Overview

Amazon Mechanical Turk (MTurk) has grown beyond its original imagined purpose of being a micro-task market and has become a great way to recruit real human subjects to do research online. However, anyone who wants to do more than a simple interaction for their experiment usually has to deal with several of the following issues:

• Loading HITs into Turk and paying workers
• Communicating with workers via a real-time interface
• Creating interactions with multiple users at once
• Enforcing limits on participation in experiments
• Logging a large amount of experimental data properly
• Testing complicated experiments thoroughly before spending $

TurkServer is designed to help you deal with all of these issues, and more, and to minimize the amount of duplicate work that researchers are doing to get their experiments on MTurk. TurkServer was built from our own experience running many experiments on MTurk involving interaction between multiple people, and is being constantly modified and improved to support more types of experiments. We hope that this can be a collaborative software effort so that we can slice away the minutiae and get down to the real, interesting work!

For a research perspective on why TurkServer was created, read the paper.