On Optimizing Human-Machine Task Assignments
This work addresses cost and accuracy challenges in crowdsourcing systems integrated with machine learning, though it appears incremental as it builds on existing methods for task assignment.
The paper tackles the problem of optimizing human-machine task assignments when using off-the-shelf machine classifiers, showing that reordering tasks for humans improves accuracy and joint optimization outperforms greedy parameter selection.
When crowdsourcing systems are used in combination with machine inference systems in the real world, they benefit the most when the machine system is deeply integrated with the crowd workers. However, if researchers wish to integrate the crowd with "off-the-shelf" machine classifiers, this deep integration is not always possible. This work explores two strategies to increase accuracy and decrease cost under this setting. First, we show that reordering tasks presented to the human can create a significant accuracy improvement. Further, we show that greedily choosing parameters to maximize machine accuracy is sub-optimal, and joint optimization of the combined system improves performance.