Exploring Archaeological Ceramics with AI

Carlotta Gardner, Associate Editor for Archaeological Ceramics

Conversations about the possibilities and limitations of Artificial Intelligence (AI) have become common place since November 2022 when ChatGPT was released to the world. In archaeology different components of AI have been used in a variety of ways for a relatively long time now. AI has been used to analyse ceramic typologies and to reconstruct forms, for pattern recognition of macroscopic and microscopic observations and for data analysis.

There have been a number of projects that use AI to analyse images to assist archaeologists in the classification and interpretation of ceramic objects. One of the best-known is ArchAide which started in 2016 and founded within the European Union's Horizon 2020 research and innovation programme. The platform utilizes AI algorithms trained on vast datasets of pottery to automatically recognise and classify different ceramic types, styles, and attributes. Described as the ‘Archaeologists dream tool’ it can accurately identify archaeological pottery and save vast amounts of time.

Beyond typological classification, AI can aid in the extraction of detailed features from ceramics. Advanced computer vision techniques enable the identification and analysis of minute details, such as surface textures, colour variations, and glaze patterns, which are crucial for understanding the manufacturing techniques and cultural significance of these artefacts. AI algorithms can detect and quantify these features, making it possible to compare them across different pottery assemblages, regions, or time periods. An example of this work is the ARCADIA project, which used Convolutional Neural Network models to classify engraved pottery from Saran, France (Chetouani et al. 2020). This automated approach allows researchers to identify subtle changes in ceramic production and trade networks, providing valuable insights into cultural interactions and technological advancements.

Most applications of AI in archaeology are in the analysis of images and recognising pattern, as demonstrated above. It is less common for AI to be used to analyse numerical data in archaeology, such as results from elemental analysis. However, machine learning algorithms can be trained on databases of known ceramic sources, enabling researchers to identify the geological and chemical signatures unique to particular regions.

There are some pioneering studies that have used machine learning to explore issues of provenance with a variety of materials, from soil (Oonk & Spiker 2015) to obsidian (Lopez-Garcia et al. 2020). Data from the elemental analysis of archaeological ceramics have also been explored in a number of papers and the results are promising. Anglisano et al. (2022) use supervised modelling for to investigate a data set acquired through ED-XRF analysis of 208 samples from the Barcelona area. The case study was used to explore the application of the method and how successful it was. The paper also provides all the code and details of how to use it in R. Over all the results are encouraging, the authors note that for this approach to be successful it is essential to obtain large numbers and reliable reference samples covering the likely areas of provenance.

The integration of AI into archaeological research has opened new doors in the study of ceramics. Through automated typological classification, feature extraction, and pattern recognition, AI algorithms facilitate the analysis of large quantities of archaeological ceramics, allowing for faster and more accurate interpretations. Furthermore, AI's assistance in provenance determination contributes to a comprehensive understanding of ancient societies and their cultural interactions.

Anglisano, A.; Casas, L.; Queralt, I.; Di Febo, R. 2022. Supervised Machine Learning Algorithms to Predict Provenance of Archaeological Pottery Fragments. Sustainability, 14,11214.

Chetouani, A.; Treuillet, S.; Exbrayat, M.; Jesset, S. 2020. Classification of engraved pottery sherds mixing deep-learning features by compact bilinear pooling. Pattern Recogniton Letters, 131, 1–7.

Oonk, S.; Spijker, J. 2015 A supervised machine-learning approach towards geochemical predictive modelling in archaeology. Journal of Archaeological Science, 59, 80–88.