From Shoebox to Toolbox: Archival Affirmation in the AI-Assisted Processing of a Media-Art Archive
Year: 2026
Type: preprint
Abstract: Large digitised media collections often remain unprocessed and unexplored, despite being technically preserved, and archival institutions increasingly turn to artificial intelligence (AI). Rather than pursuing autonomous archival systems, this article examines how AI can support archival description and classification while preserving the contextual, historical and provenance-related integrity of archival materials. Drawing on the archive of media artist Teoman Madra (1931–2025), the study develops and evaluates an AI-assisted archival tool for a large, initially unstructured collection. The proposed approach establishes an iterative, human-in-the-loop workflow in which AI-generated suggestions are critically assessed by the archivist, whose decisions subsequently inform the system's outputs. The study shows that the controlled vocabulary given to the model shapes what it can recognise, that the archivist's validated decisions guide classification reliably within an archival unit but not across units, and that the model's contribution declines as tasks demand contextual knowledge beyond what is visible. While this process can improve efficiency and reveal details a person might miss, it also introduces epistemic risks, particularly when plausible but inaccurate classifications remain undetected. Faster processing does not by itself mean more reliable description. The article conceptualises AI-assisted archival classification as archival affirmation, a collaborative process in which human judgement remains integral to the production and validation of archival knowledge. By situating the archivist's interpretive authority at the centre of this process, the article contributes to discussions of the responsible integration, critical oversight and governance of AI in archival practice.
Source: Research Square
Author: Selcuk Artut
Institution: Sabancı Üniversitesi
Language: English
Subfield: Conservation
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236606422
