Crowdsourcing global development: working theoriesMay 18th, 2011 by Ville Miettinen
Take the much quoted statistic that 33% of workers on Mechanical Turk are located in India. It sounds impressive, right? Thousands of people in a developing country using Mechanical Turk to earn some much needed extra rupees. But according to a recent study by Microsoft Research India, the vast majority of Indian Turkers are college graduates with above-average household incomes. In other words, mainly middle class kids.
We like to think of paid crowdsourcing as a truly “equal opportunities” phenomena. A global workplace where anyone anywhere can get a job. But is this really true outside developed, computer-literate countries? Has crowdsourcing lived up to its potential when it comes to employing the world’s poor?
Lost in translation
The researchers at Microsoft Research India conducted an experiment. They selected a group of Indians with lower incomes, high-school standard education and only basic English and computer skills. After researchers explained the basic concept of Mechanical Turk (it’s not mechanical, there are no Turks), the participants were set a range of simple tasks. The results make uncomfortable reading.
To cut a long study short, every participant in the group failed to complete the tasks. Why? Mainly because people couldn’t navigate Mechanical Turk or follow requestors’ “ad-hoc and complex” instructions. The researchers responded by simplifying the interface and translating instructions into the local language. Result: task completion was boosted from 0 to 66%.
Now, I know Mechanical Turk requestors are an intelligent bunch, but surely it’s a bit extreme to expect them to translate every task into multiple languages – India alone has hundreds of languages and dialects. Instead, the researchers suggest that crowdsourcing platforms create “standardized human instruction sets” (you can tell these are programmers): individual, pre-written, simple English sentences that requestors use to write tasks.
Of course, even if the instructions are crystal clear and the platform is a miracle of human-centered design, there’s still the problem of access. Only 6% of Indian households have a PC and internet connection. Some African countries are still waiting for broadband to arrive. Internet cafés are expensive, plus many people in the developing world don’t have bank accounts.
Hatching an industry
The authors of the study state that “paid crowdsourcing has the potential to improve earnings and livelihoods in poor communities around the world.” But actually creating a paid crowdsourcing industry in developing countries is clearly a mammoth task. It requires connecting more people to the internet, redesigning crowd labor platforms, improving computer skills and adapting more tasks for mobile phones. Will commercial crowdsourcing companies be willing (or even able) to invest that kind of time and money?
Crowdsourcing has been described as an “egg before chicken” industry (stay with me here, I promise it will make sense). You need the workers (or eggs) in place in order to attract clients (er, chickens). Unlocking the potential of the millions of workers in the developing world could accelerate the growth of the whole paid crowdsourcing industry. Or as the study puts it: “requesters could find lower-cost labor, platform owners could benefit from larger volumes, and the crowdsourcing paradigm as a whole could scale up to address much grander problems”. It sounds like a bright future, now all we have to do is figure out how to make it happen.