Call for Predoctoral Fellowships (IMPRS-MDH) Application Deadline: May 31, 2026
\n
Interviews: June 2026
\n
Start: from August 2026
\n
About the IMPRS-MDH The International Max Planck Research School for Multimodal Digital Humanities (IMPRS-MDH ) is a new doctoral school jointly established by the Bibliotheca Hertziana – Max Planck Institute for Art History (BHMPI ) in Rome and the Faculty of Art and Social Sciences (PhF ) of the University of Zurich (UZH ), funded by the Max Planck Society (MPG ) for six years. Building on pioneering research at the MPG-UZH Digital Visual Studies center (DVS ), the IMPRS-MDH opens a new field at the crossroads of art and architectural history, linguistics, and multimodal artificial intelligence. It is basedat the University of Zurich.
\n
IMPRS-MDH predoctoral fellows work within a uniquely rich research environment and with access to special resources : on the one hand, the BHMPI's fully digitized photographic collection of 1.3M assets, 3M pages of scanned art historical literature, and the Max Planck Research Group Machine Visual Culture (MVC ), led by Prof. Dr.
\n
Leonardo
\n
Impett; and on the other hand, the UZH's Linguistic Research Infrastructure (LiRI ), CLARIN-CH resources of text corpora and computational linguistic technology, and the Digital Society Initiative (DSI ). The program is directed by Prof. Dr.
\n
Tristan
\n
Weddigen (BHMPI /UZH ) and Prof. Dr. Noah Bubenhofer (UZH ), with scientific coordination by Dr.
\n
Darío
\n
Negueruela del Castillo (DVS ).
\n
The Research Program The IMPRS-MDH is driven by a central question:
\n
How can the nuanced and culturally diverse patterns of thought and argumentation of the humanities be modelled using AI, and how might we use these models to study both cultural history and the cultural position of AI systems themselves, across languages and cultures?
\n
We seek doctoral candidates who bring genuine intellectual investment to one or more of the following research areas, which are not fixed tracks but overlapping problem spaces, we actively encourage proposals that cut across them:
\n
1. Epistemic Modeling of Multimodal Reasoning How can the interpretive methods of art historians, scholars of literature, and humanists at large be operationalized as multimodal AI models?
\n
This area moves beyond using AI as a retrieval or classification tool toward making humanistic argumentation itself transparent, testable, and computationally legible. Projects might examine how specific methods and epistemic approaches and frameworks can be encoded as multimodal pipelines, or how large, digitized collections such as the BHMPI's, or more specific ones like the Wölfflin edition (HWGW ) can serve as grounds for training and probing such models.
\n
2. Spatiality, Temporality, and World Models How can the diversity of humanistic knowledge be mobilized to develop AI systems that are genuinely adequate to the spatial and temporal complexity of cultural phenomena?
\n
Multimodal AI research has largely addressed text-image relations while leaving spatial and temporal dimensions theoretically and computationally underdeveloped. This area investigates how perspectival systems, architectural typologies, urban morphologies, and the multilingual discourses through which they have been named and transmitted might inform the design and evaluation of spatially and temporally aware models for humanistic inquiry. It further asks how temporal processes such as gradual transformation, feedback, and drift might be modelled across modalities like text, image, and space, in ways that go beyond frame-by-frame pattern detection.
\n
Fellows may draw on spatial theory, corpus linguistics, or media archaeology, and might work with 3D reconstruction, video models, or multimodal corpora. Projects might develop cross-modal methods for representing historical process and spatial meaning or use architectural and urban corpora as evaluation grounds for culturally situated spatial AI.
\n
3. Critical AI and the Culture of AI Systems What cultural assumptions are encoded in the latent spaces of large multimodal models, and how can humanistic frameworks reveal, critique, and contextualize them?
\n
This area brings the methods of art history/visual studies, linguistics, and the history of science to bear on AI systems as objects of inquiry themselves, and not only tools for analysis.
\n
Topics include: the visual, literary, and cultural genealogies of training datasets; circuit tracing and mechanistic interpretability as methods for humanistic AI critique; the cultural position of AI-generated images and text in contemporary culture; the history of computational thought in art history, linguistics, and the humanities; and the relationship between critical AI studies and Bildwissenschaft. Fel