AI-Assisted Reconstruction of Past Environments

Reconstructing paleoenvironmental conditions in archaeology is key to understanding past human-environment interactions. However, this process often involves extensive, repetitive, and time-consuming data analysis. The introduction of Artificial Intelligence (AI) techniques can be a valid tool for assisting experts in managing this workload. This project focuses on environmental archaeology, employing AI techniques (e.g., tree-based algorithms, neural networks) to predict biomes by analysing pollen counts collected during sampling or archaeological excavations. Pollen counting involves quantifying the species or families in a sample, reflecting the vegetation composition and distribution during pollen deposition. Associating species to infer environmental conditions requires expert input and the consideration of multiple variables (e.g., soil characteristics and pollen dispersal) to explore human-environment dynamics over time. The study integrates predictive and generative AI models to enhance data analysis. Predictive AI supports experts by accelerating species identification and association, while generative AI produces visually impactful reconstructions. The tool is developed collaboratively with professionals from diverse fields (e.g., archaeologists, mathematicians, botanists) to ensure scientifically valid results. Additionally, it serves as both an educational resource and a means for effective dissemination.