Research

Research Statement

Emily L. Spratt

Dept. of Art and Archaeology, Princeton University

Spring 2019

I am trained as a Byzantine and Renaissance art historian with specialization in the early modern Mediterranean world. My doctoral thesis, "Byzantium Not Forgotten: Constructing the Artistic and Cultural Legacy of an Empire between East and West in the Early Modern Period," is a multi-field study of the response of religious art and architecture to different modes of rulership in the Venetian, Ottoman, and Slavic domains of the disenfranchised and collapsed Byzantine Empire. Utilizing icons, wall paintings, architecture, prints, and engravings, my dissertation narrates the cultural history of Orthodox Christian communities that defined themselves through tradition in a world of change. Although this material has largely been neglected in scholarship or examined in isolation due to the limited accessibility of objects from the period and the lesser-known languages in which they have been published (if at all), my research draws upon the extensive fieldwork I have conducted in the Eastern Mediterranean and the Balkans, a massive archive of unpublished photographs that I captured and assembled for my thesis, and extensive language study for my source materials.

Not only does my dissertation address a lacuna in our knowledge of the perpetuation of Byzantine visual culture in the early modern period, it advocates for the recognition of the field of Post-Byzantine art history, and the reevaluation of the traditional categorizations of medieval and Byzantine art. My theoretical premise for this position is espoused in "Toward a Definition of ‘Post-Byzantine’ Art: The Angleton Collection at the Princeton University Art Museum" and I am addressing the problem of the classification of style in objects that defy a singular formal interpretation in Beyond Hybridity: The Migration of Forms Across the Mediterranean, a volume of collected essays that I am editing.

Overall, my research is theoretically inspired by the post-colonial discourse and the Frankfurt School of criticism associated with the Institut für Sozialforschung, and is grounded in the theories and methodologies offered by religious studies, history, anthropology, philosophy, and a socio-historically-oriented art history. By utilizing these approaches and centering my research on both "high" and "low" art (a distinction that my scholarship problematizes) in different media, my project explores pre-modern identity politics, critiques the formation and promulgation of hegemonic structures in the medieval and early modern world, and offers the field of art history the writing of a subaltern culture, that of the former Byzantines, into the canon.

In recognition of the methodological limitations inherent in researching Byzantine and Post-Byzantine art, I also moved into the domain of computer vision science, where I work with computer scientists on the development and analysis of new tools of study for the history of art. It is my aim to lead a new type of cross-disciplinary research initiative, one that builds a cultural resource on Byzantine and Post-Byzantine art, which includes hard-to-source images of icons and their accompanying information, and to develop a program employing deep learning techniques for the analysis of this material that can be used in conjunction with more traditional methodologies. While icons are particularly amenable to style-based classification and therefore are an ideal subject to demonstrate the potential uses of vision technology in the arts, this project is intended to exemplify a new approach to research in the arts and sciences, which is not confined to singular fields of study, and offers a revised philosophy of approaches to visual culture in the 21st century.

Since icons are generally created in emulation of their prototypes, meaning that there are large groups of these works of art that look very similar to each other, and the published ones make up only a minority of the icons that have been produced and are cited in non-standardized ways, it is very difficult to assess this material with traditional art historical approaches. To date, only people with the most rarefied knowledge of icons and the privilege of direct experience with these hard-to-access objects have the ability to join the critical conversation about them, and there is often a lack of consensus amongst scholars on their attribution. Icons are also typically not signed, as they are usually created in an act of devotion, and are often repainted during their tenure in a sacred setting. Another complication in the identification of icons using traditional methods occurs when they have been photographed by scholars without clear indication of the conservation history of the object, which inadvertently produces a visual trail of images of the same icon that may appear as different works of art altogether.

Incorporating deep learning techniques into the visual analysis of this material will help resolve many of the above-stated issues, yet compiling the data for this type of research is no easy task. Scholars from the field of art history, managers of cultural institutions, and, in the case of Byzantine material, ecclesiastical hierarchs too, remain protective about their objects, and universal access to digital images of both published and unpublished materials, if the objects have even been digitally photographed, is a complicated and contested subject. Even the art market, which is generally open to sharing their records on the sales of icons, had non-standardized approaches in their catalogues from year to year, and not all object records have associated digital images. This research endeavor thus requires the cooperation of not only academics, but also church and state alike, and much cultural diplomacy with the custodians of private collections. The first step in realizing this project, which would help pave the way for others like it, is bringing awareness to art historians and the general public on how vision technology can be useful in the arts. To this end, I organized with my colleagues at The Frick Collection the blockbuster symposium "Searching Through Seeing: Optimizing Computer Vision Technology for the Arts."

One of the goals of my research is to demonstrate how developments in computer vision science can be fused with art history. In "Computational Beauty: Aesthetic Judgment at the Intersection of Art and Science," I have discussed how 18th-century philosophies on aesthetics still influence many present-day theories on visual perception in the humanities and require reevaluation given the advances that have occurred in all domains of neuroscience relating to the analysis of the visual world, particularly vision technology. In recognition that attitudes toward the relationship between the fields will change, my collaborator, Prof. Ahmed Elgammal, director of the Art and AI laboratory at Rutgers (where I am a member), and I surveyed computer scientists and art historians on this subject to create a benchmark for future sociological studies, the results of which appear in the paper "The Digital Humanities Unveiled: Perceptions Held by Art Historians and Computer Scientists about Computer Vision Technology" and are discussed further on our website, "The Digital Humanities Project: Aesthetics at the Intersection of Art and Science." Although this research has invoked the misleading term "digital humanities," which does not usually encompass the domain of AI, it is important to underscore that my focus is indeed predominantly focused on machine learning. To this end, I have drawn attention to the importance of the development and analysis of feature visualization techniques with deep learning for use in art history. For instance, in "Dream Formulations and Deep Neural Networks: Humanistic Themes in the Iconology of the Machine-Learned Image," I highlight the significance of the development of Google's DeepDream project for art history through its relation to formal analysis and iconography, in collaboration with Alexander Mordvintsev.

It is my conviction that vision technology will increasingly be used to augment human sight and that research in this direction will evolve in step with "Creative AI." I have had the opportunity to direct exploration of some of these concepts as the honorary guest editor for the Association of Computing Machinery's special magazine issue on computers and art. As society moves in the direction of expanding sensory capabilities, it is essential to consider the philosophical consequences of the technologies that empower these changes and to help guide their development in all neuroscience domains and in particular computer vision science. In this vein, my curatorship of "Unhuman: Art in the Age of AI," an exhibition on the results of a novel use of generative adversarial networks (GANs) to produce innovative styles of art, intentionally interrogates the definition of artistic authorship, the concept of creativity, and the meaning of art in this brave new world of machine learning. This exhibition mirrors the deeply humanistic inquiry that is embedded in the foundation of all of my research: How can we learn more about ourselves and our society, at present and historically, by exploring those uncharted territories surrounding the story of art, science, and culture?