Assessing quantitative methods in archaeology via simulated datasets: The Archaeoriddle challenge. Concept, project and motivations

Abstract

Compared to what is found in many other scientific disciplines, archaeological data are typically scarce, biased and fragmented. This, coupled with the fact that archaeologists can rarely test their hypotheses using experimental design, makes archaeological inference and our ability to assess the robustness of quantitative methods used to make such inferences challenging.

Archaeoriddle is a project that was born as an attempt to compare archaeological methods in an artificial scenario where the behaviour to be reconstructed was known. In this project we organised an experiment where a virtual archaeological record generated from a simulated interaction between hunter–gatherers and early farmers in a fictional landscape was shared with interested participants. Three archaeological questions were posed and the participants were challenged to answer them with the data that the developer team made available. The model and the generative processes behind the virtual record were known to the developers of the virtual world (Rabbithole) but not to the participants. Additionally, players were allowed to sample only a subset of the data from Rabbithole, mimicking real-life archaeological research and sampling efforts.

The long-term aim of the project is to assess how different methods performed under a controlled environment since, in this case, we knew the correct answers to the questions posed. This experience provided us with some insights into (1) how efficient various archaeological methods are in answering complex questions; (2) the degree of interest from archaeologists in improving their analytical techniques; and (3) the potential of archaeological method when free from external constraints (e.g. budget, fieldwork, etc.).