Machine Learning in Astronomy and Neuroscience: Lessons from Citizen Science

Abstract

While biodiversity monitoring has dominated the intersection of citizen science and machine learning, significant and instructive applications have emerged in astronomy and neuroscience. These domains offer unique insights into the integration of human and machine intelligence for scientific discovery. In astronomy, citizen science projects such as Galaxy Zoo and the Milky Way Project have engaged hundreds of thousands of volunteers in classifying celestial objects, with machine learning algorithms subsequently trained on volunteer classifications to automate analysis of vast image datasets. In neuroscience, projects such as Braindr have demonstrated how citizen scientists can amplify expert-labeled medical images, with deep learning algorithms learning to replicate expert quality assessments. This review examines the applications of machine learning in astronomy and neuroscience citizen science, analyzing the technical approaches, scientific outcomes, and lessons that can be generalized to other domains. We identify key success factors including the effective division of labor between humans and machines, the importance of expert validation of citizen science data, and the potential for machine learning to enable discoveries that would be impossible through human effort alone. We also address challenges including the difficulty of training models for rare or ambiguous objects, the need for careful quality assurance, and the importance of maintaining volunteer engagement when tasks are automated. We propose a framework for integrating machine learning into citizen science across diverse scientific domains, drawing on lessons from astronomy and neuroscience.