The mass digitization of historical sources and the exponential growth online of born-digital sources has catapulted the discipline of history from an “age of scarcity” to an “age of abundance”. Making sense of such “big data of the past” requires new approaches to data management, mining, visualization, and interpretation – an endeavour that poses multiple challenges to the disciplines of both history and data science. We propose to address this problem with a new Doctoral Training Unit (DTU): “Deep Data Science of Digital History”. Our DTU’s key objective is to train PhD students and engage with them in a critical study of historical data by bringing together intellectual and technical resources generated across disciplines, particularly from digital history, social sciences and data science. To achieve this objective, we propose to deepen the interdisciplinary collaboration between digital history and computer science by exploring the concepts of deep history and deep data science.
Based on the theoretical framework of “digital hermeneutics”, this DTU will tackle critical questions in historical data science by focusing on three thematic & methodological pillars:
1) deep data & knowledge; 2) deep analytics & learning; 3) deep visualization & interpretation.
• Deep data & knowledge addresses the challenges of creating digital datasets which, in the field of history, are generally characterized by their heterogeneity of data and their unstable or fluid nature in terms of volume and integrity. The axis will focus on analysis of characteristics, formats, histories, and infrastructures of historical data and train our PhD students in historical data criticism and traceable data management.
• Deep analytics & learning engages with state-of-the-art approaches in machine learning technologies and the use of artificial intelligence for analyzing large historical datasets. The aim of interdisciplinary training is to evaluate the heuristic potential of statistical modelling techniques, sensitivity analysis, and simulations for developing historical questions and interpretations and to confront computational methods with the rigor of quantitative methods at scale in historical sciences.
• Deep visualization & interpretation enters epistemological discussions about how visualization techniques and dynamic interfaces transform historical imagination and interpretation. Based on recent trends in explainable AI, information visualization, and human-computer interaction, the aim of this axis is to promote critical debates about how historical arguments can be turned into “graphic arguments”, and how new techniques of representing big historical datasets can be turned into explorative modes for the temporal and spatial sampling of historical information.
Spanning these three thematic pillars, two transversal structures will frame the Doctoral Training Unit as a coherent and shared interdisciplinary endeavour: the first builds on the conceptual work realized within the previous DTU “Digital History and Hermeneutics”, serving as a theoretical “Überbau” for organizing the collaborative work within the thematic units; the second transversal structure “Deep time & history” will serve as a common ground for discussing fundamental questions of how the collection, analysis, visualization, and interpretation of the big data of the past affects our understanding of time.
We designed the thematic pillars and transversal structures to function as “trading zones” – a concept originating from the sociology of knowledge describing the affordances and risks involved when trying to build fruitful bridges between disciplinary traditions and communities of practice. Inspired by a critical reflection of the metaphor of “deep time”, our DTU will problematize complex notions of multi-layered temporalities both in a “horizontal” (longue-durée) and “vertical” (superimposed temporal regimes).