Rethinking AI for the Humanities

Matteo Valleriani: Current debates often portray AI as a threat to the humanities, raising fears about authorship, interpretation, and research practices. Our workshop takes a different path: We focus squarely on the technical details and epistemic implications of specific AI-based methods. Our aim is to critically explore what AI actually does—and can do—when applied to humanities questions.

A distinctive feature is our emphasis on “time” and “temporality,” core topics in the humanities but difficult for AI to handle. By bringing together top-level AI experts and humanities scholars, we aim to explore not only how AI can support humanistic inquiry, but also how humanistic concepts can expand the horizons of AI research.

Oliver Eberle: From the AI perspective, the humanities pose challenges that today’s models cannot yet solve: limited and fragmented data, strong dependence on context, and the need for nuanced interpretation. While foundation models can help to organize large bodies of material, the features that define humanistic inquiry, such as heterogeneity, reliance on contextual understanding, and complex reasoning, often exceed what today’s learning-based systems can handle. These difficulties are amplified by the “low-resource” nature of many humanities sources, which provide limited data and often lack the clear task labels that AI systems require. Addressing these limitations means asking how to evaluate outputs with historical, cultural, and ethical nuance, and how to build models and interpretability methods that engage the full depth of humanities research.

Matteo Valleriani: The workshop is designed as a platform for deep interdisciplinary engagement between AI researchers and scholars in the computational humanities. Structured into four thematic sessions—Explainable AI, Historical Networks, Low-Resource Data, and Foundation Models—it brings together leading voices in both communities to share insights, discuss challenges, and push methodological boundaries. With approximately 30 additional experts in attendance beyond the speakers, we anticipate that the event will create fertile ground for synergies among existing projects, seed ideas for new collaborations, and accelerate the exchange of tools, methods, and research questions. Beyond the keynotes and panels, the real value lies in conversations, poster sessions, and informal encounters that can spark fresh collaborations.

This workshop would not be possible without the joint support of BIFOLD and the Max Planck Institute for the History of Science. Both institutions have fostered collaboration between AI research and the humanities. This workshop stems directly from that vision and serves as a model for how foundational AI research can remain connected to the broader intellectual and societal questions that drive the humanities.

To open the conversation beyond academia, the workshop is framed by two events open to academia and the public. Christoph Sander will start on September 22, 2025 by presenting his research on “EFFIZIENT ABER KORRUPT? Verwaltung im 17. und 21. Jahrhundert, untersucht mit KI” followed by a comment by Jan Mendling (HU Berlin) and a discussion. Please note the format is in German. Further information will follow shortly. If you would like to be notified once the full program is available, please send an email to pr@bifold.berlin and we will keep you updated. A scientific colloquium with Ingo Scholtes will follow on September 25, 2025 [More details].