AI, citizen science data and expert opinions: towards an improved fossil identification process

Accurate identification of fossils is instrumental to palaeontological research but requires expert knowledge, is time consuming and subject to human biases. Through citizen science platforms and apps, artificial intelligence (AI)-assisted identifications can mitigate those challenges, as shown in successful biodiversity research applications. Digitised fossil data are relatively scarce, but a large and growing body of fossil datasets is becoming available in open data repositories through collection digitisation efforts. Furthermore, data are collected and validated by fossil enthusiasts on citizen science platforms. These datasets can be used as training data for deep learning classification models to provide both experts and citizen scientists with accurate, quick and easy to use tools to collect, validate and analyse palaeontological data. However, AI model performance may be limited by the size and quality of the training dataset. To explore the use of AI in fossil identifications, we present and compare a set of convolutional neural networks (CNNs) that are trained and tested on standardised images from museum and private collections (>46,000 images) and images from the online citizen science platform Oervondstchecker.nl (>74,000 images). Both datasets consist of Quaternary fossils and artefacts from the Netherlands and the southern North Sea Basin. Moreover, we compare model performance with identifications by ten domain experts and expert citizen scientists to gain a measure of data quality. This comparison shows that the AI models can outperform experts when trained and tested on standardised data, but not consistently across both datasets. Furthermore, expert identifications are far from unanimous, affecting the training data quality. Based on these insights we make recommendations on how to account for variable data quality to further optimise AI model training and performance. The synergy between AI model predictions and domain expert identifications can potentially rapidly increase the number of high-quality identifications of fossils and flag potential rare finds. To further increase data acquisition and public engagement, the models have been made publicly available online for use by professional experts, citizen scientists and the general public alike.