Mechanical Turk Workers: Secret Cogs In The Internet Marketplace
RENEE MONTAGNE, HOST:
A lot of jobs in manufacturing and farming have become automated, of course. Occupations like doctors or nurses might be harder to mechanize. In between is an entire series of things that computers can almost do without humans, but not quite. A system has sprung up to finish these in-between tasks. Planet Money's Caitlin Kenney explains how it works.
CAITLIN KENNEY, BYLINE: In 2005, Amazon.com had this problem - they had all this stuff for sale, but the information they had about it was messy. Like, let's say you wanted a red shirt.
ISAAC NICHOLS: You know, you may call them crimson shirts or you may call them light pink shirts.
KENNEY: Isaac Nichols worked at Amazon back then. He says at that time, Amazon's computers were bad at figuring out that the same color could have many different names.
NICHOLS: And that's a very human judgment call around do you call this a red shirt or a pink shirt or a crimson shirt?
KENNEY: Amazon had so many weird things like this, stuff that it needed people to do; stuff that software just couldn't do as well as humans. So Amazon created this opportunity for people out there in the Internet. Think of it as the world's smallest job offer - take a few seconds, tell us what color this shirt is or pick out the best photo and we'll pay you a few pennies for your trouble. The program was called Mechanical Turk. Mechanical Turk - it's a sly reference to this old thing - a chess-playing robot from the 18th century that actually had a human inside, and Amazon's Mechanical Turk was a huge hit. Ten years later, thousands of businesses post tiny little jobs in this site. Hey, take this survey. Hey, translate this sentence. And half a million people are there ready to do them. These people, these Turkers, are these secret cogs in the Internet machine, and they're anonymous. So we figured the best way to find them, to talk to them, was to hire them. We posted this task that involved listening to our podcast, and we embedded a secret message among the audio files.
(SOUNDBITE OF PODCAST, "PLANET MONEY")
KENNEY: And we want to talk to you about what your job is like. What's it like being a Turk? So write to us - email@example.com. Thanks.
And they did write to us.
Hello, are you our Turk?
KRISTOPHER WATSON: Yes.
KENNEY: You did our task.
WATSON: Yeah, quite a few times.
KENNEY: We found Kristopher Watson in Indianapolis, Matthew in Canton, Ohio, Manju in Bangalore. They were doing these little tasks from their offices, their bedrooms, their living rooms. And this kind of work was perfect for Ashley Taggard in California.
ASHLEY TAGGARD: I kind of had a problem with anxiety for most of my life. And so I didn't, like, leave the house. And so I'm kind of behind where everybody else is at my age.
KENNEY: Many of the Turkers we talked to came to this because they felt like the regular working world just wasn't the best fit. Kristopher Watson said he'd been on lots of unsuccessful interviews. And right now, doing these tiny, little tasks, called hits, it's just more comfortable.
Do you always put on work clothes?
WATSON: No. I - like, right now, I got on sweats and a sweatshirt. You don't want to overdress because if you submit a hit every, you know, 10, 20 seconds, you might get hot.
KENNEY: Here's how the system works from a Turker's perspective - a tiny, little job gets posted, usually for a few pennies, and the race is on to grab it first. Here's Ashley.
TAGGARD: Here's an 11-minute survey for $1.10, doesn't tell you what it is, from the Wharton School. They put a lot of hits out. Let's - I want to see what it is. I'm kind of curious. Oh, they're already gone. See? - There are no more hits available, so they're already gone.
KENNEY: Someone else beat you to it.
KENNEY: Does that happen a lot?
TAGGARD: Yeah, it does.
KENNEY: In the real labor market, there's friction - resumes and phone calls, in-person job interviews and salary negotiations. But here on Mechanical Turk, everything happens in just a few seconds. A job is posted and bam.
TAGGARD: Yeah, it's gone; it's gone again, so I - I'll go back over here.
KENNEY: And you can imagine what this competition does to the price. All these hundreds of thousands of workers out there racing for work, they're driving the price down - the price of themselves.
TAGGARD: Today, I've earned .12 cents. And yesterday, I earned $1.81. And then I had a 10.50 day - that was good.
KENNEY: It's piece work. Most of the people we talked to were turking on the side as a way to make extra money, to fill in the gaps. But some people who've done it a long time, who are faster and do certain kinds of jobs, they say they can make enough to do it full time. But things are tough here at this line between computer and human because as Turkers get good at doing a job, computers get even better, taking a job away from them. A few years ago, a bunch of Turkers were looking at photos of food labels, typing up the nutrition information. Now software can do that faster. The Turkers aren't needed. Ashley Taggard sees this line shifting all the time. She told me about this one task where her job was to teach a computer to sound more like a human. In this case, to sound like a doctor meeting with a patient who had just hurt their leg.
TAGGARD: They say then you have to give him five different ways of saying how's your leg doing today? So is your leg better today? Have you noticed an improvement in your leg? (Laughter) Literally, this is how you do it. You know, during the week, did your leg feel like it was doing better? You know, there's all these different stupid ways of saying the same thing.
KENNEY: If enough Turkers generate enough human-like responses, the next time, the computers can do it without them and the Turkers will move on to their next task. In theory, the Turkers could put themselves out of business this way, but few people seem worried about this. They say that at least for a while, there's going to be this gap between the messy human world and the orderly computer world. And a Turker is a cheap way to fill that gap. Caitlin Kenney, NPR News.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.