Ressentiments Towards Digital Humanities And How to Deal With It
Junior researchers who analyse their sources with digital approaches can profit from productive environments where the advantages and potentials of these techniques are well known and established. In Germany, for example, this is the case if they are in the lucky position to work in one of the (fortunately increasing number of) digital humanities institutions in Berlin, Cologne, Darmstadt, Hamburg, Jena, Leipzig, Mainz, Munich, Potsdam, or Trier. I’m pretty sure the list is even longer, but if they don’t work nearby DH institutions, they are often faced with prejudices against their practices. In this post, I want to list five statements revealing biased attitudes, which I constantly hear, and propose answers to refute them.
With computational approaches, you try to depreciate traditional scholarship – but they can never replace it! They are only signs that you want to compensate your intellectual deficiencies in your own subject.
This fear is as old as Methuselah. New technologies will arrive and ignore all the findings and methods established over many decades. It’s inevitable that digital humanists will fail in their methodological limitation. But that’s not (or at least shouldn’t be) the aim of digital humanities. They don’t replace the established methods but complement them. The latest description, as far as I know, of integrating digital approaches in your research practice is provided by Andrew PIPER in his 2018 published monograph Enumerations (p. 10):
Based on your own reading knowledge, you have a certain presumption (belief), which you then test by doing distant reading (measurement). Now, you receive data, which you have to analyse by doing close reading (interpretation) again to verify or falsify your presumption. This is called a data-based approach.
|Once, I had the presumption that there is a relationship between Augustine’s attitude towards the intelligence of his female correspondence and his quoting behaviour. I thought that he quotes more in letters addressed to men or – if this is not the case – expresses quotations to women more explicitly, in order to make it easier for women to follow him. During his time, women were considered as intellectually weak. Then, I measured all the quotations Augustine uses in his letters and came to the result that he actually quotes more in letters addressed to women – explicitly and implicitly. I had to refute my presumption. But to interpret this result, I had to rely on my close reading experiences again. According to Kate WILKINSON (2015, p. 110), it was a sign of modesty for late antique women to speak in Bible verses. Since Augustine adopted the language of his addressees (see CONYBEARE, 2005, p. 58), he quoted the Bible many times. This had nothing to do with the intellect of his correspondents. I published this in 2016, if you want to know more details about it.|
Another possibility to connect distant and close reading processes would be data-driven. If you have a big corpus, you can do some information retrieval to search for keywords, topics, patterns, or whatever you want. The results can give you a hint where to start with the interpretation.
|Last summer, I created a keyword database for Augustine’s letters. In the results, I see that the term ratio is a keyword in letters 14, 120, and 162. Now, I can compare these letters in a close reading process and analyse them for Augustine’s usage of ratio, which might lead to new insights. Here, computational approaches function as GPS in a jungle of sources.|
“Distant Reading” and “Close Reading” are as prominent as problematic. I selected these terms to fall into line with the terminology used in PIPER’s figure. The problem is that Franco MORETTI’s approach, who coined the term “Distant Reading” in 2000, is neither “distant” nor “reading.” As PIPER’s figure has shown, you need some familiarity with your corpus to be able to interpret the results, so that there is no distance. And when the computer measures the text, there is no interpretive act of reading as well, but more a “systematic examination of data” (see JOCKERS, 2013, p. 25). Close reading could be confusing, because its original meaning in the New Criticism is looking at words as art piece neglecting any form of content (vgl. SCHMITZ, 2002, p. 105). Therefore, I would prefer computational approaches instead of distant reading, and deep reading instead of close reading.
However, two things are clarified by PIPER’s figure: Computational approaches do not replace traditional scholarship. It adds an important component. Furthermore, to have success as a digital humanist means to be absolutely familiar with your subject. So, the first ressentiment towards DH is certainly not true.
Learning to use computational approaches is a waste of time, which we should rather invest in a deep reading of our sources. After all, we are no trained IT-engineers.
This is often a claim of scholars, who are willing to learn Italian or French in order to cope with the secondary literature. But programming is nothing else than learning a language. In this case, it’s not Spanish but Python or R. Besides, it depends on your research question. It’s not always necessary to learn programming. Sometimes, free tools like AntConc, VoyantTools, or CATMA are sufficient enough for your plans. We provided a list with useful tools and tutorials on our website. However, you have to know what the computer is doing to avoid black boxes. Some statistical knowledge is also useful, because you should always be able to explain how the results were calculated. It’s like math: It’s not enough to know the results, you also have to know how you got it. There are also research questions where deep reading is a waste of time. This already motivated Franco MORETTI to use computational approaches. If there are too many sources, you won’t be able anymore to read them all. Your results become typological, as MORETTI calls it (2005, p. 76). When passing his judgement on deep reading, he said “Reading more is always a good thing, but not the solution” (2000, p. 55). However, we do have to admit one thing: We are no IT-engineers and, sometimes, it would really be a waste of time to do this on our own. Therefore, before starting a DH-project, it’s important to do a cost-benefit-analysis. If it gets too complicated, you should not give up, but start collaborating with IT engineers. Digital humanities belongs to a highly interdisciplinary field and a collaboration with computer scientists can be very fruitful for both sides!
Digital humanities are only able to find correlations without any meaning. The results are useless, because you never know if they were intended by the author or if they just happened by accident. Its interpretation is always highly speculative.
Yes, but no. Very often, arguments are weak if they stand alone. But correlations revealed by computational approaches give us a hint to find more arguments based on further reading experiences. Significant patterns in texts can provide strong support for other observations that we have made in a text, which would be also weak if they had no support by other arguments. In other words: Computational approaches enrich our sources with new data, which we can analyse to gain more information. All together, these newly-gained information is not speculative but conclusive.
In our field, it’s not necessary to deal with methods of digital humanities, since we have this fancy <choose a popular concordance of your discipline>.
Digital dictionaries and concordances are a vast improvement for researchers of all disciplines. Latinists use the Library of Latin Texts or the Thesaurus Linguae Latinae, Bible exegetes have BibleWorks or Accordance, and so on. These databases provide a huge range of functions to analyse the sources. Many scholars think that this is enough for their needs. The problem is that research is limited in this way, because you are only able to do what the interface allows. With BibleWorks or Accordance, for example, you can only work intratextual. The Library of Latin texts allows you to do lemmatized intertextual word searches, but there is no full text option. You wouldn’t be able to do a named entity recognition for network analysis or topic modelling with it. So, if you think you don’t need further computational approaches other than those provided by a certain (and often premium!) software, you will put your research creativity in irons.
Your sources are not big enough to justify computational approaches.
You could also hear this from computer scientists, which is also supported by Martin MUELLER’s term “Scalable Reading.” According to him, you use distant reading when analysing large digital text corpora, but close reading when you look on a single text or passage. The amount of texts decides the method. This becomes even more evident when considering JOCKERS’ term Macroanalysis (2013), which simply blanks out analyses on small units of texts. However, there can be also a reason to use computational approaches on small sources. One example is MORETTI’s Network Theory, Plot Analysis (2011), where his only source for computational approaches is Shakespeare’s Hamlet. Maybe our texts aren’t big data (and compared to business data it’s not that easy to find big data in humanities’ sources). But to conclude with Thomas WEITIN (2015, p. 5), with computational approaches, we are able to transform our unstructured small data in smart data, which provide us new information. Therefore, the size of the sources is not decisive when you have to decide for digital humanities or not.
Conybeare, Catherine: Spaces Between Letters. Augustine’s Correspondence with Women, in: Voices in Dialogue. Reading Women in the Middle Ages, ed. Linda Olson and Kathryn Kerby-Fulton (2005), 57-72
Jockers, Matthew: Macroanalysis. Digital Methods and Literary History, Topics in the Digital Humanities, Urbana, Illinois 2013
Moretti, Franco: Conjectures on World Literature, in: New Left Review 1 (2000), 54-68
Moretti, Franco: Graphs, Maps, Trees. Abstract Models for a Literary History, London/New York 2005
Mueller, Martin: Scalable Reading (Online-Ressource: https://scalablereading.northwestern.edu/?page_id=22), 2012
Nunn, Christopher: References in the Correspondence of Augustine. Chances and Boundaries of digital “Distant Reading” Processes, in: Revue des études augustiniennes 62,2 (2016), 223-233
Piper, Andrew: Enumerations. Data and Literary Study, Chicago 2018
Schmitz, Thomas: Moderne Literaturtheorie und antike Texte. Eine Einführung, Darmstadt 2002
Weitin, Thomas: Thinking slowly. Literatur lesen unter dem Eindruck von Big Data, LitLingLab Pamphlet 1, Konstanz 2015
Wilkinson, Kate: Women and Modesty in Late Antiquity, Cambridge 2015