View all Blog posts

Posted by: Hilde de Weerdt 7 years, 8 months ago

(0 comments)

Digital Interpretations

This post is based on an apology for my ecclectic use of digital research methods in the final part of a forthcoming monograph (Hilde De Weerdt, Information, Territory, and Networks: The Crisis and Maintenance of Empire in Song China. Cambridge, MA: Harvard University Asia Center). I first review the historical roots of historians' fears about the digital and proceed with an explanation of some potential and real benefits of digital methods for philological and historical inquiry.

Hilde De Weerdt.
Originally written in 2012, slightly revised in 2014.

"The Bitch-Goddess QUANTIFICATION"[1] 

            Debates about the value of quantification in the humanities in general and in history more specifically have returned to conferences and workshops in a manner reminiscent of the discussions of the uses and abuses of quantification in the 1950s and 60s. An engagement with methods that have in recent years become associated with 'the digital humanities' requires further clarification in an environment where the linkage between the humanities and the digital incites scathing critiques by literary critics or tacit rejection and dismissal among historians. Some frame their critiques against digitization in general terms, objecting to its supposed nefarious impact on linear and intensive reading, critical hermeneutics, the engagement of humanities scholars with questions of truth and falsehood, and their ability to distinguish between "the important and the trivial."[2] The role of quantification in particular has been the critical point in the cold reception of the self-proclaimed digital humanists' work amongst those who have come to accept consciously or unconsciously the effect of digitization on research routines such as searching and accessing primary sources or scholarly communication.

            Digital historians are especially affected by historians' (selective) memory of mid-twentieth-century quantitative history which was primarily concerned with topics of interest to post-WW II social scientists such as social mobility, urbanization, or party affiliation. Even though the earliest application of computation in the humanities was not in this field (digital humanities practitioners generally point to the building of a concordance, a tool that is more commonly associated with conventional philological scholarship, as the first humanities computing project),[3] quantitative historians became avid and prolific users of the computing power offered by mainframe computers. These machines allowed them to conduct statistical analyses of vast amounts of "data." One critic of the most controversial works in cliometrics recalled his reading experience of Time on the Cross: The Economics of American Negro Slavery (1974) in these terms: "the rattle of electronic equipment is heard off stage, and the reader is coerced by references to vast research effort involving thousands of man and computer hours' and inconceivable mountains of statistical data."[4] The unpleasantness of the experience extended well beyond the reference to industrial noise pollution as the authors of Time on the Cross concluded that the prolific evidence showed that the slave economy in the American South had been operating "within modern rates of expropriation"; through a variety of other numerical indicators such as the average time lag between whippings, they painted a rather benign picture of slave labour. It wasn't until the 1980s and 1990s, with the popular adoption of the personal computer and the introduction of the Internet and the World Wide Web, that historians could begin to challenge the overwhelmingly negative connotations associated with the relationship between computers and historical scholarship. By the late 1990s the methodological and interpretive shortcomings of Time on the Cross could be addressed through a range of digital projects rendering accessible and interlinking not only the census data privileged by quantitative historians but also transcripts of slave narratives, images, a sound recording of an interview with a former slave, pre-existing secondary scholarship, as well as new digital narratives characterized by non-linear narrative strands and hyperlinking.[5]   

            Nevertheless, as one practitioner, historian, and critic of digital history writes, wistfully, the disparaging sentiments towards projects "using the very latest in computer technology" "are still widely held in the discipline."[6] For many the appearance of a numerical table, a graph, a map, or a chart in secondary historical scholarship triggers an automatic response, one that associates it with positivist attitudes attributed to the natural and social sciences. The use of visualizations, especially of those types that are also used in the sciences, is thus readily seen as a sign that their author has bought into a correspondence theory of truth. Visualizations are interpreted as a sure sign that their creator believes that they are representations of phenomena that are objective or observer-independent; any reference to "data" is then linked to a denial of their constructedness and of the inherent subjectivity of interpretation in the humanities. The identification of "the whole paraphernalia of graphs and tables" with "dehumanizing methods of social sciences" already emerges in the earliest critiques of quantitative social history such as Carl Bridenbaugh's 1962 American Historical Association Presidential Address, even though Bridenbaugh admitted that, when used in combination with a concern for " men acting in time and place," quantification (in lower-case as opposed to "the Bitch-goddess, QUANTIFICATION," invoked as the new object of worship amongst historians in the same address) still had some uses.[7]

Cultural history and digital methods

            Fifty years later "maps, graphs, and tables" are still considered as stand-ins for positivist social science methods. This time Chinese historians have also become involved in the debate, this in large part due to the creation and adaptation of large-scale databases and datasets such as China Biographical Database [hereafter CBDB] and China Historical GIS. In a recent critical review of Song scholarship by Christian de Pee, for example, they are associated exclusively with the work of an earlier generation of social historians and read as vehicles through which "trends and patterns suggested by sociologists and political scientists" are shown to apply to the history of imperial China.[8] To some extent such an exclusive association of visualizations in humanities research with positivist social scientific theory and methodology makes sense: many are produced with out-of-the-box software designed for social or natural scientists and employ standard metrics.[9] However, the maps and tables in the work of the new generation of cultural historians whom de Pee credits with the rehabilitation of nineteenth-century philology[10] suggest that the critical hermeneutics of cultural historians is not incompatible with numerical and graphical representations and that the latter need not solely be "the province of social history."[11] The works of some of the most influential cultural historians like Carlo Ginzburg or Roger Chartier similarly combine close readings of texts with creative interpretations of previously compiled numerical and statistical datasets.[12]     

            Despite an emerging consensus that the history of the digital humanities (which we could define as the use of computer technology in research on humanities subjects and the critical self-reflection on how digital methods shape research) can be subdivided in "waves" or phases, with the first wave of "humanities and computing" generally being perceived as "quantitative" and the second wave as "qualitative,"[13] its early practitioners have been keenly aware of both the perils and the opportunities digital methods presented to cultural and intellectual historians. Willard McCarty, who in 1987 set up Humanist--the first listserv in the field, recalls how his attempt to model Ovid's Metamorphoses above all highlighted "the mediation of thought that [the agent-scholar's] use of the computer implies." By reflecting on what is lost when poetic phenomena like personification are translated into systematic encoding McCarty came to the conclusion that this approach opened up the possibility of multiple grammars of personification rather than the one he set out to construct. And so, "the via positiva out of which comes a useful tool is complemented by a via negativa, the illuminating path of failure by which insight comes."[14] McCarty thus articulated in this early experiment a first benefit of digital encoding: encoding raises awareness of the mediological effects of computational designs.

            The practice of encoding not only raises one's awareness of the mediological effects of computational designs, but also, as highlighted in more recent work of literary and cultural critics interested in old as well as new media, enables the researcher to interpret texts and reader responses to them flexibly.[15] A first aspect of this flexibility is the ability to reconfigure texts. Linking passages and narrative strands is a fundamental aspect of interpretation. In my research on notebooks and letters I turned to encoding, first as a way to keep track of my reading through fragmented primary sources. Encoding thus becomes as in McCarty's experiment a way of taking notes. Encoding allows one to recall text and annotations as if they had been compiled through an elaborate card-index system without the need for bulky filing systems. The reconfiguration of text and reader notes by means of the encoding criteria selected by the reader offers a different approach to reading and interpretation from the more ubiquitous and far less self-reflective use of natural language searching. Natural language searching allows for the selection of passages on the basis of a literal match between terms of interest to a contemporary reader (in search interfaces usually one or a very limited list) and a fixed set of digital resources. The latter tools have allowed historians to identify and gather together passages on the basis of particular characters or phrases, but the use of smaller lexical units in scanning the text has led to a fragmentation of the text and the loss of an overview of the subjects and arguments covered in the notebook in the aggregate. In other words, the original content map has been forgotten and the present-day reader in most cases is not replacing that map with a topic map that may help capture the content of the entire notebook for present-day historical analyses. In the case of Chinese notebooks, for example, present-day readers relying on natural language searching approach them as collections of historical materials from which to pick whatever passages come up. I began to appreciate encoding as a substantive part of a new approach to the close reading and interpretation of sources when it became clear that, both for the reader tagging content and for a later reader accessing the text through such tags, narrative strands within the text can be pursued without the need for insisting on the use of particular terms. The interpretive benefits of digital analysis apply to both text and graphic representations. For example, a digital reconstruction of reading instructions for a twelfth-century map can help recover how an editor or prefectural school instructor in the twelfth century intended the map to be read and helps the modern reader question the reading conventions with which he/she approaches a map.[16]   

            A second aspect of the flexibility of digital encoding is the ability to connect texts with other texts, including digital reference sources. By encoding authors, interlocutors, titles, or quotations in a notebook, for example, we can open the individual text up to the social and textual world within which it was conceived. We can link to the multiple reference sources embedded in CBDB or other databases and can thus learn about the family history of or the places visited by the acquaintances notebook authors referred to. Or, we can link to descriptions of or other comments on a particular title, author or place in other encoded texts, both contemporary and later, regardless of how that author, text or place was referred to in the original text--through encoding readers normalize various ways of referring to the same thing and thus all instances can be accessed regardless of the particular characters used. Encoding can thus help recover dimensions of the social and cultural history of texts that present-day readers removed from the social, political, and textual contexts within which these texts were produced will always struggle with. Creating a network of all authors and conversation partners and analyzing their family and career histories provides an understanding of the social and cultural world within which the author wished to portray himself. Such network diagrams are of course approximations of the relationships and memories that inhabited the texts. Such approximations, nevertheless, bring in context that was in many cases known and relevant to the author. More generally, they invite us to read texts not as isolated products but as embodiments of personal and collective practices of writing and publishing. This is not to say that consulting a range of traditional reference sources might not yield similar results. Encoding and the creative visualization of the annotations encoding produces are, however, rendering access to relevant reference sources and contemporary sources much more easily and systematically available. They are integrating reference work into the reading experience. They are updating and improving traditional philological tools and making it possible for larger numbers of scholars to be involved in such work.[17] 

            Besides the reconfiguring of text and notes and the integration of other sources of information, there is another aspect of the flexibility of encoding, one that has proven most controversial. Encoding practices allow the present-day humanities scholar to interpret texts in the context of larger sets of texts through quantitative indicators. Numbers and frequencies are relevant to much historical understanding, although they are never transparent or univalent. How many relevant texts were written and printed over time, how often a particular author or historical actor appears in particular texts, in particular genres, or with whom or what places and institutions that author or actor is most closely associated are the sorts of questions that can, within the limitations of the existing record, be tabulated and visualized in a variety of ways in digital projects. Many fear that this sort of quantification diminishes the value of historical and humanistic inquiry. If this were to become the only thing that historians would now set their hearts on, they would be right. The vision underlying the provision of this level of contextualization is, however, usually not to render all history quantitative but rather the improvement of the contextual knowledge that should go into the selection of primary sources and their interpretation as well as into the construction of questions based on comprehensive rather than predominantly anecdotal evidence.

            Let me illustrate some of the ways in which quantitative indicators can enhance cultural historical inquiry on the basis of my work on Wang Mingqing's twelfth-century series of notebooks Waving the Duster. First, before selecting this series of notebooks for further investigation, I created a database of all notebooks that had been printed during the Song Dynasty (960-1276). This dataset suggested, among other things, that with its 6 or 7 attested printed editions between the 1180s and 1270s this was one of a small number of notebooks that circulated in several printed editions. It also showed that its printing by private, government as well as commercial publishers was not an uncommon phenomenon. Second, by looking at the temporal distributions of all authors and interlocutors mentioned by Wang Mingqing as well as the frequency with which they were mentioned we learn that Wang Mingqing (and it turns out from later tests many notebooks authors from the twelfth century onwards) recalled a wide and diverse range of relationships of mostly contemporaries and increasingly men of lower literati status. This raises questions about the presumed centrality of standardized texts through examinations and the impact of (Neo-)Confucianism. Third, through a content analysis based on quantitative indicators (alongside qualitative annotations of each entry in the notebook) I am also able to revisit cultural historical questions other than those relating to the social and cultural history of notebook writing, questions such as how Song Chinese literati perceived non-Han peoples and states. From selected entries the reader searching for the equivalents of "Jurchens" or "non-Chinese peoples" could conclude, alongside modern historians, that Wang largely adopts the terminology of equality used in diplomatic exchanges. Or s/he could conclude the opposite that Wang recalled passages showing that the term non-Chinese was a derogatory and degrading term. When reading through all individual entries discussing Jurchens (a people originally living to the northeast of Song China whose military invaded the Song Empire an eventually occupied about half of its former territories) in Wang Mingqing's notebook, the present-day reader should be struck by the range of terminology used to refer the Jurchens. The reader may interpret this at first reading as a sign of the multivocality of notebooks that draw in differing genres of contemporary texts. In an appropriately encoded text we can go further. We can read the terminology used in the entry at hand in the context of its frequency and statistical relevance across the entire notebook and, based on our encoding of the context and situation within which the term was used, interpret it along a dimension that is otherwise lost to historians and literary critics. It has been lost to most historians because they have tried to fix the meaning and connotations of ethnonyms divorced from their rhetorical contexts. Tables of frequencies are therefore relevant to the discussion of the representation of the other. They do not, however, speak for themselves nor are they generated without human interpretation. Pace Stanley Fish who recently reduced the variety of methods of digital humanists to one in which "first you run the numbers, and then you see if they prompt an interpretive hypothesis. The method, if it can be called that, is dictated by the capability of the tool."[18] Numbers are not given or pre-existing to the interpreter's analysis.

            The above examples further suggest that digital readings allow for a multiperspectival exploration of texts. By linking into external databases and other texts, we can explore texts, people, events and opinions from multiple perspectives. For the twenty-first-century historian to examine what it meant to have one's notebook circulate in print, tabulations of all editions known to have been printed, however imperfect, provide a level of contextualization unavailable in the text or its prefaces. Such contextualization is relevant to a cultural historical interpretation even if the data would not have been known to those whose work we read. On the basis of a systematic exploration of the biographical information, social, geographical and political backgrounds of the persons the author recalls in his published notes we can explore the ways in which his interests and the positions he takes took shape within concrete historical relationships and debates. 

            I have focused mostly on what encoding can enable historians to do and less on the pros and cons of ways of exploring, tabulating, and visualizing data. As articulated most emphatically and clearly in Johanna Drucker's recent critiques and as illustrated in Franco Moretti's work,[19] humanities scholars would benefit from designing visualizations that are more closely aligned with their methods and goals. Rather than a table with numerical values for the frequency with which different topics are discussed in particular notebooks, I found a topic cluster diagram to be a better way to visualize the main concerns and interests as I had identified them. The diagram uses size as an indicator of which were the more frequently occurring concerns but it also organizes these on the basis of the cultural historian's understanding of how these interests can be best interpreted and explained. Moreover, the visualizations are contextualized and interpreted in the main text. The modern reader, who might attribute objectivity to the map or the table, is advised to read and evaluate them as arguments alongside the textual arguments. As it turned out absences of concerns that we have come to associate with the period may be very significant--such topic maps also serve the practical purpose of guiding researchers to the notebooks and the passages most relevant to their topics of interest. 

            Drucker further proposes that digital humanities scholars use "capta" instead of "data" to refer to the material represented in visualizations, this in order to prevent our colleagues in the humanities from automatically assigning us to the camp of the positivists. Even though the new term captures the intervention of the interpreter more effectively than the old,  I have not attempted to systematically shift to the new vocabulary here; as Drucker admits, natural and social scientists are generally attuned to the idea that data are constructed, implying that "data" also means "capta." I'd rather continue to insist on the important point that data are interpretive rather than conceding that some historians have data and others capta. The materials represented in tables, maps, and charts should not be read in isolation from the historical narrative within which they are embedded. As Tim Hitchcock's successive self-reflections on his large social historical digital projects underscore, the data and visualizations we create only become history when they are crafted and read as "purposeful history."[20]

Concluding remarks

Share on Twitter

Share on Facebook