Adaptive Crowd Algorithms for Open-Ended Problems

Lydia Chilton is a graduate student in Human-Computer Interaction at the University of Washington working with James Landay and Dan Weld. She is currently a visiting student at Stanford University. Her foundational work in crowd algorithms began as an undergraduate and MEng student at MIT. She was awarded a Facebook Fellowship and a Brown Institute Grant to continue her work on crowdsourcing open-ended, creative artifacts. She likes Star Trek to a degree that embarrasses at least one of her advisors. 

Title: Adaptive Crowd Algorithms for Open-Ended Problems

Abstract: Although computers are powerful in their systematic approach to solving problems, people have open-ended problem solving skills that computers do not yet have. My research combines human and machine abilities to systematically solve problems too challenging for either to handle alone – problems that are big, hard, ill-defined, and require creativity. I introduce crowd algorithms that coordinate people's conceptual understanding to solve problems with these characteristics. Traditional algorithms tend to be rigid and highly parameterized. But to solve open-ended problems, crowd algorithms must adapt to the context of the problem.  I present four adaptive mechanisms for crowd algorithms: control structures, recursive generate-and-test, actionable feedback, and constraint satisfaction with backtracking. I show that adaptive crowd algorithms with these mechanisms can solve open-ended problems such as synthesizing disorganized information into taxonomies, creating conference programs, and generating creative artifacts such as humor.