Digital Humanities Research. How do digital methods challenge the paradigms of the humanities?

---

Paradigms describe the elementary principles of scientific disciplines. Such principles relate to, for example, the fundamental theories of a discipline, its research objects, its methods as well as its governing concepts. Paradigms describe the general outlook and practices of a discipline. These principles and regularities, though relatively stable, do change – they change historically, due to the development of new methods, technological advancement and innovation, or due to societal change. Paradigms shift, as Thomas S. Kuhn, a prominent researcher in the field of the philosophy and theory of science, put it and so does the way in which scientific disciplines generate knowledge.

The humanities, as a group of partly very different disciplines, share certain paradigms – especially concerning their research objects (human, cultural artifacts such as texts, artworks, for example) and the principles of their methodological approaches. In short, most humanities scholars read texts and try to understand them, they relate these documents to social contexts, they elaborate on intentions of authors and actors, they interpret the written word or the manufactured artifacts. These interpretations, the core of the hermeneutical method, are the way in which humanities scholars aim at understanding past and present societies. We want to understand communities, human interactions, societal institutions and organizations – historically, we want to understand cultural transformations.

One of the biggest, if not THE biggest cultural transformation that our society is facing right now, is the digitalization of almost all realms of our lives. Digitalization does not stop at the doorstep of scientific research in general or the humanities in particular. Digitalization changes all that we are interested in, it changes communities, human interactions, societal institutions and organizations. It changes the objects of our research through digital editions of (historical) texts and artifacts. In many areas digitalization poses challenges. And this is also true for our work in the humanities.

Reading and Understanding

If reading and understanding are two of our most fundamental practices, how do you teach that to a computer? The short answer is: you don’t. Computer tools do not read and understand in a way that humans do – or even in a way that we understand. It has become apparent in recent studies that machines do not »read«, they do not try to make sense from what they »see«. Maybe it would be better to talk about »detecting« instead of »reading« when we talk about computer programs, since what they do much better than humans is detecting patterns in large amounts of data. Still, we talk about making text material and data »machine readable«, transferring our own practice of information gathering into the digital realm and even into the machines themselves, all the while creating the challenge that we fear most through our own choice of words. Reading and understanding, making sense of the world around us does for now and should remain in the future a human task.

Then, why make use of digital methods and pattern recognition?

One of my main arguments would be that it offers us a chance of »productive irritation« by allowing us to interact with text in a different way than by reading.

Let me give you an example. Humanities scholars like myself tend to do a lot of reading when they want to answer a research question. And coming from medieval studies, my reading involves not only medieval writings as primary sources but also quite an amount of secondary literature, i.e. books about the medieval material in question written by other humanities scholars. In this way, I build up quite a bit of knowledge about a topic, learning about different interpretations of the material. And while this is an unquestionably important thing to do, the knowledge that I build up blocks my view of the original, metaphorically speaking, and hinders or predetermines my own interpretations.

Computer programs do not build up knowledge about something in the same way and that lays the ground for »productive irritation«. Using digital tools offers us the possibility to transform textual material into different states of matter. Text becomes a bag of »words«, relieved of all coherencies, their frequencies quantifiable, and their relations computable by statistical measures. Algorithms detect patterns, calculating frequencies and distances, for example, and rendering that information into a new proposition to generate meaning. Now all we have to do, is to read those new propositions, to understand diagrams and visualizations, to decode the patterns. Why do I call this »productive irritation«? Because in this way, digital methods help me to get rid of all context information and look at text as »raw data« again. Tables, lists, and diagrams allow me to see new patterns and coherencies and thereby irritating the knowledge I had built up before.

Visualizing and Interpreting

Looking at text as data produces new sources of information that require new modes of presentation. Patterns detected in a »bag of words« are usually depicted so that they can be »seen« rather than »read«.

Fig. 1: Word Cloud visualization on the basis of the edited volume »Die Corona-Gesellschaft«, ed. Michael Volkmer and Karin Werner, Bielefeld: transcript Verlag, 2020. The visualization was realized with Voyant Tools (https://voyant-tools.org/).

Word Clouds such as this one show the most frequent words in a text corpus represented by the size of the writing. The bigger the word, the more frequent it is. Obviously, »corona« and »2020« are the most frequent words in a book dealing with the early effects of the COVID-19 pandemic on the society. What this graphic also shows is that there are many words depicted in roughly the same size suggesting that their frequencies do not differ very much from each other. Does this mean that the vocabulary used in the different contributions to the edited volume in question is rather broad, that there are many different words used, than dense which would lead to less, but more frequent words?

Looking at a different pattern might help answering that question. Plotted are the positions of all 40 single articles of the edited volume in relation to the frequency in which they use the words »solidar(ity)« and »pandemic«.

Fig. 2: Mandala visualization on the basis of the edited volume »Die Corona-Gesellschaft«, ed. Michael Volkmer and Karin Werner, Bielefeld: transcript Verlag, 2020. The visualization was realized with Voyant Tools (https://voyant-tools.org/).

Not many articles seem to be talking about »solidar(ity)« – in fact most of them do not. Manipulating the picture further by adding »measures (Maßnahmen)« to the words of interest, changes the pattern and shows other similarities and differences between the contributions.

Fig. 3: Mandala visualization on the basis of the edited volume »Die Corona-Gesellschaft«, ed. Michael Volkmer and Karin Werner, Bielefeld: transcript Verlag, 2020. The visualization was realized with Voyant Tools (https://voyant-tools.org/).

Calculating distances between texts by looking at their respective vocabulary and its frequencies gives us a visual impression of their similarities and differences. The visualization of the results suggests a pattern that provokes an interpretation. And that interpretation relates to questions of a shared vocabulary which we might expect from contributions in a volume that should all relate to the same topic. But is it a shared vocabulary? How many texts represent single voices because there seems to be no connection? What does that tell us about the way in which the people represented here talk and think about the global pandemic?

Visualizations such as these draw attention to relations between words and texts that would not have been as apparent through reading. They provide new information as a basis for new questions and interpretations. But they do not make that interpretation. They do not make sense of the pattern they present. That remains a human task, a task for the humanities researcher accepting the challenges provided by digital methods.

---

See also the video interview with Silke Schwandt about the new series »Digital Humanities Research« ↗ on this BLOG.

---