Crowdsourcing the Hardest Questions

Abstract

Crowdsourcing has emerged as a revolutionary approach for quickly solving problems that resist fully-algorithmic approaches and require human knowledge or judgment. Since individual crowd members are human, they can err, and a variety of quality-control algorithms have been developed to ensure accurate output. Unfortunately, most existing approaches rely on aggregation of multiple, independent answers in their attempt to ensure quality. While these methods yield accurate results in many cases, they fail on extremely difficult problems where the majority of workers get the answer wrong. This has caused many researchers to conclude that crowdsourcing can never achieve an accuracy of greater than 85-90% on some tasks, no matter how many workers are involved. In contrast, we believe that the fundamental problem stems from the simplistic aggregation approaches used to date. We propose four methods for improving the accuracy of crowdsourcing: microtalk, Bayesian crowd reflection, decision-theoretic question routing, and differential instruction. If successful, our methods will significantly improve the accuracy of both commercial and research crowdsourcing methods.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2016
Source ID
N000141512774

Entities

People

  • Daniel S. Weld

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Washington

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Systems Analysis and Design
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks