Shaping History, Responsibly: Seven Principles to Guide the Design of Archival AI Assistants for Cultural Heritage Collections

Major archives and libraries increasingly deploy archival AI assistants. Many of these tools not only offer users new ways to navigate but also to analyse collections, providing summaries of relevant documents and answering historical questions. Drawing on established understandings of historical practice and archival theory, this paper argues that these analytical affordances risk distorting and introducing further biases to our understanding of the past and offloading important interpretative decisions away from human experts. To combat these risks, we offer seven principles aimed at promoting the development of archival AI assistants that facilitate accessibility while enhancing research integrity. We demonstrate how these principles can be applied in practice through real-world examples, with particular focus on Barista, a chatbot designed to help users navigate digitised cultural heritage collections processed by the “Impresso - Media Monitoring of the Past II” project. Our strategy prioritises effective communication of archival AI assistants' capabilities; narrow, tailored approaches that speak to the specificities of particular corpora; and innovative design choices that encourage, rather than hinder, critical thinking and responsible research.