ESHS/HSS Annual Meeting

Re-envisioning Scholarly Editions in the AI Age: Methodological Challenges and Opportunities

• In Event: The Impact of the Digital Revolution on the History of Science and its Social Implications-3

English Abstract

Long-standing scholarly editions, often the product of decades or even centuries of meticulous work, represent invaluable corpora of curated textual and intellectual knowledge. The challenge now facing these projects is their probably inevitable paradigmatic transformation into the age of Artificial Intelligence. One or two of these traditional editions may serve as case studies for this transition from a static print collection to a dynamic digital laboratory for the history (of science). We offer some already tested solutions and propose ways to explore, discuss the obstacles and put the process into a perspective of technological development and a meta-history of science.
The foundation of this transformation was the migration to a TEI-XML framework, which enables the application of computational methods. Initial opportunities lie in leveraging AI for the automatic transcription of complex early modern manuscripts using emerging Multimodal Models (MMLs) and for developing bespoke Named Entity Recognition (NER) pipelines to identify key actors and locations, or using the latest DH technologies to reconstruct lost text. However, this integration presents profound methodological challenges, chief among them establishing rigorous standards for the citation, validation, and reusability of AI-generated data in concert with traditional manual scholarship.
Beyond these foundational tasks, the implementation of granular TEI encoding schemes—for instance, systematically tagging all mentions of illness or scientific instruments —can unlock unprecedented thematic research into the history of medicine or the material culture of science, hardly present in earlier editorial work. This structured data can then be organized within a comprehensive thesaurus, linking concepts, experiments, and formulas within the corpus and, crucially, to external datasets, fostering new interoperability in historical research. Furthermore, the creative application of Large Language Models (LLMs) opens avenues for visualizing historical thought experiments or even generating video reconstructions, transforming abstract textual descriptions into accessible, dynamic representations.
By confronting the challenges – epistemological, ethical, technical, dealing with transparency (black boxes), issues of long-term preservation, and the like – the evolution of editions will do more than simply illuminate the scientific past; it will actively shape the future of historical inquiry, forging a new synthesis between scholarly expertise and AI-driven analysis, leading into a faster paced future with ever more automated procedures.

Author

• Martin Kurz, University of Zurich