The “Holy Grail” of Digital Humanists – Some Thoughts About “Understanding Data” (for Humanists)

As a reader of blog posts, you might be well aware how important role commentaries happening on blogs can play in the establishment of a specific scholarly discourses. Running parallel with more traditional forms of scholarly communication, blogging gives us unique possibilities to quick reflections, to rapid exchange, to give sneak peeks into work in progress, or to just empty our mind to a white surface functioning similarly to Dumbledore’s pensive. In this piece, you can read our guest author’s, Stefan Karcher‘s reflection on Jim Casey’s recent blog post, Taste the Data. Enjoy!

When the blog post „Taste the Data” by Jim Casey (CDH Princeton) was nominated for republishing at Open Methods, my reaction was a little bit divided. On the one hand the nomination had my full support, but on the other hand, I couldn’t held back a tiny critical remark: great didactic idea, but … really??? I followed my first impulse to mock the article a little. It seemed too silly to explore “data” with jelly beans, Peeps and various sauces in order to discover important information and structures in a data set. Do most people not take in important information about their environment with their eyes? Isn’t it the same when dealing with data?

There are countless books on data visualization and discussions (or rather an exchange of beliefs) on how to best visualize data to convey the most important information (Cao/Cui 2016; Grant 2019; Miller 2017; Nussbaumer Knaflic 2017). Charts and maps of various kinds have been optimized for different “use cases” for years, and the advantages and disadvantages of different ways of presentation are explained in manuals. This opinion is based on the fact that information conveyed through visualization contributes to a more precise understanding of what is actually meant by the author of a text. A simple example is the following sentence: “He drew a beautiful cow.” With an additional picture, e.g. here or here, it is much easier to interpret this sentence and to put it into the right context. The example is striking and underlines that data visualizations are the best way to give the reader access to the hidden information in the dataset. Of course, visualizations are also highly manipulative and can control the readers. Yet, this isn’t something I would like to discuss here; especially since I have already talked about the principle of “sincerity” in “5 Thesen zu ‘Warum wir visualisieren sollten’” (Karcher 2017) and in an additional InFoDiTex-blog post (Karcher 2018).

But “Taste the Data” takes a completely different approach. In the tradition of “Data Cuisine”, another sensory organ is addressed to capture the informative content of data sets: “taste.” The “Data Cuisine” uses statistical information in a “shared research experience” to create dishes that taste well, but also represent information included in the data sets. For example, they transferred the proportion of women in science in Belgium, Japan, and Turkey onto fried eggs (Jansen 2016). Taste and visualization are obviously combined in this representation and it is easy to imagine how the “data cuisine” works: The important statistical information is conveyed by weakening or strengthening a certain taste nuance in a dish. As mentioned in “Taste the Data,” the process from a recipe to a finished dish is enormously cooperative and communicative process, which is obvious in the “egguality” example. It must be discussed whether the proportion of women is represented by the egg yolk or the egg white. Does this decision not perhaps distract from other information in the data set? Is it only a matter of taste or is it also “You eat with your eyes”? — Would you still ask yourselves these questions while choosing the colors for a simple bar chart? Or does habit and aesthetic perception predominate?

“Taste the Data” takes the didactic challenge of creating an awareness of the data by making taste-related decisions in the team. Are all data tasty like jelly beans or is there data that puts mustard on sweets? Students are encouraged to make decisions for and against a certain “recipe,” which becomes a presentation method. The data is made tastefully comprehensible, which, in a next step, leads to the transfer of the interpretation into better and more meaningful visualizations. Similarly, data analyses have already been set to music in literary analysis, so that the visualization has been replaced by “audio” (e.g. the project Music from text).

This way of dealing with data analyses is certainly not suitable for the publications by humanists. In 2020, however, it should no longer be a big secret that in the Digital Humanities the view of large amounts of data is sometimes over-strained and new didactic approaches are needed to find relevant information in the data sets. New perspectives on data – here based on taste – point researchers to new questions, which can be answered and presented using familiar methods. “Taste the Data” presents this in the iterative process of data interpretation as a step between information filtering during the process of data mining. Casey observed that the students “began to raise creative questions,” which finally led to clearer, more intuitive, more meaningful (“humane”) visualizations.

Such a didactic approach can only enrich DH-research. Although digital humanists usually describe themselves as innovative and modern, they often follow the usual logic of established research in the humanities. In principle, research in the humanities works – to put it simply – as follows: Within one’s own discipline, a scholar/student deals with a specific topic (reads publications, looks at sources, etc.). Then he/she discovers a desideratum within the field of interest and begins to ask questions (Why does this gap exists?). If we exclude the ideal case that the clue to find the desideratum was discovered through an unknown source, a humanist would now begin to look for sources that helps to answer the research question. Using the methods of discipline, the sources are interpreted and embedded in the research context, aiming to produce a progress of knowledge within the research discourse. In this way or similar, seminar papers, dissertations, articles, or monographs are produced.

The research procedure of digital humanists is quite similar, but differs in one important aspect: “Data” are obtained from material sources (texts, images, finds etc.), which are then processed using “new” D (digital) or C (computational) methods. Digital Humanities enrich the source analysis with a broader data basis, which helps to weight arguments by new (often quantitative) insights. This kind of a data-based knowledge is a great challenge for many humanists, because some continue to be anxious to the methodology. Also, the process of getting data from source is sometimes hard to follow. Sometimes, we are blinded by spectacular visualizations, which lead to spectacular interpretations of data sets. For this reason, it is more important to understand the data than to interpret it.

“Understanding Data” seems to me like finding the Holy Grail of Digital Humanist. Because we often work on a “data-based” basis, we have developed and become familiar with many methods in recent years that support data interpretation and help us to recognize connections. “Taste the Data” caricatures the approach of humanities work to data interpretation by collecting the data as a source and making the data analysis into an inner source analysis. It is no longer a question of what contribution the data makes to answering the research question, but rather what research questions arise from the data. Digital humanists should adopt this perspective in order to detach themselves from their disciplinary perspective and from their question and to be able to look at the data set without bias.

On a science theory level, this results in two consequences that provide digital humanities research with considerable added value:

1. Those who generate “data” from sources transform the haptic material into digital source material, in which basically the same information is contained, but which can be technically processed with new methods.
2. Digital source material is not suitable for singular projects, but can be further used in many ways if researchers without the bias of a pre-formulated research question open up the data set. While the first aspect touches more on the area of recognizing the creation of a database as a scientific achievement in its own right (after all, no one would seriously deny this when creating a scientific edition), the second aspect has a dramatic consequence: if data from sources are actually produced in such a way that they can be disclosed for later use, there must inevitably be further insights than those already produced by the original project.

Illustrated by a simple example from the established humanities, even if the editorship of a full edition of the author’s oeuvre is driven by a particular research interest, the scientific edition of the works is accessible to other fields of research. It is obvious and self-evident that a collection of letters, diaries, documents, etc. is not only useful to the project of the person who originally published it. But does one understand data on the same level as the source edition? Why is it not possible to continue using them in the same way?

This is where “Taste the Data” makes a remarkable contribution. The didactic approach addresses the challenge of obtaining information from source types other than the usual (text, image, etc.): namely data. While methods of textual exploration have been established for centuries, the methods of data exploration in the humanities have been marginalized. Sure, there are explorative tools and various possible ways to visualize. But do they help when it comes to developing new questions and insights from a data stock – particularly those “new” questions that one hopes to find in the DH? “Taste the data” shows a completely unconventional way of dealing with a data set and developing questions from it that are not immediately obvious.

I consider it necessary to deal with this didactic approach for two important reasons. (1) It seriously takes the potential that DH has created in recent years. We can now convert a multitude of sources into “data sources” and then analyze them, but the so often “questions we have not thought of yet” are hardly ever asked. So, although DH do innovative research, they ask very conventional questions. (2) It ensures the re-use of data sets that are published after the end of a project. While some people have evaluated the data to answer their own questions, others can develop new questions using innovative data exploring methods. For the Digital Humanities, this would be an important step in the direction of “scientific” humanities. With published data sets, the principle of being able to reproduce results is followed, but the data can also be used to produce findings for many other projects. For the second part of the task, creative and didactic methods are needed to identify such information in data sets. Ultimately, however, researchers who work in this way must also ask themselves whether the “Digital Humanities” should not rather call themselves “Data-Driven Humanities.” Think about it!

https://twitter.com/streka/status/1224712927300792320?s=20

References

Cao, Nan/Cui, Weiwei (2016): Introduction to Text Visualization, Paris: Atlantis Press, https://doi.org/10.2991/978-94-6239-186-4.

Casey, Jim (2019): Taste the Data! CDH Princeton, https://cdh.princeton.edu/updates/2019/05/15/taste-data/.

Grant, Robert (2019): Data Visualization. Charts, maps, and interactive graphics, London/New York: CRC Press.

Jansen, Wiebke (2016): Egguality, Data Cuisine, http://data-cuisine.net/data-dishes/egguality/. Karcher, Stefan (2017): Thesenpapier zu: „Visual Windows –Welchen Erkenntnisgewinn liefern Visualisierungen?“ Vortrag, Universität Heidelberg am 28.11.2017, https://heibox.uni-heidelberg.de/f/ae8bc0be67d348939c09/.

Karcher, Stefan (2018): “Visual Windows – The Added Value of Visualizations in Humanities,” INFODITEX-BLOG, https://infoditex.hypotheses.org/156.

Miller, James D. (2017): Big Data Visualization. Learn Effective Tools and Techniques to Separate Big Data into Manageable and Logical Components for Efficient Data Visualization, Birmingham: Packt Publishing.

Nussbaumer Knaflic, Cole (2017): Storytelling mit Daten. Die Grundlagen der effektiven Kommunikation und Visualisierung mit Daten, München: Verlag Franz Vahlen.

Cover image source: Pixabay, CC0.

---

---