TurkServer

multi-user experiment framework for Amazon Mechanical Turk

October 2013 update: We are building the next generation of TurkServer. If you are just getting started with an experimental project, you will want to use the Meteor version of TurkServer instead. It accomplishes everything that this version does with even less code and allows you to focus on designing and running your experiment, not technical minutiae.

TurkServer builds a framework for real-time communication and group tasks on top of Amazon Mechanical Turk (MTurk). Features include real-time communication between clients and a server, user grouping, lobbying, and coordination abilities, and capabilities for tracking users over time. The framework is targeted toward experimenters interested in human computation, social science, user interface experiments on MTurk but can also be useful in many other circumstances.

Turkserver isn't a point-and-click framework like zTree; you need to have some programming experience in Java, and knowledge of user interface design using Java's Swing or client-side Javascript. However, TurkServer simplifies many other things that you would otherwise have to do from scratch, such as network communication, posting hits on MTurk, paying workers, and collecting data. Please read the wiki to learn more about how to use TurkServer. As with any other open-source project, we welcome useful collaboration: bug reports, help with the documentation, and good pull requests.

There are no versioned releases yet, but you can check out the core code via Git. We are currently updating documentation for TurkServer, and we plan to release some example experiments and interfaces in the near future once the corresponding projects are finished.