As I prepped for the ImagePlot workshop in the seminar this week, I was reminded again about the difficulty DH scholars often confront in translating their work outside of (or even sometimes to!) DH circles. ImagePlot can be very bewildering the first time one encounters it, and I remember the first time I presented on it at a conference I spent most of the time fielding questions about what ImagePlot *is*, rather than what it can do or show us. I think there are several reasons for this. First, ImagePlot as a visualization tool requires a decent amount of abstraction: it’s taking a lot of images, making them small, and arranging them according to particular qualities such that a pattern emerges. Because of these things, it can be difficult to figure out exactly what it is you’re looking at, and it can take some time to explain that. Second, while ImagePlot is open source, it’s not eminently accessible: it involves a number of steps that are easy to get stuck on if you don’t know Command Line, don’t have the right settings or directories, etc. It’s even more confusing (speaking from experience) when you don’t know what the different steps you’re doing are actually doing, or why they are necessary. Finally, ImagePlot as a method is what we might call distance visualization (as opposed to distance reading), because it gets away from particular images and looks at a large collection of images in order to discern patterns. These distance approaches are themselves tricky to translate to traditional scholars, because it can be difficult at first to see what they show us that we couldn’t figure out otherwise–what it is that’s new or original about them.
Because of these realities, it’s especially important that DH scholars practice and be able to translate their work to audiences outside of DH. If no one can tell what it is that we’re doing, then what we’re doing actually isn’t worth much–it’s not helping anyone then. So we need to be able to explain what we’re doing: what our questions are, what our methods are, and, importantly, how the technology works that we’re using to accomplish these things. In my own experience, this involves a lot of trial and error in order to see what makes sense to folks, and what does not. Or perhaps rather what is recognizable, and what isn’t. For ImagePlot, this involves being able to quickly explain what the final product of the process is, and what it shows us. I don’t typically go through each step of the process–that would just be more bewildering–but I have developed quick summaries of the overall process because that helps folks understand what’s happening in it. I learn a bit more every time I translate this work to a new audience, and modify the explanation to help make it more understandable.
I think that’s crucial to improving the legibility of DH as a field, and is something that all DH scholars have to practice. It’s not easy, and it is a bit of an unfair disadvantage: we don’t get to operate under the assumption that everyone knows what we’re talking about, the way folks do when we talk about a symbol or theme in a novel. We have to both do the work and constantly explain it at the same time. But I think this can also work to the advantage of DH studies. It keeps us honest, and keeps us critical about the work we do. And, hopefully, it results in better scholarship in the end–more critical scholarship that is more able to help people.
As I’ve been working on my project for the DH seminar this semester, it’s occurred to me in several of our discussions that mapping and visualization aren’t so different. Indeed, we might even say that they’re the same thing in different terms: maps are abstractions of space meant to make large quantities of space readable, along with their social and cultural attachments (towns, regions, nations, etc.). Visualizations are abstractions of data, meant to make large quantities of data, and trends within them, visible. So what I’m doing with ImagePlot–visualizing game narratives–really isn’t so different from someone doing a mapping project visualizing the *where* of a set of data.
Many DH and software studies scholars have noted how visualizations rely on these abstractions, the curious concoctions of distance, transformation, and relationship that visualizations require. Yet as we read about mapping projects, and particularly spatial humanities this week, it struck me that all spaces rely on abstraction, or perhaps mediation. It’s easy to see how maps are abstractions of space, but even as we navigate the spaces around us, the way we perceive and navigate those spaces is entirely dependent on sensory inputs that are interpreted and rendered to consciousness by the brain. In this sense, all spaces and all of our interactions with spaces are built on abstraction, even in our moment-to-moment experiences that seem to be immediate.
I think this realization is important because it deconstructs dichotomies that say some things are “real,” “actual,” or “natural,” and other things are “constructed,” “virtual,” “fake.” The history of maps demonstrates how untenable that distinction is: the abstractions of space have very real, direct consequences for our experiences of space. And I think we see something similar in contemporary gaming, and attempts to write off virtual or digital spaces as not being “real.” This argument is especially prevalant in online trolling and harassment cultures: it’s ok to treat people like garbage, because it’s all online and therefore has no consequences. The realization that all of our spaces are abstractions, that all spaces are virtual ones, helps us reject that premise that is causing a lot of harm in contemporary social spaces.
For this week I’m updating the project plan for the Archive of Player Experience. As I laid out in the last proposal post, for this semester I plan to complete the initial stage of the Archive using ImagePlots of the games I’ll be using in my dissertation on queering game narrative: Gone Home, The Vanishing of Ethan Carter, SOMA, and The Talos Principle. I have completed a worksheet planner for the project, and am including the link here:
As I worked on a separate project in the past week (on Pokémon GO and narrative construction of reality), I realized a way to frame my argument for the Archive. My interest in using ImagePlot with games has always been to establish a better handle on narrative variance in games–how each player has a different experience of the game, and to what extent. By comparing different players’ playthroughs of the game, we can get a sense of the general form of a game’s narrative, including the critical mass of similarities that players’ experiences generally share. What I’m measuring, then, is how games cohere as a narrative while still allowing players to have different and emergent experiences. Ultimately this determines how a game shapes its reality for the player, including how much variance and possibility that reality has built into it. This allows us to account for difference while still acknowledging commonalities.
This week I was especially drawn to Matthew Jockers’ blog post, “A Novel Method for Detecting Plot” (Jockers). In the post, Jockers discusses how his project of tracing sentiment in 19th century fiction using sentiment analysis also lead him to visualizing the plot structures of the novels he looked at. Basically, he found that the shifting emotional valences in a novel (measured through certain words and sentiment markers) are also a proxy for the rising and falling action of the plot. Presumably the moments of positive sentiment are also the high points of the plot, and the moments of negative sentiment are the low points in the plot. There are still some questions and potential problems there––for example, aren’t the low moments also the ones with the most conflict and action?––but it’s an interesting way of visualizing how the narrative moves and varies throughout a work of fiction.
I’m very interested in the possibilities of visualizing narrative forms and structures because it allows us to see how narrative develops and grows: in other words, how it is a living process. This is one of the central goals of my ImagePlot study of narrative variation in games. If we can see how much narrative changes even within a single text (like a game), then we can get a better handle on how narrative operates and the potentials of what we can do with it. The idea that narrative has a shape is interesting as well: what does it mean that the abstract and affective qualities of narrative have specific forms? Is the shape of a narrative ever anything more than an abstraction or a metaphor? Or if it is just that, what meaning does it have? It seems like common or repeated shapes could act as a key of sorts: not in a determistic sense as unlocking every story in a particular form or archetype, but in terms of establishing a set of common meanings that relate to each other across stories.
One of the issues I see with this approach (one that I’ve encountered with my own ImagePlots) is that while these visualizations allow us to see change over time, they’re not great at capturing the fluid, variable, active elements of the experience. In other words, these visualizations present the plot, whether it be represented in rising or falling action or in images, as something static––the graph itself does not move, nor does it capture all the possibilities present in each moment of the text. It is an arresting of motion that is itself quite dead. That might be a necessary evil in order to do this type of work and analysis, but I still wonder if the tools themselves could get better at demonstrating some of the same action that they represent.
Jockers, Matthew. “A Novel Method for Detecting Plot,” http://www.matthewjockers.net/2014/06/05/a-novel-method-for-detecting-plot/, June 5, 2014.