Palladio is a free web-based network tool. It allows users to upload their own data to create both maps and network graphs. Users don’t need an account and can download project to their computers or save the url to return to their projects.
When embarking upon a networking project, I think it’s important to be conscientious about what kinds of relationships networking can best represent. The idea is to see relationships between data that otherwise are hard to conceptualize. My trials with Palladio have some good examples where networks effectively convey information, and good negative examples as well. As I describe how to use the tools, I’ll included some commentary about what I learned in the process.
After entering http://palladio.designhumanities.org/#/ into your address bar all you have to do is press “start.”
From there you’ll have to load some data. As far as I can tell you can’t create data in Palladio. My data were stored in .csv documents and prepared for me by my instructor. Hopefully you’ve organized your data before you’ve come to this point. I loaded the primary data set I would be drawing from- in other words the spread sheet all other sheets would relate to. From there press “Load.”
Your primary dataset will appear in the “data” tab. Here you’ll connect other data tables to make relationships. In this step you’ll already start making decisions about what kinds of relationships you want to show by what data you choose to add extensions to. In my project we were concerned with showing relationships between people and places, so this is the data we linked.
By clicking on a data field (in my case “where interviewed”) the editing window opens. Here I selected “Add a new table” to start creating relationships. I uploaded my “locations” spread sheet. Then I selected the “subject interviewed” data field from my primary dataset and added the locations of their enslavement. Finally, in the Enslavement table that I just uploaded I clicked on the drop down “Extension” menu and selected “Locations” to link the two datasets. At the end I had this set of data:
With all my data uploaded and connected, I was ready to start exploring Palladio’s visualizations.
Mapping with Palladio
It is possible to map with Palladio, but that doesn’t mean you should. Creating the map is fairly easy. Select “Add new layer” then the above editing window appears. Just select what kinds of points you want and the two datasets you’d like displayed (source for me was interview location, target where_ensalved). You might have to “ctrl -” to zoom out enough to see the “add layer” button.
For some projects the mapping feature is probably sufficient, but for mine, it was far from. Although this map does show the movement of individuals from their places of enslavement to the location of their interview, the directionality is not clear. Palladio does allow for layers, but the information available for display is not nearly as rich or customizable as CartoDB and other GIS specific applications. This map does convey that there was a significant movement of people after slavery ended, but other questions can be better explored and asked in GIS applications.
The networking tool is pretty intuitive. All you have to do is select that data you want to relate (a source and a target). The “facet” tab allows users to focus in on certain aspects of their data (for example in the above visualization I could limit the interviewees to those over the age of 80). In the above example I selected the interviewer as the source data and the interviewee as the target. The resulting visualization shows which interviewers interviewed which ex-slaves. I would say this is a good visualization because the viewer can see the intended information with ease. So, instead of looking at a spreadsheet organized by interviewer, I see all the their interviewees in one page. This visualization doesn’t elicit many novel questions, but it does provide solid information.
Within the source data, you can select other facets to relate. In the above example I visualized which interviewers met with male and/or female informants. This helps me ask questions about gender bias in the interviews, or how gender norms in 1936 influenced the interviews.
Networking the relationships regarding topics had mixed results. The graph above, showing the relationship between the type of work the ex-slave was engaged in during their enslavement and what topics they discussed is a rather successful graph. I would have expected there to be more variation in the topics, but upon seeing this graph I was reminded that the interviewers used a script, asking about particular topics and really engaging in natural conversations. The few outliers probably represent occasional spontaneity arising from the script.
The above graph is a fairly clear example of an instance where networking doesn’t work well. The relationship between which interviewees and which topics they discussed. The result is too highly clustered and the size of the nodes too large to make anything out. This unfortunate grouping is due to the afore mentioned scripted. Almost every person discussed the same things sine they answered the questions they were being asked. This consistency also reflects what Mark Twain refers to as “corn-pone,” when enslaved or previously enslaved people who tell white people what they wanted to hear. It could be said that there was already a script, even before the WPA produced one.
A Brief Reflection
Networking visualization can be a very powerful tool when the investigator is conscientious of what information the graphs can provide. Palladio only computes bimodal networks, which is usually the best thing. Looking at these networks allowed me to ask questions about why these relationships looked the way they did. So, why did female interviewers tend to meet with female informants? Why were ex-house slaves the only people to discuss “mammy?”
I was also able to draw some preliminary conclusions. The WPA script was, arguably, effective. Conversations stayed relatively on script and recorded consistent types of information. These observations would lead me to look at the scripts themselves to see to see if I’m correct. Which leads to another take-away: no tool replaces close reading.