Data-Driven Maps of Art History

Introduction

Especially after the inception of art history as an academic discipline in the late 18th century, diagrams have been used as visual devices to convey views on related developments. They were included in publications and textbooks and served as communicational means to support scholarly positions and as educational tools providing bird’s eye views on large-scale historical developments. Famous examples such as the “Diagram of Stylistic Evolution from 1890 until 1935” from (Barr, 1986/1937) sought to present visual aggregations of multitudes of individual processes as macroscopic, law-like views on artistic evolution. As discussed in (Schmidt-Burkhardt, 2005), such charts were often praised for their educational value but also challenged for lacking objectivity due to omissions or special foci introduced by their authors.

In recent years, the increasing availability of both digitized and born-digital art history resources went in parallel with the emergence of new paradigms of data-driven analysis which can be subsumed under the umbrella term “data analytics”. The emerging field of digital art history increasingly uses algorithmic methods to analyze the accumulating body of historical material, while recent trends in museology propose digital means of presenting art collections to an increasingly tech-savvy audience, seeking to contextualize exhibits using multimedia guides and interactive installations for local and dedicated online presentations for remote visitors. Information visualization plays an important role in both scenarios, supporting researchers in the interpretation of large data collections and the communication of their findings, but also providing virtual visitors with navigational guides for finding their way through vast online art collections.

The original aim of this work was to explore the potential of historical social network data, represented through interlinked artist biographies, for providing bird’s eye overviews on art historical developments beyond the relatively limited scope of existing, often manually created maps of art history. This was driven by the expectation that large-enough amounts of recorded ties between artists and/or other important persons from the art world would aggregate into large-scale networks spanning multiple centuries whose visualization could support the contextualization of artworks in virtual presentations. The analysis of existing biographical data sources, however, revealed that they were not free of cultural/institutional biases which appeared in form of differing compositions of nationalities and/or roles of covered persons. This provided an additional and relevant research perspective which directly linked to existing discussions about modes of inclusion and exclusion in the formation of the canon of art history, global aspects of which were often subsumed under the notion of the core and the periphery of the art world (Joyeux-Prunel, 2014).

The aims of the work presented in this article were therefore (1) to demonstrate the possibility of creating data-driven maps of art history and (2) to show their use for revealing specific biases present in different data collections. This paper unifies the authors’ previous related work (Goldfarb, Arends, Froschauer, & Merkl, 2013) (Goldfarb, Arends, Froschauer, Weingartner, & Merkl, 2014) under a common data processing framework and discusses large-scale network visualizations of art history biographies generated with content from the Getty Union Artist Names and Wikipedia, comparing them with each other and with existing scholarly examples. Different filtering approaches are used to highlight data specific aspects, including means to unravel chronological structure embedded in highly interlinked sets of historical entities and to reveal hidden interactions between subgroups of them. The results show that data-driven maps of art history successfully extend their manually created scholarly counterparts by putting them in larger historical context and at the same time serve as tool to reflect upon the nature of the used data sources themselves.