The cost of labour on Mechanical Turk is both tantilising and unsettling. Tantilising because it’s so cheap. That makes it possible to outsource skilled tasks like translation and transcription at remarkably low cost. As a journalist, it’d be great if I could afford to routinely transcribe interview tapes. Prices are still a little high for that, but my hunch is that they will come down.
The unsettling aspect concerns the impact of these wages. Some folks who work flat-out on Mechanical Turk probably earn $2.00 per hour. That’s okay in India, but many Turkers are based in the United States, where the federal minimum wage is $7.25 per hour. Discussion forums like Turker Nation are full of people complaining about low wages, but there are still plenty of US-based workers on the site. Sure, some workers find it a pleasant place to waste time and don’t see it as a full-time job. But that doesn’t account for all workers. I’d like to know how many people depend on the site.
The Turker forums are also full of workers complaining about scammers — employers who post jobs and don’t pay, or reject work arbitrarily, thus removing the need to pay. There is a lot of frustration directed at Amazon also. It’s clear that workers think that the company could be doing a better job of cleaning up the site.
I’ve been taking a look at these issues over the past couple of months. The results came out last week in New Scientist. The story is subscriber-only right now, but the brief summary is that scammers continue to evade Amazon’s moderator system. There are jobs that attempt to trick workers into giving away credit card details. Plus lots of jobs that are clearly lead generation scams. The worker fills in an online survey, pretending to be interested in, say, going back to college. Then the employer gets paid a referral fee from colleges looking for new recruits. So the college is being conned and, in the jobs I tried, the workers were too, because the employer didn’t pay me after I filled in the survey. (I’m still getting calls from colleges).
I was struck by the variety of these scams, but they don’t seem to bother experienced workers too much. After the story came out, I thanked the workers on Turker Nation who had helped me, and posted my piece on the forum. One response to my story noted that the scams were easy to avoid and that it was low wages that are the problem.
Scams such as things offering $20 for me to sign up at whatever are not a real big concern for me on Mturk. Do most workers fall for this often? More than once? Ever?
The real problem for workers on Mturk is lowballing requesters. This is how I see it. It’s as if I were walking on a road that was 90% strewn with sharp slivers of glass, and someone is saying – The problem is that there are some huge potholes!”
TurkerNation is the last place that you should be conducting research about the attitudes of workers. There is a huge selection bias: they are workers with enough time in their hands to participate in forums, instead of working, and people that have enough time to spend it on Turk. This is a tiny fraction of the workers on MTurk, and definitely not representative of the overall population.
Having a worker blaming the employer for paying low wages? Really? How original! Do you think this never happens, in any profession?
All employees, in all places outside Mechanical Turk, feel that they are richly rewarded, and paid unjustifiably high wages? Well, I would believe that *every employee* thinks that their employer is low-balling them.
Ask these TurkerNation workers about the tasks that they like to complete: I bet it will be mostly about content generation. And you already know what is the purpose of the content created on Mechanical Turk.
Posted by John Draper on July 14th, 2011.
John has a point. I don’t think that makes Turker Nation a bad place to do research, but it does mean that the comments left there should be taken as representative of the subset of Turkers that hang out there, not the whole of MTurk. My guess is that Turker Nation is almost entirely US-based Turkers.
And when he says that I already know about the purpose of content generated on MTurk, I think he’s talking about this post.
Posted by Jim Giles on July 15th, 2011.
College and school HITs on mturk were really numerous as were tons of car insurance HITs. There were alot of those “do offers for a gift websites”. They wanted zip codes, emails, many wanted full personal info down to birthdates. All these “employers” were were lead generators and affiliates of other lead generators. Turns out there is a big US company in California that specializes in schools AND in car insurance – has several .coms for matching schools and a couple in insurance. Turns out, this company is a big affiliate for the “gift for offer websites” and there are myriad of those, all aliases of 2 or 3 main “icky” companies. The car insurance, school, affiliate of “ick” company has an A rating on the BBB. It has also been through legal procedings with the FTC (2006 or 7?)and settled (it breached ethics and proper advertising. I think many of the scams getting all the personal data from zip codes to emails to more were actually of this one company, but I do not know for sure. I think the company has left mturk, as its work no longer seems to be there, or I may be wrong. Did they pay the worker? Sometimes. Sometimes they say they want a work product on page such andsuch of the survey, then, whenyou fill in the data to look for the work product it is not even remotely there. And, by then, they’ve got all the data to sell, as do their affiliate webpage. It is bad when this is combined with pressure for valid data under threat of rejection and leaving no time to read policies or think of who the work is for. The spam that is received from some of these lead generators is 99 percent predatory, taking expense of the consumer so the advertiser can profit at his expense. Many of the numerous school ads were useless, non-accredited, saying no tests were needed to get degrees in months – i.e. school scams. Worse yet, all the solits indicate that someone tags the recipients with derogatory labels – can’t pay bills, need money bad, willing to get a degree without working for it, certain health problems, stupid or greedy and falls for scams, this problem, that problem, etc. Eventually, worse fraudsters buy the data and it is impossible to stop the predatory and illegal daily deluges; they will go on for years, as will any negative descriptions that seem to be attached to the recipient data, shown by the continual onslaught of certain types of spam, and even by the same industries targetting the recipients’ sidebars and online ads.
Posted by AvoidingScams on August 29th, 2011.