Stefan Heßbrüggen-Walter

In my previous post on visualizing philosophy, I have tried to characterize ‘visualizations for/as scholarship’ by two criteria: They must contain information that is relevant for experts in the field and try to provide insights that cannot be conveyed easily in a different medium. And they must allow for ‘inspection’: Implicit claims made in a visualization must be open for a closer examination of their corroboration.

In previous posts (here, here, and here), I have written on ‘nanopublications’ as a medium for what I call ‘collective doxography’. In this post, I will demonstrate how nanopublications can be visualized in order to gain new insights into the structure of a philosophical debate.

Three visualizations

The visualizations are related to a recent paper of mine (now published in Archiv für Begriffsgeschichte (only in print)) on the concept of philosophy in early modern Spanish Aristotelianism. After submitting the paper, I went through my excerpts and entered their content as a test case for ‘doxographical nanopublications’ into the EMTO-Nanopub database. This is my first attempt of visualizing this subset in a meaningful way (clicking on the static image should open a zoomable version: right-clicking into the background allows you to move (‘pan’) the graphics. The mousewheel is used for zooming in and out.)

Some explanations: Text fields denote authors in the debate and views they have asserted, denied or reflected upon. If a view is asserted, the arrow connecting it to an author is green. If a propositional content is denied, the arrow is red, and if the content serves as an object of reflection (i. e. if the author does not make a claim about it), the corresponding arrow is grey.

Clicking on a proposition or an author node opens the corresponding page on EMTO Nanopub. Clicking on an arrow tip opens the page that represents the ‘connecting fact’, i. e. the fact that an author has the epistemic attitude signified by the color of the respective arrow.

So what you get, is a visualization with ‘footnotes’. For every graphical element, a corresponding nanopublication exists that allows the viewer to investigate the correctness of the visualized claims, because the nanopublication gives a bibliographical reference that allows to locate the claim in the original text.

The question now is: what can you learn from viewing this visualization that could not be conveyed equally well in a text? Two aspects have surprised me:

• There is a marked difference in what we may call the ‘argumentative heat’ of authors. Compare Antonio Rubio (at the top of the visualization), the Complutenses (to the right of Rubio) and Gaspar Cardillo de Villalpando (at the bottom). Most of the arrows belonging to Rubio are grey: He is content to mostly paraphrase options that are available for defining philosophy without taking a stance on most questions. For the Complutenses, there are no grey arrows. They deny assertions made by others and no propositional content they accept is shared by others: all red arrows go to the center, green arrows are not connected to other authors. In Villalpando, there are only green arrows: he is not interested in contributing to a debate or weighing pros and cons, but wants to drive home the points he makes.

• differences in the relevance of propositions: some (e. g. in the center, “human thing is perceivable”) are mentioned, presumably because they belong to a tradition that must be reviewed. But they are neither asserted nor denied by anyone - apparently they must be referenced, because they are contained in an authoritative source, but they have no impact for defining philosophy. Others (e. g. “meditatio mortis is end of philosophy”) are more controversial.

A second visualization takes out all the grey arrows, reducing complexity. It only contains those authors who assert or deny at least one claim that is relevant for at least one other author (this excludes Rubio, because the claims he accepts or denies are not shared by the other participants in the debate):

Here, the structure of the debate becomes clearer. It can be seen that the debate is split into two main camps: the Conimbricenses, Saenz de Aguirre, and Diego Mas vs. Gaspar Cardillo de Villalpando and Toletus. Montanes has a ‘bridging function’, sharing views with both groups. The Complutenses are ‘outliers’: they argue against claims made by Villalpando and Toletus, but without sharing views with the Conimbricenses, Saenz de Aguirre, or Mas. We can reduce complexity even more and visualize only those propositional contents that are endorsed by at least two authors:

In this visualization, the Complutenses are no more part of the picture, because they only assert contents that are not asserted by anyone else. Now the grouping mentioned in the last graph becomes fully transparent. And one other insight may come as a surprise to some: There is not one thesis that all authors under examination can agree upon. In other words: instead of merely laying out agreed dogma, as some views of this period assume, early modern Scholastics (or Aristotelians or School philosophers - there are many soubriquets, none of which fits exactly) were connected by a shared style of reasoning rather than a common ‘world-view’.

How it is made

This is where most colleagues in philosophy will probably stop to read (some of them probably even earlier - the subject discussed is, as I readily concede, quite arcane). However, some degree of ‘data literacy’ is a requirement even for those who are content to ‘consume’ visualizations (instead of producing them themselves). In this case, it is important to understand that the graph was not layouted with some preconceived notions of relevance in mind. The structure of the graph is dictated by numbers (in this case the number of in- and outgoing connections of nodes). So the insights I have articulated are ‘data-driven’: in principle, they could have been gained by looking at a table that contains numerical values. This is important, because it helps to understand why a graphical representation of a debate can contain new information even for its creator.

The diagrams have been produced using Gephi. The data are generated using the export function of SMW (the software used for EMTO Nanopub). Currently, Gephi does not allow for an export of dynamic directed graphs. So I have gone with simple vector graphics as an export format. The exported SVG file is then ‘animated’ manually, using the excellent SVGPan library. Finally, the graphical elements (text boxes and arrow tips) must be linked to their correspondent EMTO Nanopub pages, using a Python script. This last step is technically uncomplicated. But in my view, it has enormous potential for the digital humanities, because it allows to build bridges between image and text. It should even be possible, although I have not yet tested this, to link into vector graphics, so that an excerpt of the graph could serve as a ‘graphical footnote’ for some textual comment or assertion.

And this need not be the end. Different principles of organisation for nanopub visualizations are conceivable: a timeline of ‘the most popular theses’ in a given period, a geospatial distribution of asserted contents on a map etc.

Conclusion

The strategy presented here depends on the essential presumption that something like ‘the doxographical attitude’ is really possible for philosophers. The methodological (or metaphilosophical) problem we face is whether disagreements in interpretation and biases are introduced in higher-order levels of historiographical work (e. g. weighing of passages, differences in emphasis regarding certain theses etc.) or whether such biases may go ‘all the way down’. Nevertheless, I think that introducing nanopublications as data that are based on ‘qualitative observation’ rather than quantitative analysis may come in useful for doing data-based work in the humanities in areas, where approaches based on ‘big data’ or purely quantitative analysis reach their limits. For the domain presented here, the history of early modern philosophy, this is certainly the case: research on natural language processing in Latin is woefully underdeveloped. Until then, we may have to rely on doxographical approaches (and a community willing to invest in them) in order to develop digital tools that function in this domain. Even on a small scale, such activities can provide interesting ‘microscopical’ insights into debates and arguments in the history not just of early modern philosophy.