Enhancing stone provenance studies through software built with language model artificial intelligence (AI): An example of ancient Calabrian quarries (southern Italy)

Abstract This study represents the first attempt to develop archaeometric software that enables researchers without programming knowledge to address archaeometric challenges, specifically determining the provenance of rocks extracted from ancient quarries. Through interaction with ChatGPT 4.0, an advanced artificial intelligence (AI) language model, the authors guided the AI to develop StoneScanalyzer 1.0 software in Python programming language. The step‐by‐step collaborative process resulted in software capable of automatically extracting 43 quantitative variables from sets of images of cut, wet rocks acquired under reflected light, thin sections of rocks acquired under natural and polarized transmitted light using a flatbed scanner. Data elaboration using linear discriminant analysis (LDA) models and principal component analysis (PCA) led to the construction of discriminant diagrams for 250 samples taken from 10 quarries located in Calabria (southern Italy). StoneScanalyzer 1.0 software can be easily used by researchers without basic petrographic or geological knowledge, making it highly appealing as a first step for archaeologists, architects, art historians and anyone interested in studying rock provenance without expertise in mineralogy, geochemistry or petrography.