Recent social movements research that has excited and motivated me groups around three themes: 1) community-level effects on movements (Reger 2012, McAdam and Boudet 2012), 2) the path-dependent nature of movements (Blee 2012), and 3) the application of computational methods to study social movements (Hanna 2013) and culture (Bail 2013). My research fits loosely at this junction: I use computational methods to study the structure and culture of local social movements over time.
To do so I, like many others, conceive of local movements as fields. I formalize fields in two ways: a social field consists of 1) a structure—a set of actors that are in some way related to one another (DiMaggio and Powell 1983), and 2) a culture—taken-for-granted assumptions that both enable and constrain action (Jepperson 1991). While network analysis is an established way to measure structure, quantifying culture has proven more difficult.
The challenge: quantify both the structure and culture of social movement fields to analyze geographical differences within movements and change/stability over time. My movement of choice: feminism.
My research up until now has focused on measuring the local feminist movement fields in two cities, Chicago and New York City, in two time periods, the first wave and the second wave. Fitting with previous research on different movements cited above, I find the feminist movement fields in these two cities were distinct, and each field exhibited distinct path-dependent trajectories. Mapping the structure of each field, however, was not enough; I also had to analyze their respective cultures.
To give a taste of my research agenda I highlight below a few of my findings and methods, ending with hopes for future research along these lines.
1. Substantive Findings: Structure, Culture, and Local Feminist Movements
Using network analysis, I found that the structure of the feminist field in Chicago in both waves was centralized and embedded, while in New York City it was decentralized and independent. In the first wave in Chicago, for example, feminists worked largely through existing left organizations, such as the Women’s Committee of the Socialist Party, and through the dominant and central women’s organization Hull House. In the second wave the feminist movement was again centralized, this time through the Chicago Women’s Liberation Union, which kept close ties with other left organizations. In New York City, alternatively, the feminist movement was decentralized and was much more independent from the larger left in both the first and second wave.
The culture underlying the feminist fields in these two cities was also distinct and, like the structure, the distinct cultures were continuous within each city over time. To formalize this culture I use the term cognitive framework, which is similar in some ways to institutional logics. Using a variety of text analysis methods on the literature produced by feminist organizations in these two cities I identified two distinct cognitive frameworks embodied in feminist organizations. Women in Chicago in both waves assumed change happens through institutions and is achieved by winning concrete reforms that affect daily lives, while women in New York City in both waves assumed that change happens through individuals and is achieved by raising consciousness about the general political roots of the problems women face.
I found that these local feminist fields persisted between the waves via two mechanisms. Organizations institutionalized cognitive frameworks that persisted in these communities over time (Greve and Rao 2012). Even if organizations themselves do not persist, the cognitive models they institutionalize do. But, I also found no less than 39 organizations that did persist between the waves, providing many concrete opportunities for the direct transfer of ideas, tactics, and models between first- and second-wave organizations. My findings support and expand on the findings of many who have noted continuities between waves of feminist action (Reger 2012, Reger (ed) 2005, Staggenborg and Taylor 2005, Staggenborg 1996, Whittier 1995, Rupp and Tailor 1987).
2. Methodological Advances: Measuring Cognitive Frameworks
While network analysis has a long history in social science, identifying and measuring cognitive frameworks, or culture, required a new set of tools: automated text analysis. Sociology has a developed theoretical framework for understanding culture as I define it—the concept of institutional logics—but we still struggle to find a way to precisely identify and measure these underlying frameworks. Thornton, Ocasio and Lounsbury (2012) suggest using ideal types; others, including Armstrong (2002), begin with the data and systematically construct underlying logics from it. The first method risks the problem of forcing a square peg (data) into a round hole (ideal types), while the latter method risks confirmation bias. Recently, scholars have turned to automated and semi-automated text analysis as a tool to study text-based data (see also the December 2013 issue of Poetics), and I use these tools to uncover the cognitive frameworks underlying social movements.
In the case presented above, I used four quantitative methods to identify feminist cognitive frameworks. First, I identified the words that most defined one city compared to the other. The defining words in Chicago tended to be particular and concrete (for example school, abortion, Nixon) and those in New York City tended to be general and abstract (for example feminist, radical, history). I next ran a topic modeling algorithm called Latent Dirichlet Allocation on the same text. The literature produced in Chicago was more often dominated by concrete topics, like Labor Laws and Anti-War Organizing, while literature produced in New York City contained more abstract topics, like Feminist Theory and Women’s Experience. Next, I tested the hypothesis that the language used in Chicago was more concrete and the language used in New York City was more general by calculating a concreteness score for each text. I constructed the measure from a lexical resource called WordNet which, among other things, organizes nouns and verbs hierarchically, through hypernyms and hyponyms. I calculated the path for each verb and noun to get to its root (general) word, and found that the texts produced in Chicago were on average more concrete compared to the texts produced in New York City, confirming my hunch. Finally, using a named entity recognition algorithm, I found that organizations in New York City were more likely to mention individuals, while organizations in Chicago were more likely to mention state institutions and organizations. My final step was a qualitative analysis of the same text, which filled out the quantitative word-based analyses and led me to the cognitive frameworks detailed above.
Quantitative text analysis reduces unstructured text into informative groups of words that can suggest latent categories within the text. Identifying these categories can then shape and provide context for qualitative analysis.
3. Future Research
Research on contemporary movements must account for the role of the internet in collective action but the vast amount of stuff, often text, on the internet produces unique methodological challenges—but not insurmountable ones. I plan to use a variety of computational methods, in particular computational text analysis, to study contemporary feminist movements, something that has not yet been done in social movements research.
3 responses to “Feminism, Culture, and Computational Sociology”
Pingback: Emerging Stars in Social Movement Research | Mobilizing Ideas
Pingback: Emerging Stars in Social Movement Research | Mobilizing Ideas
Pingback: Text as Data: A Call to Standardize Access and Training | Mobilizing Ideas