Change in Mturk HIT behavior

Question

I am running an experiment on Mechanical Turk that has several different lists. The task runs in the browser inside of Mechanical Turk (i.e., I don't direct people to a different webpage). These HITs worked fine until yesterday.

Yesterday something changed, and now when a worker accepts my HIT, an HTML file is downloaded to their computer/device. In the past, the task just started in the same browser window. If the worker opens the downloaded file, they can go through the task but they get an error when they try to submit it. This was reported to me by workers and I was able to replicate this behavior in the sandbox.

The day before this issue started, I ran two batches of HITs with no issues. I went back to those HITs and now they are having the same problem too. Nothing has changed in my code... one day it worked fine and the next day it didn't. I can't figure out how to fix this issue. I'd appreciate any help!

LM · Answer

Update: After quite a bit of troubleshooting, I've found that the only difference between my HITs that are working and the HITs that are not working is the number of lines in the code. If the number of lines is above a certain number (I haven't pinpointed the exact number yet), the HIT will not run properly. Accepting the HIT starts a file download. If the code is shorter, the HIT will run in the browser as expected. Is there some kind of limit that has been imposed on the length of the code?

Aparna_P · Answer

Hello,

I am very sorry to hear this is happening. Unfortunately, there is no general guidance we would be able to offer here. I would advise you to contact our MTurk Support using this form:
https://support.aws.amazon.com/#/contacts/aws-mechanical-turk

They have the tools and insight to assist with issues such as this, please be sure to include all the details the space will allow.

Slava · Answer

Any chance the we get help from AMT support here?

Change in Mturk HIT behavior

Relevant content