From text to image: Visual and multimodal AI in the Digital Humanities

As historians increasingly engage with Artificial Intelligence, the challenges and opportunities of working with images in digital humanities research are coming into sharper focus. While text-based AI applications are well established, images present unique challenges: they are not inherently machine-readable, posing significant challenges for large-scale analysis. With recent advances in Artificial Intelligence, historians now have powerful tools for detecting, classifying, and comparing visual material, offering new ways to study artistic traditions, print networks, and iconographic patterns. This lecture explores the role of AI in historical image analysis, from computer vision techniques that identify illustrations in early modern books to machine learning models that group repeated images and track their circulation. It also examines how AI-generated metadata and visual search tools are transforming iconographic research, allowing historians to trace the reuse of imagery across religious, political, and artistic contexts. While these advances open new possibilities, they also raise critical questions about interpretation, technical limitations, and the role of human expertise in computational approaches. By reflecting on both the successes and challenges of using AI in digital humanities, this lecture provides historians with a practical framework for engaging with visual data in an AI-driven research landscape.

DREW THOMAS | University College Dublin

Seminar cycle: “AI and history“

Organization:
Andrea Pojer (Università di Trento – FBK-ISIG)
Massimo Rospocher (FBK-ISIG)
Sandra Toffolo (FBK-ISIG)

The event will be held in English.

The presentation will take place in presence in FBK’s Aula Piccola, subject to availability, and online.

Registration is mandatory by 28 April 2025 at noon.

***
Image: AdobeStock_1037404632

***
Ai sensi e ai fini del Regolamento UE n. 2016/679 - Regolamento generale sulla protezione dei dati (GDPR) e come descritto nell'Informativa sulla privacy per i partecipanti all'evento di FBK, si informa che l'evento potrebbe essere registrato e divulgato sui canali istituzionali della Fondazione. Per non essere ripresi o registrati, si potrà sempre disattivare la webcam e/o silenziare il microfono durante eventi virtuali oppure informare anticipatamente lo staff FBK che organizza l'evento.