Dafna Shahaf and Eric Horvitz
9 December 2009
We lay out the challenges and opportunities of defining and optimizing the operation of human-computer task markets where planning and learning methods are used to identify and enlist both computational and human expertise to jointly contribute to the solution of problems, based on the competencies, availabilities, and pricing of different problem-solving resources. The approach melds the area of human computation with machine learning and decision making. We present details about the hardness of plan generation and provide methods for solving them. We illustrate key ideas in the context of a prototype named Lingua Mechanica that brings together human and machine resources for translation.
© 2009 Microsoft Corporation. All rights reserved.