SAN JOSE, Calif. — The job didn’t pay much: four bucks an hour if you really hustled. But for Catherine Fraser, a recent community college grad from Mountain View looking to pick up a little extra spending cash, the work was a hoot.
“I told a friend ‘I’m now working in the porn industry’ because I had to watch little clips of adult movies for a minute or two and then give them titles,” says Fraser, 35, part of a growing global army of people making pennies in their spare time doing piecemeal — and often quirky — online micro-tasks. “We were limited to a small number of characters and encouraged to get creative.”
So Fraser would come up with, well, creative euphemisms “just to spice things up.”
She soon graduated to more savory micro-gigs– taking little surveys, transcribing insurance claims, penning product descriptions for $2 a shot– dipping her toes into a sprawling and little-known global subculture of digital grunt-workers. Largely unregulated — a spokesman for the California Labor Commissioner’s office said it was “starting to pop up on our radar”– this massive labor pool is starting to transform the traditional workplace, helping to power the tech renaissance unfolding in Silicon Valley.
“Crowdsourcing harnesses the collaborative nature of the Internet and enables us to connect, from a labor perspective, in ways we could never do before,” said analyst Martin Schneider with 451 Research. “Like manufacturing has done forever, crowd-labor lets us break down a job into tiny components, where one bit of fact-checking or writing a few sentences is now the equivalent of gluing that chip onto a computer board.”
Whether it’s rating the relevance of a search-engine’s results to help train its algorithms, or grading the sentiment of customer tweets (angry? irritated? happy?) for Fortune 500 companies, or screening dating-site photos for inappropriate content, this cadre of anonymous workers is supporting huge swaths of the social-networking empire.
“Crowd-sourced labor started off as this weird thing with people doing these funny little jobs in their spare time, but now it’s really catching on,” said Bill Quinn with Boulder, Colo.-based Trada, which hires people to help advertisers beef up online search campaigns. “I think 2013 will be the break-out year because the concept’s not so strange to companies anymore.”
The numbers back that up. A study by industry group crowdsourcing.org said crowd-labor revenues were up 75 percent in 2011 to $375 million. And the number of crowd-workers is growing even faster, climbing more than 100 percent last year, with about 40 percent of the six-million-member workforce living in developing countries.
While some workers can make six-figure annual salaries on more sophisticated tasks, one of the fastest-growing job segments, up 133 percent last year, is micro-tasks like the ones Amy Ellis of Alpine, Texas, does when her 2-year-old daughter’s at school.
“I’ve done things like 100-word product descriptions that pay $2.25 each, like why you should buy this brand of fluffy towels,” she said. She gets her work, like many of her fellow micro-taskers, from Amazon’s Mechanical Turk, an online labor market that pairs businesses, or “requesters,” with “Turkers” who compete for HITS, or “human intelligence tasks,” for a specific price and time-frame. One recent job listing on Turk, for example, offered to pay two cents to “copy text from business cards.”
While Ellis’ husband works full-time, she uses her average weekly take of $70 to buy groceries. “It’s not a lot of money,” she says, “but I love the ability to log in and do a few minutes of work whenever I feel like it.”
But like others, Ellis has also discovered the sleazier side of crowd-labor, starting with “scammer workers doing product reviews who just rip off content from other sites, and scammer requesters who are looking for fake reviews of hotels, restaurants and other businesses.
“My favorite one,” she said, “was a hotel in Virginia that had all these negative ratings on Yelp — cockroaches, hookers in the neighborhood, that sort of thing– and they were paying Turkers four bucks to read about the hotel online and write a 4-star review, even though they’d never set foot in the place.”
The community, says Mechanical Turk’s vice president Sharon Chiarella, is largely self-policed. “Workers and requesters earn a reputation on Turk,” she said. “And we make it clear to requesters that if they treat workers poorly they won’t work for them anymore.”
The larger question — and one with huge global implications as crowdsourcing redefines and in some cases kills traditional jobs and long-established labor-management models– is whether the crowd-labor pool could essentially become one big worldwide digital sweatshop. While industry studies show average hourly earnings across all categories range from about $7 in India to $16 in Western Europe, the fast-growing segment of micro-taskers earn half that on average, and some make only $1.50 an hour.
Lilly Irani, a PhD candidate in informatics at the University of California at Irvine who has studied crowd-labor, says these unrepresented workers remain ripe for exploitation. After all, someone is profiting from their underpaid work.
“We need to make sure that workers are part of the conversation about how these systems are designed,” said Irani. “And we must never forget that these are real people out there doing real work, sometimes making an extra $120 a week that literally helps them keep a roof over their heads.”
Here are a few of the so-called “microjobs” done by contributors to CrowdFlower, one of the largest crowd-sourced labor firms:
Categorization: Contributors match products, URLs, images or other items/information to their appropriate categories. For example, e-commerce sites with long lists of product names also need product attributes, so customers know exactly what a product is.
Search Relevance: Contributors rate the relevance of search results on a website in order to help “train” a website’s search algorithms and essentially make company’s websites more searchable.
Content Generation: Contributors are asked to write short pieces (length varies according to individual client needs) to quickly improve a company’s search-results ranking
Listing Verification and Enrichment: Contributors verify and/or correct existing business listings by searching business websites for contact information and checking websites to add supplemental information like hours of operation or menu data.
Image Moderation: Contributors tag images with descriptive terms or flag ones with unwanted content.
Sentiment Analysis: Contributors rate the sentiment or emotion of tweets, blog comments and other content so that the client company can gauge consumer feelings about its brand.
Transcription: Contributors take images of handwritten text or audio content and convert it into digital forms, creating versions that are easier to store and process.