For this genealogy of feminist media studies, we attempted to develop a visualization that illustrates ties between people, institutions, and fields of study central to our discipline. Each step of developing this genealogy—such as formulating survey questions, setting up the survey, collecting the data, “cleaning up” the data, choosing a metaphor, a technology, and a design for the visualization, and modifying the result—created different opportunities for different kinds of feminist interventions. The contribution here is in thinking of a genealogy as a set of relations among people, fields of study, and institutions, visually communicating how the data was collected and appropriating a flow diagram as a social network analysis.
This is not an academic genealogy in the traditional sense, which usually shows mentoring relationships across generations as an academic lineage or an academic ancestry, such as the Mathematics Genealogy Project or the Academic Family Tree project. Instead, we wanted to reveal how people are connected both directly (with advisors or mentors) and indirectly (as scholarly influences), where and at what institutions these scholars and subfields exist (both as degree conferrers and as workplaces), and what fields and subfields of study intersect in feminist media studies.
We collected the data through a Google survey that yielded a tabulated data set. A total of 266 participants completed the survey online, submitting their names, the names of three mentors and five scholars who have influenced their work, the names of institutions where they received their PhD or highest degree and where they have taught, and three subfields that describe their work within feminist media studies. We also asked for the year they completed their degree, as we considered highlighting different generations of scholars, but eventually decided against it because there was already too much data to visualize.
Without counting duplicates, the data for the 266 participants included 536 mentors, 661 scholarly influences, 104 subfields, 123 institutions where respondents completed degrees, and 331 institutions where respondents taught at for at least two years. This resulted in 3,429 relationships. These numbers convey exactly the complex levels of interconnection we hoped to trace, but proved challenging from a data visualization standpoint. When placed in a circular layout, the data looked like the image reproduced in figure 2. Although the data is all in view, the visualization is unsatisfactory because it is overwhelming on a computer screen. A user would not be able to see a label without enlarging it or magnifying it, but as soon as the labels became visible, their connections to others would become too distant and removed from the screen view, thus making it impossible to see what a specific data point was linked to.
Using the “sticks-and-balls” layout typical of social network analysis resulted in visualizations captured in figures 3 and 4.1 These visualizations were a little clearer, but it remained difficult to read individual data points.
Figures 5, 6, and 7 illustrate a heavily curated version of the same visualization: in its initial state, when the circle representing Stuart Hall is rolled over, and when zoomed in or enlarged. There are multiple problems with these visualizations, and I highlight three here. First, there is the above mentioned problem of not being able to see what the lines are linking without magnifying the visualization. Second, the visualization does not offer any new insight into the data. Instead this method of analysis encourages looking for the most influential node, which could be easily acquired through a simple sort in the spreadsheet. Using social network analysis lingo, users would be looking at “ego-nodes” and their “alter-nodes.” This seemed contrary to the nonhierarchical, intersectional approach we were aiming for. And third, this visualization was also limited by the binary relationship of survey question and participant response, whereas we were looking for potential links that could result from further relations among the participant responses.
After considering different options such as chord diagrams, we decided to use (or, more accurately, misuse) a Sankey diagram.2 Sankey diagrams are traditionally used for energy flows from one location to another; they are named after Captain Matthew Henry Phineas Riall Sankey, who developed this diagram to visualize energy in a steam engine. A specific Sankey diagram that is always invoked as a foundational example in data visualization or infographics is one avant la lettre: “Carte figurative des pertes successives en hommes de l'Armée française dans la campagne de Russie en 1812–1813,” Charles Minard's 1869 graphical illustration of Napoleon's loss of French army soldiers in their goal to conquer Russia in 1812. It is regarded as one of history's greatest visualizations, as it involves six different measurements: the number of soldiers in Napoleon's army, their direction of travel, their distance from the Russian border to Moscow, the temperatures (which became a relevant factor during the winter), time, and latitude and longitude.3 Although it is visualizing people, Minard's graphic is considered a Sankey diagram because it visualizes flow and its direction: in brown, the graphic points to the army's march toward Moscow, and in black its return. The width of these colored zones represents the number of men in the army (each millimeter is ten thousand men).
We misused the Sankey diagram because we didn't make use of its purpose—visualizing flow in a specific direction—nor its potential to visualize the amount of influence or any other value with the width of the lines. Instead, we were interested in the Sankey visualization for its potential to be used as a social network analysis, while pointing to where the data originated from. As in the first tests shown above, social network analyses usually do not distinguish the participants in the visualization from their answers. With the Sankey diagram, it can immediately be made clear: to the left are the survey participants from which this visualization takes place. Without their participation and input, we would not have this data or resulting visualization.
Instead of using the design of a Sankey diagram to show flow by assigning values, we gave all fields the same value and ended up with a visualization that reflects the multiple connectivities of individuals, institutions, and fields. On the left we listed all the participants’ names, and on the right, all the other fields: scholarly influences, mentors, institutions, subfields. Overlaps are pulled to the middle, highlighting relationships and connections. For example, Elana Levine filled out the form as a participant, but was also listed as a mentor and an influential scholar by other survey participants. She works within the subfield of television, she attended the University of Wisconsin-Madison, and she now teaches at the University of Wisconsin-Milwaukee, resulting in multiple interconnections with other feminist media scholars. With this visualization, we were also able to convey what we were trying to capture: when a participant's name is chosen, not only do we see what people, institutions, and subfields they are associated with, but the other immediate relationships for each of these nodes and their overlaps, revealing what fields are tied to which scholars and to which institutions who compose a section of feminist media studies.
Above all, this visualization breaks down the linearity in academic genealogies and uncovers hidden connections and patterns of influence. So, is this visualization feminist? We are not sure. On the one hand, it began with a survey that returned tabulated data, which means it was already structured and inherited a specific kind of knowledge system and argument. On the other hand, it does go against a linear progression; it reappropriates a flow diagram meant to be used for resource allocation and redeploys it as a social network diagram; it places people, institutions, and fields of study as part of a genealogy; and it is clear where the data originates from. Perhaps a good start? Get in touch if you would like to try another version.