Collecting Data with Mturk: The Nuts and Bolts and Issues
Associate Professor Haipeng (Allan) Chen, Mays Business School, Texas A&M University
27th Jun 2011 02:00 pm - Room 214/215, H69 - Economics and Business Building
Online data collection is becoming increasingly popular, but cost effective sources of respondents are often hard to find. A new resource, Mechanical Turk or Mturk, developed by Amazon offers a potential solution. At Mturk, researchers as "requesters" post work to be done, and qualified "workers" do the work at their location/pace for a very small fee (at a rate of about $1.40/hour or $.25-$.35 for 10-15 minutes). Therefore, for behavioural researchers Mturk can be an excellent source of respondents. To get interested researchers up to speed, especially with the limited documentation and human assistance on the Mturk website, this talk will offer a tutorial and answer questions on how to use Mturk for data collection from a user's perspective. I will address issues of data validity and participant demographics, and offer a step by step demonstration of how to use Mturk to run a simple experiment, including how to sign up and navigate the system, post a request, download data, reward workers, and problems you might encounter. Mturk is a potentially valuable tool not only for those interested in experimental or survey research but also for open-ended responses, pilot testing and any other tasks that can be conveyed electronically (e.g. data cleaning, transcribing, translation).