Crowdworking, where small tasks that computers can't handle are outsourced to human workers online, is one way to teach algorithms how to recognize faces, places, and context. And with machine learning on the rise, crowdwork is only going to become more common. So is it the bellwether of a dystopian future, or the democratization of labor?
Your typical crowdworking task asks you to do something that's just a little too complex for a computer. A crowdwork platform like Amazon's Mechanical Turk—where more than 10,000 people are working at any given moment—might show you a picture of a glass shoe and request your associations (Cinderella, fairy godmother, Disney). Depending on the difficulty of the task, you might get paid a few cents to a few dollars. Meanwhile, the translation-based platform CastingWords will pay you $2.50 a minute if you're working for a whole day (which equates to $150 an hour).
Still, it's not necessarily a great way to make extra cash. "Clickworkers," as they're called, have no benefits and no rights, and there are no regulations about minimum wage—similar to other elements of the gig economy.
While the critics argue about crowdworking's ethics and efficacy, the interaction designers Stephan Bogner and Philipp Schmitt are imagining how crowdworking could better fit into our lives through UX and hardware, creating three experimental products that harness crowdwork for good. The duo, who recently graduated from the University of Design Schwaebisch Gmeund in Germany, call the project Human Element.
The idea of "micropayments," where consumers can pay a few cents to read an article or watch a video instead of looking at an ad, is an emerging idea in media, led by micropayment platforms like Blendle, where price of newspaper and magazine stories range from 9 to 49 cents.
Bogner and Schmitt think crowdwork could help. They came up with Workwall, an paywall-style interface that asks a reader to complete short crowdwork tasks in order to unlock an article, using the prices-per-article found on Blendle.
Crucially, Workwall would give you the option to see what larger project your task is part of, revealing identity of your temporary "employer" as well. It's a solution to a UX problem Bogner noticed during the research process: Some crowdwork tasks he came across fell under a category that Bogner calls "creepy tasks"—where the goal is dubious in nature. One job asked Bogner to rate the "Turkish-ness" of different people in photographs. Another asked Bogner to rate a customer service representative based on an exchange; though it purported to be a fictional scenario, Bogner still didn't know if his answers would contribute to getting someone fired. If tasks like that were to show up in Workwall, he says, you also have the option to skip them.
This kind of exchange model, where you offer two minutes of your time in exchange for two weeks of the journalist's, could also be applied to downloading MP3 files or video content.
Where Workwall exchanges crowdwork for content, another design—called the Spare-Cycle-Workstation—makes use of the spare moments of our days.
Suppose you're waiting for the subway to arrive, or sitting in a waiting room at the doctor. Instead of mindlessly flipping through your smartphone or reading, Bogner and Schmitt propose, you could perform short crowdwork tasks delivered on workstation kiosks stationed in public spaces, such as train platforms. By bringing digital labor into a public space, the workstation allows anyone to monetize their spare time.
While anyone can crowdwork through a smartphone app, Bogner says a modular workstation is uniquely valuable. These kiosks could include more complex technology to capture more sophisticated reactions, even measuring blood pressure or tracking the direction of eye movements in response to a task. And as machine learning makes more and more crowdworking tasks obsolete, other kinds of technology could easily be swapped in. This kind of crowdwork raises privacy concerns, especially if health data is involved. But it's already been shown that crowdworkers will share selfies—for the right price.
The designers' third concept envisions a way for anyone to submit tasks for crowdworkers around the globe to complete. It takes the shape of a scanner, where the employer scans a sheet of paper they want translated, typed, spell checked, or uploaded into an email, then inputs the speed and cost of work they're looking for. Once the job is done—it'll be finished quicker if the user pays more—the employer can send validation of a job well done using the "good job" button.
While this idea may be the least practical of the three, it is an attempt to imagine what a crowdwork-powered object would look like and, perhaps, restore some of the humanity to the process. Critics of crowdwork say it's easy to forget that human labor is hidden behind the computer—but the scanner gives employers a visualization of where on earth their work is being completed. The addition of the "good job" button serves as another reminder—you'd never tell a computer it's done a good job for running an algorithm, but all workers need praise when they've succeeded.
Crowdworking is already a part of the global economy, and it has the potential to impact the relationship between humans and machines by either improving our technology or facilitating unethical standards for digital work. For Bogner, Human Element was designed to confront us with this ambiguity, asking questions such as how much would we pay for a transcription or a translated email if it means the worker on the other end might be paid a living wage, or what kinds of tasks we're willing to take on if we know they're aiding discrimination by computers.
"Some people see a huge potential," Bogner says. "Others think it's dystopian, now we'll have to work all the time. We leave that open. It's up to you to decide."
[All Photos: Philipp Schmitt]