Abstract

We present research on complementary computing for task markets, where automated planning procedures are used to enlist computational and human expertise to jointly contribute to the solution of problems, based on a consideration of the competencies, availabilities, and pricing of the different problem-solving resources. The methods meld the area of human computation with automated reasoning. We describe a prototype named Lingua Mechanica that creates plans for harnessing people and computation to perform translation among multiple languages. With the prototype, people with different skills are recruited to refine rough translations generated by a machine translation system. We present details about the hardness of plan generation, provide methods for solving them, and review studies with tools for capturing human contributions.