To improve scalability and data quality, the eBird Human-Computer Learning Network (HCLN) blends emerging techniques that integrate the speed and scalability of mechanical computation, using advances in Artificial Intelligence, with the real intelligence of human computation to solve computational problems that are beyond the scope of existing algorithms. In addition to developing emergent filters and new models of contributor expertise, this work makes extensive use of the semantic links between observations and observers to mine additional information from the existing data in order to strategically address data quality issues.
Our extended research team have found evidence of learning-through-doing and improvements in participant performance with accumulated project experience. Initial analyses also confirmed expectations about localization of observers’ knowledge of species, and showed that measures of performance effectively distinguish between observers who are and are not able to detect and identify “secretive” or ambiguous species.
This project is a collaboration with the Cornell Lab of Ornithology, eBird, University of Michigan, and Oregon State University.