Characterising Learning in Informal Settings Using Deep Learning with Network Data

Online Citizen Science (CS) projects represent informal settings in which volunteers can learn and discuss about different areas of research while participating in scientific activities. In such settings, however, volunteer involvement is geared by project needs and individual learning occurs more as a side-effect. Data-driven, longitudinal studies examining such learning impacts are scarce. We study the user activity in the Chimp &See discussion forum on Zooniverse through the lens of social network analysis (SNA) to detect emerging user roles and evolutionary changes in behaviour indicative of learning. We explore the potential of structural network embeddings to identify similarities in relational patterns in comparison to externally assigned roles. Our analyses show that explicit roles such as “moderator” exhibit a high proximity in the embeddings, and that external promotions in the form of assigned role changes are preceded by a convergence of the corresponding behavioural patterns towards the ones of already established moderators, which is indicative of a profile change based on engagement and ensuing skill acquisition. Implications and potential applications are discussed.