Skip to content

Latest commit

 

History

History
162 lines (108 loc) · 7.52 KB

readme.md

File metadata and controls

162 lines (108 loc) · 7.52 KB
Raw

Supersubmiterator

A tool for managing external HITS on Amazon Mechanical Turk. This tool makes it easy to post external HITS and to download the data after participants completed the experiment. Supersubmiterator also supports automatic batching, i.e., splitting one HIT with a large number of assignments into several hits with fewer assignments per HIT.

The behavior is similar to the original Submiterator tool by Dan Lassiter and Erin Bennett. However, this is a pure Python implementation which does not require the Mechanical Turk CLI tools (which are no longer supported).

Setup

(Note: supersubmiterator is written in Python 3. If your default Python command is python2, make sure to run pip3 instead of pip and change the first line of supersubmiterator.py to #!/usr/bin/env python3.)

  1. If you do not already have an MTurk requester account, follow these instructions to sign up as a requester and create an access key/secret pair on Amazon Mechanical Turk.

  2. Install the boto3 and the xmltodict packages:

    pip install boto3
pip install xmltodict

  3. Install your MTurk credentials (access key and secret), following the boto3 directions. There are multiple ways to set these up, but the two easiest are:

    1. Put them in a file called ~/.aws/credentials, formatted like

       ```
 [default]
 aws_access_key_id = YOUR_ACCESS_KEY
 aws_secret_access_key = YOUR_SECRET_KEY
 ```

    2. Add environment variables in your Bash (or other shell) profile (e.g., ~/.bash_profile on most Macs). Add the following lines:

       ```
 export MTURK_ACCESS_KEY=<YOUR MTURK ACCESS KEY>
 export MTURK_SECRET=<YOUR MTURK SECRET>
 ```

    Supersubmiterator will use the environment variables if they are available by default. Otherwise, it will fall back to boto3 to find the credentials. For details on how this works, see the boto3 documentation

How to use supersubmiterator mturk tools

To post the HIT, first setup the config file.

Give this config file a unique label as its name: [LABEL].config.

{
"liveHIT":"no",
"title":"a title to show to turkers",
"description":"a description to show to turkers",
"experimentURL":"https://www.stanford.edu/~you/path/to/experiment.html",
"keywords":"language research stanford fun cognitive science university explanations",
"USonly?":"yes",
"minPercentPreviousHITsApproved":"95",
"minNumPreviousHITsApproved":"1000",
"frameheight":"650",
"reward":"0.00",
"numberofassignments":"10",
"assignmentsperhit": "9",
"assignmentduration":"1800",
"hitlifetime":"2592000",
"autoapprovaldelay":"60000",
"doesHaveQualification":"<ID_TO_MTURK_QUALIFICATION> or none",
"doesNotHaveQualification": "<ID_TO_MTURK_QUALIFICATION> or none"
}

The tool supports the following options:

Option Value Description
liveHIT "yes" or "no" If set to "no", the HIT is posted to the sandbox (useful for debugging).
title string A title that is shown to MTurk workers.
description string A description that is shown to MTurk workers.
experimentURL string A public HTTPS URL where your experiment is located.
keywords string A list of keywords that is shown to MTurk workers.
USOnly? "yes" or "no" If set to "yes", only MTurk workers with a US IP address can see and accept your HIT.
minPercentPreviousHITsApproved integer or "none" If set to an integer x between 0 and 100, only participants with at least x% previous HITs approved can see and accept your hit.
minNumPreviousHITsApproved integer or "none" If set to an integer x , only participants with at least x previous HITs approved can see and accept your hit.
frameheight integer The height (in pixel) of the iframe in which your experiment is displayed. Set this to at least the height of the largest trial in your experiment.
reward float The reward (in USD) that MTurk workers get for completing your HIT.
numberofassignments integer The total number of assignments (i.e., the total number of participants) for your experiment.
assignmentsperhit (optional) integer The number of assignmets per HIT. If this is set to a lower number than numberofassignments, the tool will automatically create multiple HITs with at most assignmentsperhit assignments per HIT.
assignmentduration integer Maximum time (in seconds) for MTurk workers to complete your experiment.
hitlifetime integer Lifetime (in seconds) of your HIT. After this period expires, MTurk workers can no longer see and accept your HIT.
autoapprovaldelay integer Time (in seconds) after which completed assigments are automatically approved.
doesHaveQualification (optional) string If set to a qualification ID, only MTurk workers with this qualification can see and accept your HIT.
doesNotHaveQualification (optional) string If set to a qualification ID, only MTurk workers without this qualification can see and accept your HIT.

Once you have setup the config file, run the following command in the terminal:

python supersubmiterator.py posthit [LABEL]

And then when you want to get the results:

python supersubmiterator.py getresults [LABEL]

This will create a long-form table of your data (several [LABEL]-*.csv files).

If you want to assign workers to a qualification (workers should be in [LABEL]-workerids.csv):

python supersubmiterator.py assignqualification -qualification_id <ID_TO_MTURK_QUALIFICATION> [LABEL]

How to make this even cooler

N.B. this will only work on unix.

If you want, you can make supersubmiterator a system-wide command, so you can just type (for example):

supersubmiterator posthit [LABEL]
supersubmiterator getresults [LABEL]

To do this, save the Submiterator repo somewhere where it won't move, copy-paste and run the following command:

chmod u+x supersubmiterator.py

Then make a directory called "bin" in your home folder and make sym-links to the Submiterator file:

cd ~
mkdir bin
cd bin
ln -s [PATH_TO_SUBMITERATOR_DIRECTORY]/supersubmiterator.py supersubmiterator

Then open up or create the file .bash_profile or .bashrc in your home directory and add the following line:

PATH=$PATH:~/bin

Then once you open up a new terminal, you should be able to use the supersubmiterator command as above.

Assigning qualifications

You can also use supersubmiterator to assign qualifications to workers:

supersubmiterator assignqualification -qualification_id [MTURK_QUALIFICATION_ID] [LABEL]

This will assign the qualifiaction [MTURK_QUALIFICATION_ID] to all workers in [LABEL]-workerids.csv.

Bonus payments

To pay workers a bonus, create a file [LABEL]-bonus.csv in the same directory as the [LABEL].config file, using the following format:

workerid,assignmentid,bonus
XYZAUMQ231,ABCCAHTNV34VFG76320,3.00
XYZPXP2345,DEFCAY5L23TKNC75GTVIP,0.50

Then run

supersubmiterator paybonus [LABEL]

This will pay the worker with the ID XYZAUMQ231 a bonus of $3.00, and the worker with the ID XYZPXP2345 a bonus of $0.50.