MobileWorks: Designing a Crowdsourcing Platform That Works (Against Poverty)

Crowdsourcing platforms like Mechanical Turk allow businesses to solve problems that cannot be easily solved using computers. These platforms act as a marketplace for online workers who complete small tasks of several seconds and earn small amounts of money. While these services allow businesses to automate hard problems, it provides a population of people easy access to online labor - work which they might not have access to otherwise giving them an alternative source of income. However, crowdsourcing market-places like Mechanical Turk are also very prone to several problems. First, a significant amount of online tasks are spam tasks asking workers to "like" their website or write a positive comment about their service. Second, answers of workers are often incorrect which is due to intentional mistakes or misunderstanding of instructions. Third, projects often do not complete because they appear in the search results of online workers order by time and the older a project the less it is recognized by workers just like the 2nd page of a Google search. Fourth, workers in these platforms usually receive an hourly payment that is well below market rates and goes as low as 0.75$/h which incentivises them to maximize their hourly wage by answering tasks as quickly as possible which means they answer nonsense.

There are several approaches to these challenges. On the one side, several algorithms have been proposed to improve quality and performance but most of them require the design complicated work-flows which most businesses are not interested engaging in. On the other side, businesses have started approving all answers that workers provide to reduce the incentives to answer nonsense. This, however, incentivies these workers even more to continue giving incorrect answers.

We present MobileWorks, a new micro-work crowdsourcing platform that moves away from the traditional market-place model to service based model that alliviates the burden of businesses and ensures workers a fair payment. As an intermediary, we provide a crowdsourcing service which gaurantees quality, fixed turn-around times and also pays fair wage to our workers. We do this using a combination of statistical and social computing approaches. Each piece of work is done by a number of workers. We model the answers of different people and select the statistically probable answers. We also provide workers with a platform to collaborate and a platform for experienced workers to help the novice workers. We believe that such approach makes crowdsourcing more usable for everyone while providing better quality of life to the workers.