On platforms like Amazon’s Mechanical Turk (Mturk) or CrowdFlower, large companies, such as Facebook, Google, Twitter, eBay, and LinkedIn, have workers fulfill orders of billions of microtasks each year to improve their products, often to provide large datasets that guide machine-learning algorithms.
Tasks range from recording which way a photo is rotated to captioning pictures of Pusheen the cat. The market depends on humans easily performing certain tasks that computers find difficult.
The business of microwork is big: experts’ estimates range from 40 million workers—put forward by the International Labour Organization—to 300 million in China alone, the World Bank’s guess. And it’s booming: in 2015, the World Bank predicted CrowdFlower (now Figure Eight) would quadruple their revenue from 2013 to 2016.
But on the old, centralized platforms, things have already got out of hand. It hardly comes as a surprise that these platforms have been labeled as the sweatshops of online labor.
A study released by the International Labour Organization last month said wages on these platforms averaged just $2.16 per hour, after the platforms had taken their cut. And because lots of these workers don’t have ready access to bank accounts, many complain that checks can take months to arrive—if they come at all.
But put the marketplace on the blockchain and the hope is that workers could sweat a little less, and the blockchain factory could continue to grow.
Companies like Gems, Storm, and Coinworker are already using blockchain to undercut services like Mturk, with hundreds of thousands of users between them. Storm alone claims to have over 250,000 users.
Their success has attracted the attention of bigger fish. Vodi, a messaging app that’s popular in Asia, just partnered with a major Indian Internet provider, Bittel. That could mean tens of millions of users in the microtask economy over the next few years.
The blockchain factory
One of the main problems with platforms like Mturk is the difficulty that comes with verifying the accuracy of tasks. It’s such a big problem that these centralized platforms require each task to be verified up to 15 times before the system is convinced of a task’s accuracy.
Picture the scene: if one worker’s task is to click on all the pictures of balloons in a gallery, there will be several others asked to do the exact same job. Each job will be checked by another worker, tasked with verifying that the pictures are indeed of balloons, and maybe even another verifying the verifier.
But with blockchain, the promise is that anyone can use micro-task workers without needing to worry about task verification, trust, or payments.
Companies like Gems, Storm, and Coinworker, are beating MTurk at their own game because their technology doesn’t require workers to have bank accounts, so they can access millions of unbanked workers that Amazon—using a centralized system—won’t hire.
“Every bone in my body is screaming out money grab.”Reddit user CryptOG613.
And because these decentralized platforms don’t require a rent-seeking middleman like Amazon to process transactions on their platform, both miners and requesters get a better deal. If Gems’ whitepaper is correct, requesters and miners will return to a fair market equilibrium.
Gems, a company founded by Harvard drop-outs and Thiel Fellows Rory and Kieran O’Reilly, says its platform offers a better solution for both work “requesters” and the “miners” who complete those tasks.
And it has already attracted some of blockchain’s biggest stars: co-founder of Augur Joey Krug, co-founder of reCAPTCHA Ben Maurer, co-founder of Aragon Luis Cuende, and co-founder of District0x Joe Urgo have all come on as advisors.
Gems' protocol is built on Ethereum, and their whitepaper claims the Gems platform will ward off malicious actors and reward fair players.
On Gems, miners “stake” a small bet on themselves that their work is correct—that the picture they captioned as a balloon is indeed a picture of a balloon. Then, verifiers—those miners promoted to oversee it all—independently bet for or against the miner’s work. The idea is that miners would only bet that their own work is accurate if they were doing honest work, so there’s less room for foul play. And because a verifier’s reputation is on the line, they’re likely to tread carefully.
So where Amazon might require 15 workers to verify the accuracy of a single task, the reputation system of blockchain companies like Gems or Storm could bring that number down to 10. It’s also cheaper for requesters—those asking the miners to complete the menial tasks. The Gems platform doesn’t charge a central fee and offers free, open-source modules for requesters to submit to miners. But like so many blockchain projects, the paradise painted by a whitepaper has turned into something a lot less rosy.
Captioners of the world unite
As part of Gems’ initial coin offering, Gems kept around 75 percent of the tokens raised—something not mentioned in their whitepaper—losing the trust of public investors. “Every bone in my body is screaming out money grab,” writes Reddit user CryptOG613.
“We, as a community, feel backstabbed after having given so much to the project without any guarantee of a reward,” writes Titus Decali, an ICO advisor in an update to an original article which praised Gems for its “outstanding” team and “clear” goals.
It seems that though Gems claims that blockchain ensures transparency, the same can’t be said for the company itself. In recent months, Gems has stepped out of the limelight as it continues to fight a lengthy lawsuit concerning its domain name. (Gems did not respond to Decrypt’s requests for comment.)
Other platforms aren’t much better. On Reddit, forums quickly reveal how microtask company Storm delayed paying users for tasks completed when the token value was high until a year later, when token values had dropped.
“If you're not paying workers when the coin is down, then the whole system falls apart.”Wulf Kaal, researcher at St. Thomas University.
In a Reddit post titled “Did they steal millions from us?,” Brooke Brody, a microtask worker who previously worked for Storm, writes that “[Storm] abused and defrauded us in what could amount to millions collectively.”
When Storm finally decided to pay, they only offered Brody 25 percent of the agreed payment—and that’s after the coin had already fallen in value.
Storm replied to Brody’s Reddit post that the delay was due to false claims filling up their email backlog, and, in and email Brody sent Decrypt, that they had to implement the policy “[t]o protect our system from being abused and defrauded.”
But Brody questions why legitimate users were penalized for the backlog and has already filed complaints with the Better Business Bureau.
The "sockpuppet" protocol
But Wulf Kaal, a researcher at St. Thomas University, says that the current generation of blockchain microtask platforms isn’t equipped to solve these problems.
“If you're not paying workers when the coin is down, then the whole system falls apart,” says Kaal. He says that blockchain companies like Gems and Storm are also corruptible because they’re based on the “Web of Trust,” a 30-year-old idea where people trust others’ ratings to create your own reputation scores.
“But if you can corrupt it, you will,” Kaal says. That’s because there’s a huge incentive to create several of what Kaal terms “sockpuppet” accounts.
If a fraudster opens a bunch of sockpuppet accounts, they can quickly upvote themselves to the higher-paying positions and behave well for a while. Then, before the system catches on to their trick, the fraudsters make several lucrative transactions between the high-rated accounts and pockets the cash.
Kaal claims—somewhat unsurprisingly—that his own research project, Semada, is the only reputation-verification platform that’s truly incorruptible. Kaal says that his platform could help microtask platforms prevent malicious actors from gaming the system.
First, Semada has developed a low-volatility currency (SEM) that’s always valued at around a dollar. If microtask platforms build with Semada, the idea is that their stablecoin disincentivizes corruption from shady business tycoons trying to exploit already cheap labor by refusing to pay out until the value of its coin is at a low, or fraudsters trying to game the system with fake accounts.
Second, Kaal claims that Semada is fully resistant to sockpuppet accounts because it incentivizes legitimate reputation scores and stops malicious actors from gaming the system. That’s because, alongside the stablecoin, Semada issues separate, non-fungible reputation tokens which are dished out to the workers once tasks are complete. Requesters don’t pay each worker for their work individually, but instead pay fees to the platform. When work is complete, these fees are distributed according to each worker’s reputation.
The more reputation tokens you have, the better your reward.
For instance, imagine that a research scientist wants a bunch of workers to caption some memes. If the scientist uses Semada, they’d pay the fees to a central pool, and then, once the tasks are complete, the platform would distributes those fees in proportion to the reputation a worker might have for captioning memes.
The more reputable the meme captioning market, the healthier the market. And because the meme captioners are then working for reputation tokens, not fees, there’s no incentive to game the system. The fraudster with five different accounts, who holds one reputation token on each account, receives the same amount as one person holding five tokens in a single account.
Because the system can’t be defrauded in the same way as companies like Gems or Storm, requesters can also be sure that the reputation of workers is correct; that those rated as highly skilled workers are indeed capable of accurately carrying out tasks. All this means is that where Gems and Storm might require 10 workers to verify a task—and Amazon’s Mturk 15—Semada could reduce that number to two or three.
Tapping the untapped
Semada’s technology has already interested huge conglomerates hoping to move into the space. Currently, Kaal’s advising Vodi—a social messaging app that has over 4 million users, largely from developing economies in South-East Asia and India—on how to create a decentralized microwork platform.
If VodiX comes through, blockchain could increase the global reach of microtask economy by tens of millions
In 2016, Vodi started to allow users to complete simple tasks, like watching videos or downloading games, to earn credits that could be redeemed for gift cards or mobile credit top ups.
“People realized they can earn coins, and now they want more coins,” says CEO Darren Lu in an interview with Decrypt.
“They want more jobs.”
Vodi’s begun work on a decentralized microtask platform called VodiX, but their existing user-base on Vodi is what’s impressive: there’s already 4 million users who currently know how to make money using Vodi. And Vodi just signed a deal with Bittel, a large Indian telecoms network, which plans to onboard some 100 million customers onto Vodi’s products.
Semada hasn’t officially partnered with VodiX—Kaal’s only advising them on the product. But both Kaal and Lu are open to future collaboration between VodiX and Semada. Nor is Semada yet fully operational: they’ve built their core, but they don’t plan to put the blockchain up until April at the earliest, after which they’ll list the stablecoin. But if all goes well for Vodi, their microtask platform could go live as soon as October next year.
Overcoming the overlords
If VodiX comes through, blockchain could increase the global reach of microtask economy by tens of millions.
And if they manage to finally deliver on those promises left unkept by the first wave of blockchain microtask platforms—that workers could earn a higher wage, tasks could become easier to verify, that millions of the world’s unbanked could be uploaded onto the digital economy, that the middleman could be cut-out—then it’s time to ready yourself for the grubby work of the Pusheen economy.