Historicising uncertainty in photo archives for semantic modelling purposes

Abstract

Cultural Heritage (CH) data are inherently ambiguous: attributions shift, authorship is debated, and titles or dates are often uncertain or contested. Cataloguing conventions vary across institutions and time, leading to inconsistencies that affect interoperability. Photo archives amplify these challenges, as early cataloguing practices developed in the absence of shared standards and often treated photographs as secondary to artworks. Linked Open Data (LOD) technologies have opened opportunities to restructure and reconcile such records. However, most models and workflows assume certainty and one-to-one matches between entities, producing overly “clean” interfaces that risk erasing interpretive diversity. Fully reassessing records to resolve uncertainty is rarely feasible, and archivists increasingly face the need to historicise metadata and embrace ambiguity as part of the record itself. This article argues that modelling uncertainty should be treated as a first-class requirement in CH infrastructures, not as a data-cleaning problem. Drawing on cases from PHAROS photo archives, it categorises recurring forms of ambiguity and proposes design principles for representing uncertainty across both data structures and catalogue interfaces, balancing scholarly accuracy with usability.