Archival Paradata and Artificial Intelligence in Archaeology

The increasing use of Artificial Intelligence (AI) systems presents significant challenges regarding transparency and accountability. This paper investigates the application of paradata – information detailing the procedures and tools used in AI-facilitated processes – as a strategy for enhancing transparency in a context of particularly complex and historically-rooted practice: archaeology. Despite growing recognition of the need for transparent AI use, current documentation practices are often inadequate. This paper argues that paradata offers a valuable approach to addressing these deficiencies, particularly when considering the inherently heterogeneous and temporally-extended nature of archaeological data and methods. Focusing on archaeological practice, the potential of paradata is explored for documenting not only how an AI system operates, but also how it is utilised within complex, historically-situated processes. We propose a ‘paradata matrix’ as a practical method for systematically identifying and documenting the necessary paradata, recognizing that a multiplicity of transparencies – rather than singular, context-independent understanding – is often crucial for navigating the socio-technico-informational complexities of AI use.