Unless you have a clear
strategy, MTurk work is a
complete waste of time.
True. I perused Mechanical Turk tasks some months ago, and was confronted by an array of plaintive, useless demands to perform what amounts to unproductive garbage picking for listless and disinterested college grad students running unimaginative projects without a clue as to whether the requested task is even possible.
It would be a request like:
Gather phone numbers from
this queue of web pages.
And they'd seem to have paid for a "database" of "leads" which was more than likely an excel spreadsheet of "hyperlinks" categorized according to a search query of keywords from the data provider.
You get into the queue, and start pulling up each URL in series, and they're all these expired domains with parked registrar pages for GoDaddy and Tucows or whatever, Along with some Geocities, Angelfile, Tripod and AOL home pages thrown in. Quickly, you get a sense that some fool of data science masters program enrollee paid good money for a dusty, mouldering text file, didn't even look at it, and dropped it right the fuck into a template for a Mechanical Turk task.
Now, there are three immediately obvious courses of action. One, abort and never again consider Mechanical Turk as a useful platform for operating an exchange of effort for rewards. Two, plead with the task owner by reporting feedback to them and ask them to stop for a moment and consider the flaws inherent to this framing of a human activity deemed worthy of compensation. Three, obey the letter of the law, and not the spirit, hold your nose, bellow the words "you asked for it!" and proceed to fill the task with the tech support numbers for all of the domain registrars, hoping that you'll not only get paid for grifting on the task owner, but also possibly inundate all these domain squatting registrars with robocalls trawling for psyche student surveys and questionnaires that will attempt to publish similarly terrible research papers designed with the intent to ostensibly "prove" a flawed hypothesis of human behavior with results that couldn't possibly be replicated because the hypothesis itself begs its own question.
Valuing my time, I just logged out, and haven't looked back since.
They need moderators to mechanical turk the quality of each task, because it benefits no one and wastes people's time, to even propose fruitless, unredeeming tasks.
Unless things have changed since winter, from what I witnessed, there is perhaps zero review of tasks to assess whether a request fits the profile of anything even remotely possible or worth trying.