Mturk Now More Expensive

As many readers have probably heard, Amazon's Mechanical Turk will soon (July 22, 2015) be charging substantially more for its services. Before, Mturk took 10% of what researchers ("requesters") paid workers. Now it will be effectively 40% (in the rare case that you need fewer than 10 distinct participants, then it's only 20%).

Naturally, some researchers are upset, as documented at The Chronicle of Higher Education ("Researchers Complain About Changes in Amazon Tool Used for Surveys").

Two quick downsides that stand out to me:

(1) Running even a small x-phi study on Mturk will more likely require grant money, which is scarce in the humanities.

(2) An increase in operating costs will probably motivate researchers to still get the same mileage out of their funds. This makes it even less likely that workers will get paid a living wage, which is an ongoing ethical issue with use of the site.

Any reactions from the x-phi community?

(Note: If interested, this blog recently featured an excellent post on the status of Mturk as a research tool.)

Comments

[Shen-yi Liao]

It's actually not 20%. For most studies, which require more than 9 participants, Amazon will now charge a 40% fee. So it's bad, really bad.

My sense is that things are still in the WTF stage right now. People are discussing alternatives. There have always been a number of other sites, but none of them critical mass like MTurk. Commercial ones like Prolific Academic are offering discounts to try to get some researchers in: https://prolificacademic.co.uk/rr?ref=5ZFZ276D . There are also institutional ones but they are hard to access for those outside of the respective institutions. And finally there have been talks of building non-profit ones (apparently Brian Nosek wrote many grants on this) but that's a ways away.

There is a Facebook group called "WTF Amazon mturk?!" that may be worth following.

[Josh May]

Yikes! Thanks for the correction. I think I misunderstood the "10 assignments" clause as meaning 10 tasks per worker. But it looks like it really is number of people who perform the task. It's even worse than I thought!