Citizen science through the lens of AI & serious games

This thesis examines the intersection of citizen science, serious games, and artificial intelligence, focusing on the methodological challenges that arise when participatory projects generate large and complex datasets. Citizen science has broadened public involvement in research, but recurring issues such as missing data, inconsistencies, observer variation, and limited reproducibility continue to affect data quality. Serious games also offer opportunities for engagement and learning, yet their behavioural data is often underused due to analytical constraints. By applying data science, machine learning, and explainable AI, this thesis shows how these techniques can improve the reliability and interpretation of citizen science data.

The research was carried out in collaboration with JGM Serious eXperiences, a serious-gaming organisation in Leeuwarden that develops escape-room-based games for professional training. These settings provided a controlled environment for observing real-time team behaviour under pressure. The collaboration reflects a form of participatory citizen science: JGM provided data and practical questions, while academic analysis refined behavioural indicators and strengthened scientific validity.

The thesis consists of three interconnected studies. The first, a systematic literature review, synthesises 90 publications on the use of AI in citizen science and highlights challenges such as data imbalance, bias, and limited transparency. The second study develops a data science framework to clean and validate JGM’s behavioural dataset, addressing missing and inconsistent entries. The third study applies machine learning and explainable AI to identify behavioural predictors of team success.

Together, these studies demonstrate how data science can enhance methodological rigour in citizen science by improving data preprocessing, analysis, and interpretation. The findings offer practical guidance on integrating data science and AI techniques into participatory projects to support more reliable, transparent, and impactful research.