Consensus Building in On-Line Citizen Science

A number of initiatives invite members the public to perform online classification tasks such as identifying objects in images. These are crucial numerous large-scale Citizen Science projects different disciplines, with volunteers using their knowledge and support tools to, for example, identify species wildlife or classify galaxies by shapes. However, complex tasks, this case study on bumblebee, reaching an agreement between - even experts~-~may require consensus-building processes. Collaboration teamwork approaches problem solving decision-making have been widely documented improve both task performance user learning real world. Most these processes mediated through feedback delivered asynchronous manner, article thus addresses a central research question: How do participants involved identification respond forms provided collaboration, designed peer-learning performance? We tested four within collaboration task, where reviewed previously annotated data based information curated from peers long running citizen science initiative. The selected interfaces strong foundation social psychology literature can be applied practices well other communities. Results showed that while all increased accuracy, there were differences types consensus existed before collaboration. Such highlight usefulness during increasing accuracy furthering users' expertise tasks. found anonymised goal-directed free text comments posted most effective improving creating opportunities peer-learning, particularly was more difficult. This has significant implications extending practice across formal informal environments out variety users.