Abstract
We conducted a quantitative analysis of ten citizen science projects hosted on the Zooniverse platform, using a data set of over 50 million activity records and more than 250,000 users, collected between December 2010 and July 2013. We examined the level of participation of users in Zooniverse discussion forums in relation to their contributions toward the completion of scientific (micro-)tasks. As Zooniverse is home to a multitude of projects, we were also interested in the emergence of cross-projects effects, and identified those project characteristics, most importantly the subject domain and the duration of a project. We also looked into the adoption of expert terminology, showing that this phenomenon is dependent on the scientific domain which a project addresses but also affected by how the communication features are actually used by a community. This is the first study of this kind in this increasingly important class of online community, and its insights will inform the design and further development of the Zooniverse platform, and of citizen science systems as a whole.
Introduction
Citizen science has evolved from a niche hobby into a powerhouse of "human computation," where hundreds of thousands of volunteers help professional scientists classify galaxies, transcribe historical records, or identify wildlife. While previous research has focused on data outputs or basic user motivation, this study investigates the social dynamics of these platforms, specifically the relationship between "Task" (scientific work) and "Talk" (forum discussion).
Methodology and Datasets
The researchers analyzed a massive dataset from the Zooniverse platform, comprising over 50 million activity records from 250,071 users across 10 diverse projects (e.g., Planet Hunters, Galaxy Zoo, Snapshot Serengeti) between 2010 and 2013. The methodology included:
- Community Modeling: Treating discussion threads as user networks to identify core vs. peripheral members.
- Time Series Analysis: Applying seasonal decomposition to identify engagement patterns.
- Vocabulary Shift Analysis: Measuring domain learning by tracking changes in the top 10 most frequent terms over a project's lifecycle.
- Cohort Analysis: Tracking user groups to measure retention and lifespan.
The "Wisdom of the Few" and Microposting
A primary finding is the stark power-law distribution of activity. A tiny "super-user" elite (less than 1%) drives over 70% of discussions and nearly 30% of scientific classifications. While 40.5% of users contribute to both Task and Talk, the intensity varies significantly; the median user performs 600 tasks but writes only 14 posts.
Furthermore, over 90% of "Talk" content consists of "untyped" microposts—short comments linked directly to specific tasks. This suggests forums function more as rapid-fire coordination tools than traditional long-form discussion spaces.
Vocabulary Dynamics and Learning
Vocabulary shifts are not uniform across projects or board types. In established domains like Astrophysics (GZ, PH), vocabulary shift was low, suggesting users arrive with existing knowledge or established norms. In contrast, projects like Snapshot Serengeti (SG) showed shifts toward expert terminology (e.g., species names).
Interestingly, shifts weren't always indicative of scientific growth; in some cases, language evolution was driven by technical troubleshooting and UI bug reporting.
Conclusion
This study highlights that community engagement is deeply tied to project domains and that forum activity often serves as a coordination tool. Future work should focus on predictive models for identifying super-users and experiments in forum design to encourage deeper scientific collaboration.
Summary updated