Presentation #306.05 in the session Innovation and Discovery in Solar and Space Physics Enabled by Citizen Science.
With the unprecedented rise in data volume of observational astrophysical data, there is a requirement for robust infrastructures that are able to efficiently and accurately process the data. Machine learning, particularly deep neural networks, have shown good performance, especially in classification and image segmentation problems relating to very large datasets, but for complex tasks and when there is large confusion between classes in the data, these networks have required human supervision to provide the necessary accuracy. On the other hand, citizen scientists have played a large role in doing simple analyses of the data with very little technical training, and demonstrated excellent efficiency in analyzing intermediate scale (~1 million samples) datasets. With over 2.5 million volunteers, Zooniverse is the leading platform that hosts citizen science projects, and over the past decade has led to a large number of discoveries, particularly pertaining to objects which are unique or unusual in the dataset (e.g., the discovery of the Green Peas from the Galaxy Zoo data). However, with the ever growing volume of astrophysical data (e.g., over 10 billion images from Vera Rubin Observatory), it is impossible to use only citizen science, or machine learning to tackle big data analysis problems. There is a growing need to marry the two ideas into an efficient pipeline: one that handles the large data volume with machine learning, but is also able to find new and interesting objects through human intervention. Recently, several projects on Zooniverse have made great strides towards this ‘human-in-the-loop’ framework, and have produced a wealth of knowledge on this front. In this talk, I will detail the citizen science efforts that are undertaken on the Zooniverse platform. I will also present some frameworks that are being designed to create deeper engagement paths for citizen science volunteers, in an effort to enable better avenues for lifelong learning for volunteers and for easier serendipitous discoveries for the research teams.