Digital Humanities and Generative AI: yet another methodological turn?

Abstract

Generative Artificial Intelligence (AI), particularly Large Language Models (LLMs), is redefining the boundaries and methodologies of Digital Humanities. Its impact extends beyond practical applications, challenging the epistemological foundations of the field while opening new research frontiers.

Unlike the explicit formalization and modeling of theoretical concepts and analytical workflows typical of Digital Humanities, generative AI relies on inductive, probabilistic, and sub-symbolic methods. It captures and models implicit structures of meaning found within vast cultural datasets, representing them as multidimensional and continuous vector spaces. This paradigm appears more adept at addressing the complexity of cultural phenomena, which often resist explicit formalization. LLMs, in particular, partially implement interpretative processes that underlie the comprehension of cultural dynamics, decoupling these processes from the subjective human dimension traditionally tied to hermeneutic methodologies.

This epistemological shift not only opens up new avenues of research but also presents significant challenges. The Digital Humanities community is tasked with defining a new research agenda that accommodates these changes, setting out fundamental lines of inquiry. This talk aims to contribute to this agenda-setting effort by 1) exploring how LLMs and other generative AI systems can be theoretically conceptualized and understood; 2) identifying potential areas of investigation in computational and digital studies of cultural phenomena; 3) illustrating some case studies within the subfield of computational literary studies.

The recording of the lecture is available here: https://zenodo.org/uploads/15069064 and https://www.saw-leipzig.de/2025-ciotti