Comparing Competing Approaches to Crowdsourced Classifications of Biological Species: Work in progress

The world’s ecosystems are undergoing rapid changes driven by climate change and human development, leading to accelerated habitat loss. These combined processes have already resulted in extinction and decline in many species population, threatening the sustainability of multiple ecosystems, and, ultimately, the survival of all life on earth, including human life. Facing this biodiversity crisis, scientists and nature conservation organizations lack important data regarding the state of the populations of most species on the planet, with only a limited fraction systematically monitored. Current population estimation methods are too slow to match the rapid extinction rates. This paper outlines initial stages of our work, which ultimately aims to ecological monitoring by integrating machine learning with crowdsourced citizen science. Specifically, here we report on our work-in-progress testing two approaches for classifying camera-trap data: engaging school students through Zooniverse and using a gamified platform with a broader audience. Our goal with this work is to determine the better approach to approximate expert classifications, in terms of reliability, scalability, and speed. Eventually this research aims to enhance ecologists’ ability to cover more species, and create timely reports about the state of nature, to inform the creation of interventions and policies for a sustainable future.