the research presented in the poster consists largely of employer organization publications, union magazines, and general newspaper and magazine writings that are now digitally available on a significant scale, and to which text mining tools can therefore be applied. Since computers can process far greater amounts of text than humans, computer-assisted text analysis can highlight aspects of the material that normal scholarly reading might overlook as well as uncover relationships in data too extensive for a human researcher to process.

The poster displays the results of two main lines of computer-assisted research: topic modeling employer, union, and general publications, and experiments with applying network analysis to early-twentieth-century employer organizations.

Topic modeling — the automatic extraction of "topics" or groups of related words from a large set of text — is an increasingly popular tool in digital humanities, as it is a technique that, though based on complex algorithms, is fairly easy to apply and produces results that a human reader can interpret and at least partially verify. Topic modeling employer, union, and general publications from the late 19th and early 20th centuries allows one to outline the topics in a large selection of employer and union publications, analyze the change in topics over time, and also e.g. examine how prominent similar topics were in general-interest publications, and therefore in at least a section of general public discourse. While similar tasks can obviously also be carried out without the aid of computers, the systematization and scale topic modeling offers is useful in confirming one's intuitive grasp, calling attention to minor but significant themes, and helping one trace change over time.

The other technique applied in the poster, network analysis, is a much older tool that has been successfully applied to e.g. analyzing 19th-century U.S. women's organizations or the political power of the Medici family. Here, it is used to trace interconnections between U.S. employer organizations in order to examine questions like overlaps in membership, regional groupings, or the identity and significance of particular persons serving as "bridges" between organizations, and how these may have shaped the efficiency and power of these organizations.

See more of: Poster Session #2
See more of: AHA Sessions