Human computation is a growing research area that studies how to harness the combined power of humans and computers to solve problems that would be impossible for either to solve alone. The goal of this project is to introduce online education as a new vehicle and incentive mechanism for human computation. The central hypothesis is that problems that are difficult for computers can be transformed into tasks that are also educational, so that students solve the problems at the same time as they learn. With millions of people learning online, education could provide a powerful motivator for participation in distributed human computation. This project will demonstrate that education allows significantly more complex problems to be attacked with human computation than has been possible with previous paradigms for human computation. The project will also explore whether human computation can be a motivator for education.
The hypothesis will be tested on a new large-scale system called Duolingo, a free language-learning site in which students will solve problems that computers cannot yet solve. The site will present students with many types of activities, each exercising a different aspect of the foreign language while simultaneously channeling the students to perform a different task that artificial intelligence cannot yet accomplish. Some of the tasks that students will perform include: language translation, audio transcription, and image tagging.
Broader impacts. The project will provide a free language-learning site expected to help millions of users learn a foreign language. It will also provide large quantities of useful data in many languages to train more accurate machine learning algorithms for language translation, voice recognition, and computer vision. Duolingo will also serve as a platform for performing large-scale experiments on how people learn languages online. In addition, undergraduate and graduate classes will be improved and developed using this research.
