Managing Artificial Intelligence in Archeology. An overview

• 27,100 publications on AI and archaeology in the last decade, doubling annually. • Archaeology embraces Big Data, requiring rich metadata for recontextualization and reuse. • Simplifying archaeological data for AI risks deterministic views, fixing fluid complexities into rigid categories. • Challenges related to data availability, transparency, ethical concerns, and technical difficulties. • Archaeology must address the ethical implications of its narratives, avoiding perpetuating academic power and digital divides. The integration of AI in archaeology poses several risks due to the oversimplification of complex archaeological data for computational ease. This reductionist approach fosters a deterministic view, treating provisional classifications as definitive truths and influencing subsequent interpretations. The reliance on legacy data and Big Data for AI training risks perpetuating outdated ideas and frameworks. As AI expands from automating tasks to interpreting and creating reconstructions, archaeologists must adopt a critical approach to avoid biased and harmful outputs. The deterministic view of AI hinders informed debate. Archaeologists should engage in discussions that address the classificatory, and ethical aspects as well as the materiality of AI. The accumulation of data in AI mimics storytelling but lacks the interpretative depth needed to understand historical human perspectives. Developing theories and narrative practices is essential to making archaeological data meaningful. The shift from a representational to a co-creative view of data is necessary to understand its re-use and the power dynamics involved. Finally, to normalise AI in archaeology, a critical and sceptical approach is needed to integrate AI into the real world and understand its implications and ethical considerations.