Proceedings of the Twenty-Fourth Innovative Appications of Artificial Intelligence Conference eBird: A Human/Computer Learning Network for Biodiversity Conservation and Research Steve Kelling, Jeff Gerbracht, and Daniel Fink Carl Lagoze Cornell Lab of Ornithology, Cornell University Information Science, Cornell University
[email protected],
[email protected],
[email protected] [email protected] Weng-Keen Wong and Jun Yu Theodoros Damoulas and Carla Gomes School of EECS, Oregon State University Department of Computer Science, Cornell University
[email protected],
[email protected] [email protected],
[email protected] Abstract solve [2]. Now the World Wide Web provides the In this paper we describe eBird, a citizen science project that opportunity to engage large numbers of humans to solve takes advantage of human observational capacity and machine these problems. For example, engagement can be game- learning methods to explore the synergies between human based such as FoldIt, which attempts to predict the computation and mechanical computation. We call this model a structure of a protein by taking advantage of humans’ Human/Computer Learning Network, whose core is an active puzzle solving abilities [3]; or Galaxy Zoo, which has learning feedback loop between humans and machines that engaged more than 200,000 participants to classify more dramatically improves the quality of both, and thereby than 100 million galaxies [4]. Alternatively, the Web can continually improves the effectiveness of the network as a whole. be used to engage large numbers of participants to actively Human/Computer Learning Networks leverage the contributions collect data and submit it to central data repositories.