Milestone 1 Cumbre
Experience the life of a Worker on Mechanical Turk
No access to Mechanical turk.
Experience the life of a Requester on Mechanical Turk
Signed up from India, we were not allowed to work on Amazon Mechanical Turk.
Explore alternative crowd-labor markets
GALAXYZOO:- a crowdsource in astronomy
no signup option as a requester or a worker deals in space and science provides an api to give users another access point.
Volunteers make contributions to the GZ to help
Amazon mechanical turk:-
Signup options for both Provides an api to give users another access point. People submit there work as worker and request for hits as requestors.
Strengths of Framework:
The framework utilises mobile as a means of completing human optical character recognition (OCR) tasks and the penetration of mobile phones in the market is very high in comparison to pc’s and desktops [ In case of India, mobile penetration is 50% compared to the 0.9% penetration of computers ]. So the framework is reachable to a larger audience willing to work. The tasks small and easy. Most of the people can do the tasks. No high qualification needed for the tasks. The tasks are divided into multiple micro-tasks ( like digitising a single character or a word of a scanned document ). This gives a great flexibility to the workers. They do not need to devote a whole lot of time to do the tasks. Some given case studies shows that people consider this a great advantage. They can work on these tasks while watching tv or while travelling. Using Iterative method, very high accuracy can also be achieved. In developing nations, the wages workers earn by doing these tasks can easily be matched with their regular wages ( per hour basis ). The biggest strength is that people who were studied during the survey (as given in the research paper), they were happy to work and were ready to recommend the system to their peers.
The tasks are micro. This may lead to lower wages per task. Also the wages per hour cannot be matched with the per hour wages of a larger division of society. So the interested number of candidates will involve only a very small division of society.
Presented Framework and the operational system only provides a mechanism for completing OCR tasks. Other tasks can also be included which can provide a better pay scale per task. The Framework uses simple “candy-bar” kind of mobile phones for the task. Now given the growth in the mobile market and the rapidly increasing penetration rate of smartphones, the operational system can be upgraded to newer technologies which will allow us to include a lot of other tasks other than just OCR tasks and with the capabilities of smartphones, the range of tasks that can be done is pretty wide.
Strengths: Everything is operated using mobile phones. so, it is easy to interact with the workers living in very low working conditions where access to internet and computers are not that much frequent. anyone with sufficient technical skills and has a basic mobile phone with messaging function can earn money. Workers can frequently complete tasks using cell phone then computers, as restrictions to a cell phone is very less. Less costlier than internet services. Tasks are divided into micro-tasks ( like dividing a whole image into several smaller images containing single word ). so user doesn’t have to devote hours for completing it. Students from low financial background can generate 6K to 10K(INR) from these kind of tasks. Users who are not fluent with english language can respond in their local language which is then converted into english languages with 90.1% accuracy. If the have any queries the can just miss call to a number and clear their doubts. so, its not costing them even a penny. Users with free SMS packs can complete higher no of tasks per day.
Problems: User is paid after 20 responses. if worker gets impatient he may stop working. In india, according to TRAI rules a user can only send 200 messages per day, so it will be difficult for a user to earn considerable amount of money. if we send a fixed no of messages in a group to each user daily it will be a great difficulty if user doesn’t responds in the same sequence. so the user has to put a particular id before answer and that will be a great problem. In other case if we send tasks one by one user misses a response it ends up in halting users reply.
The idea behind Flash Teams is to combine crowdsourcing with organizational behaviour to form dynamic teams which can reflect the works of a fully functional, tightly coupled teams of experts. Crowdsourcing platforms that exists at present, focus on dividing a single task into micro-tasks to accomplish work reflecting the work of a single expert. Flash teams aims to combine experts and form a team in a way that this dynamic temporary team can reflect the work of a traditional organizational team of experts.
The key features of Flash Teams:
With current crowdsourcing platforms, there is no connection between the entities doing the micro works making it difficult to get an idea of a clear goal. By combining the experts as Flash teams, the larger goal becomes clearer.
As Flash teams are loosely coupled, they can connect and interact in variety of configurations which follows the management modularity theory.
Flash teams work as a system of interconnected Blocks. One Block can be defined as one or more experts performing a task.
Flash teams working mechanism includes pipelining. The output of one block works as an input for the next block of the Team.
Each Block is self-contained. The whole block can be connected to another team where the set of inputs and corresponding outputs match.
To avoid the risk of diffusion of responsibility, each block has a Directly Responsible Individual, or DRI which is assigned by the Project Requester. This person acts as the manager of the block.
The Blocks are connected via intermediate results and this provides a clear picture of the work done and the goal.
Flash teams can grow, adapt, and recombine into larger organizations.
It is easy for the requester to form Flash Teams using Foundry, which let user to create new teams based on previous teams and workflows.
Each block has to be recruited using a crowdsourcing marketplace like oDesk.
Through combination, users can create not just teams but the equivalent of entire small organizations for an afternoon or a day.
This model reflects the work of a whole team or organization, not only a single person’s work divided into micro-tasks. It provides a way to involve experts in the crowdsourcing works. Flash teams can complete a complex task which require a higher level of expertise in various fields. At every instance of time, the requester is aware of the progress and a clear goal of the project because of the intermediate results coming out of every block. Many parallel tasks can run simultaneously and can reduce the time factor. The size of the team can be changed as and when needed. It is a complete reflection of any organization for one day.
Problems and Improvements:
There is no social interaction between the people of a team or block. There is an assumption that workers can be recruited within a very small time interval (15 minutes as the paper states). This is not completely true in the practical environment. Recruiting right workers who are trustable, capable and motivated for the task will take longer time. This will increase the load on Requester. The most important factor here will be the difficulty in time management. It is very common to see delays in various tasks because of the unforeseen problems arising for the completion of the task. Because of the delay by one block, all the other blocks will have to change their schedule. Only notifying the members of the next block will not solve the issue for them. They might have some other engagements during that time. It cannot be assumed that workers will be available all the time.