Guest post by Gideon Goldin and Adam Darlow
As MTurk was not designed for psychological research, it cannot be expected to provide the experimental control that psychologists typically exercise when recruiting participants for laboratory studies. In particular, MTurk lacks:
(I) The ability to exclude participants that have already participated in related studies.
(II) The ability to prevent study previews.
(III) The ability to verify participants’ completion of a study.
TurkGate, or Grouping and Access Tools for External surveys (for use with Amazon Mechanical Turk), gives researchers an easy-to-use web application for providing such control when using MTurk with externally hosted studies (e.g., Qualtrics surveys).
TurkGate groups related HITs together, such that participants may only access one HIT per group. These HITs only link to surveys after workers accept them, and only if they have not already accessed related HITs. Workers attempting to preview HITs receive information about the HIT, but no link. Once the HIT is accepted, the worker’s ID is tested against a database to verify eligibility before the worker is granted access to the study. TurkGate also generates completion codes that can be automatically verified.
Pe’er et al., (2012) have already created a relatively simple method to resolve some of these issues, which we refer to as the Qualtrics method since it is based on functionality inherent to Qualtrics. Like TurkGate, their method supports hosting studies on a variety of sites, as users can adopt (the free version of) Qualtrics for the purpose of screening participants before redirecting them elsewhere. In addition, their method also uses a script to prevent study previews. Researchers who use the Qualtrics method–and who also track response IDs–can address all of the MTurk limitations described above without needing to configure or maintain any soft- or hardware.
In contrast, TurkGate needs to be downloaded and installed on a web server (e.g., Apache HTTP Server) with a database management system (e.g., MySQL). As such, the security and reliability of TurkGate will depend on the system it is installed on. If the server goes down, TurkGate goes down with it–an unlikely occurrence with a professional service like Qualtrics. And although TurkGate is designed to work out-of-the-box, it may require some administration, such as updating versions or setting up database backups. These requirements suggest that TurkGate is best suited for an entire laboratory or department of researchers, where a single computer-savvy individual or IT professional can maintain it.
In return for this investment, however, TurkGate offers a streamlined workflow with several, distinct advantages:
(i) TurkGate manages related studies with minimal overhead. Researchers using either TurkGate or the Qualtrics method add a script to their Web Interface HITs that checks workers’ IDs against a list of restricted IDs. The difference is that the Qualtrics method maintains the list of IDs within each survey, whereas TurkGate maintains a global list of IDs for all surveys, separated by group. Having a centralized database has several advantages, especially when researchers collaborate. Namely, there is no need for researchers to store, share or update lists because all surveys use the same list and multiple researchers can use the same TurkGate installation. To make any pair of studies mutually exclusive, researchers simply assign them the same group name. Another benefit is that researchers can run their studies simultaneously, since TurkGate’s list is updated automatically and in real-time. Obviating the need to manage multiple lists of IDs makes creating surveys faster. A researcher simply submits a URL and group name for their survey into TurkGate to get the aforementioned script. They are then ready to create their HIT and run their survey. It is this highly optimized workflow that represents the original raison d’etre of TurkGate.
(ii) TurkGate disables HIT previews while preventing unnecessary HIT returns. The Qualtrics method, like TurkGate, prevents HIT previews by sending workers to an intermediary page prior to the actual study. However, explicit care was taken in developing TurkGate to prevent workers from ever needing to return HITs for which they are not eligible. This prevents the artificial and undesirable inflation of workers’ return rates. Instead, workers are provided with a group name, and if they recall having participated in the group, they know not to accept the HIT. If they do not recall participating, they can simply verify their eligibility by submitting their worker ID.
(iii) TurkGate offers intuitive, verifiable, and anonymous completion codes. TurkGate’s completion codes were crafted to support a number of features. First, the codes themselves contain useful information in human-readable form, including MTurk worker ID, group name, survey identifier, and an optional Qualtrics or LimeSurvey record ID (researchers can augment their codes with any number of additional key-value pairs). Critically, each code also contains an encrypted segment used to prevent fake codes. After running a batch of HITs, researchers simply copy and paste their MTurk results file into TurkGate, which then instantly flags invalid records and duplicates. For experimenters (or IRBs) concerned about anonymity, TurkGate can also verify participation without using response IDs (that are coupleable with study data). Those using Qualtrics often use response IDs as completion codes (http://www.qualtrics.com/university/researchsuite/faqs#codenumber), but this requires manual verification and precludes complete anonymity.
For researchers in psychology laboratories or departments that already have access to IT support capable of configuring and maintaining a web server and database, TurkGate will likely serve as a convenient and long-term solution. However, the Qualtrics method is better suited for researchers who are uninterested in the overhead of deploying a separate system, especially if they already use Qualtrics.
TurkGate is an actively developed, open-source project that users are free to download (and modify) via GitHub. It is used in multiple laboratories and continues to evolve based on the feedback and contributions of its users. Learn more about TurkGate here.
(Suggested citation: Goldin, G., Darlow, A. (2013). TurkGate (Version 0.4.0) [Software]. Available from http://gideongoldin.github.com/TurkGate/)
Pe’er, Eyal, Paolacci, Gabriele, Chandler, Jesse and Mueller, Pam, Selectively Recruiting Participants from Amazon Mechanical Turk Using Qualtrics (May 2, 2012). Available at SSRN: http://ssrn.com/abstract=2100631 or http://dx.doi.org/10.2139/ssrn.2100631