top of page

The Rise of the mTurk Super Worker...and Why it Shouldn't be a Problem at All

For a little while, I've been hearing this term "Super Worker" thrown about a bit concerning highly experienced, or career-level Mechanical Turk workers. For the record, I'm one of them. If you work mTurk for spare cash, if you complete less than 50 Hits a week; you're not a Super Worker. Super Workers make up just over 10% of the mTurk workforce, but account for something like 60% of completed Hits on the platform.

Okay, so what ?

The notion seems to be that the main power Super Workers wield is an acquired immunity to the standard manipulations that are currently employed by academic requesters, mostly within survey type Hits. Some have argued that, for example, the fact that we're not easily duped into believing that we're "working with a partner" in a Hit after completing thousands of other Hits that disclose to us after the fact that we were in fact being deceived, inadvertently clears our consciences to make overly self-serving decisions within the Hit.

Those with an axe to grind say that the Super Worker is on the platform too much, and therefore naturally going to grab the lion's share of work...well, yeah...but who wouldn't expect that? They say that the Super Worker employs technology, scripts for example, that skew the game in our favor. Of course we are, but scripts are so pervasive that we see it more as handicapping ourselves if we didn't. The requester naturally wants as large of a sampling as they can get/afford. What isn't being recognized by requesters is that if they were somehow able to weed out the Super Worker, they wouldn't get their data completed in the time frame that they need. The regular worker just isn't on the platform often, or long enough to sustain the appetite for data that is demanded by requesters.

So, what actually IS the problem ?

To start with, requesters are setting up Hits with qualifications that prevent the less experienced workers from getting to do their Hits. This is largely the fault of the mTurk platform. The platform either defaults, or encourages requesters to opt for qualifications/requirements to filter the worker pool down to exactly "who you're looking for". What this actually is for is to up-sell naive (to the system) requesters and this is backfiring upon them. For example, by requiring workers to have in excess of several thousand Hits under their belts to qualify...thereby steering themselves directly into the wall (the Super Workers) they're trying to avoid.

Secondly, when you're served as many meals as you can extremely small portions, you can learn to eat fast, or you can starve. Super Workers aren't here in some altruistic gesture to contribute to science...we're trying to survive.

It's Human Nature

Mostly, the problem boils down to plain old Human laziness. We like to think of requesters as a group comprised entirely of serious social scientists. The truth of the matter is that many of them are undergrad or graduate students that are trying to finish up some research for a grade...period. They don't really care what the results are, or if scientific knowledge is furthered...they just want to graduate. What any Super Worker can tell you for sure is that most Hits are set up on very similar templates...that's why we spend most of our days answering to what extent we'd change our lives if we could do it over again, or how much we'd donate to a particular "theoretical" charity. Would it kill anybody to write their own questions? How much scientific progress can be made by rehashing the same template over and over?

So, to sum it up, the mTurk platform would dry up and blow away without the Super Worker. To actually solve the problem, requesters would have to increase their rate of pay, Amazon would have to decrease their fees, requesters would have to compose their own studies (no templates), and all of these things would have to be done to such an extent as to attract an enormous (10x) increase in the pool of available workers to make up for the loss of the Super Worker.

27 views0 comments


bottom of page