Microwork is a series of many small tasks which together comprise a large unified project, and it is completed by many people over the Internet. Microwork is considered the smallest unit of work in a virtual assembly line. It is most often used to describe tasks for which no efficient algorithm has been devised, and require human intelligence to complete reliably. The term was developed in 2008 by Leila Chirayath Janah of Samasource.
Microtasking
Microtasking is the process of splitting a large job into small tasks that can be distributed, over the Internet, to many people. Since the inception of microwork, many online services have been developed that specialize in different types of microtasking. Most of them rely on a large, voluntary workforce composed of Internet users from around the world.
Typical tasks offered are repetitive but not so simple that they can be automated. Good candidates for microtasks have the following characteristics: They are large volume tasks They can be broken down into tasks that are done independently They require human judgement
It may also be known as ubiquitous human computing or human-based computation when focused on computational tasks that are too complex for distributed computing.
Microtasks are distinguished from macrotasks, which typically can be done independently. They require a fixed amount of time and they require a specialized skill.
The wage paid can range from a few cents per task to hundreds of dollars per project.
Examples
Toloka and Amazon Mechanical Turk are examples of micro work markets, and they allow workers to choose and perform simple tasks online, reporting directly through the platform to receive payments in exchange. A task can be as complex as algorithm writing or as simple as labelling photos or videos, describing products, or transcribing scanned documents. Employers submit tasks and set their own payments, which are often pennies for each task. This crowdsourcing project was initiated by Amazon as a way for users to find duplicate webpages, and soon it became a service for individuals to contract computer programmers and other individuals to finish tasks that computers are unable to accomplish. Since then this project has expanded from its original form; nowadays, there are people who will complete various Mechanical Turk projects as extra income on the side.
LiveOps uses a distributed network of people to run a "Cloud Call Center", which is a virtual call center. Contracted workers can answer calls and provide other call center facilities without the need for the physical building or equipment of a traditional call center. The Red Cross used this system successfully during Hurricane Katrina in 2005, to process 17,000+ calls without having to open or hire staff for a call center.
A similar model is used by text message question-answering services like 63336. Researchers connect with the service at home and receive questions as they are submitted.
InnoCentive allows businesses to post problems and offer payment for answers. These questions are often far less simple than tasks posted on services like Mechanical Turk, and the payments are accordingly higher. For example: "Think you can find a way to prevent orange juice stored in see-through bottles from turning brown? There may be $20,000 in it for you."
Samasource is a non-profit organization that allows people living in poverty the opportunity to complete microwork for a living wage. The service specializes in online content moderation, digital transcription, and data gathering and promotion.
Galaxy Zoo is a scientific effort to use online crowdsourcing to classify a very large number of galaxies from astronomical images.
In 2010, the company Internet Eyes launched a service where in return for a potential reward, home viewers would watch live CCTV streams and alert shop owners of potential theft in progress.
Uses
Most uses of microtasking services involve processing data, especially online. These include driving traffic to websites, gathering data like email addresses, and labelling or tagging data online. They are also used to accurately translate or transcribe audio clips and pictures, since these are activities that are better suited to humans than computers. These are used both for practical data conversion purposes, but also to improve upon and test the fidelity of machine learning algorithms. Identification of pictures by humans has been used to help in missing persons searches, though to little effect.