Retrieving Context, Re-centring Interpretation: AI Hermeneutics and the Democratisation of Reading

In digital humanities, retrieval technologies have always mediated how we read, from concordances to search engines. With the advent of large language model and retrieval-augmented generation (RAG) which was made available by the former, interpretation itself becomes procedural: machines now arguably perform a version of the critic’s movement between evidence and hypothesis, inviting debates and questions on the way (literary) reading might have been changed or even challenged. This talk introduces AI hermeneutics, an approach that treats RAG not as automation but as interpretive infrastructure, opening up an opportunity to integrate AI into the existing reading experience.

Drawing on experiments with the Conflict Archive on the Internet and conflict-era poetry, the  project explores how retrieval systems can surface overlooked cultural memory while foregrounding the ethical boundaries of machine reading. It demonstrates how model “(mis-)readings” reveal the biases and expectations embedded in both human and computational interpretation.

This talk also considers questions of accessibility and sustainability: how lighter, discipline-specific models and transparent documentation can democratise advanced AI methods for scholars without extensive computing resources and its challenges. By reframing retrieval as a human-machine dialogue, AI hermeneutics invites the digital humanities community to see algorithmic mediation not as a threat to interpretation but as an opportunity to make the labour of reading visible and rewarded again.

This webinar was delivered as part of the DARIAH Friday Frontiers Autumn series 2025. The video was edited by Vicky Garnett, and the subtitles were provided by Áine Foley.

Learning Outcomes

After watching this webinar, learning will be able to:

• recognise issues of bias embedded in both human and machine interpretations of texts.
• critique the accessibility and sustainability of advanced AI models, and the opportunities that they offer in textual analysis.