Milestone 3 AltaMira PowerIdea 1: Secondary markets

From crowdresearch
Jump to: navigation, search


Empower workers to have a bigger stake/share in the system, give them a clear path to promotion/growth. Get work done for requesters.


Secondary markets - request for requesters.


What requesters want from MTurk is a solution, quality work that they are willing to pay for. As with all real world systems, this means that managing thousands of workers is not ideal or scalable. In the real world, this is done by having several levels of managers (Oh no!!!) but it accomplishes the problem of not having to manage people individually. Imagine in Larry Page had to manage every Googler, it wouldn't make sense but this is exactly what happens on MTurk. Every requester has to manage all workers and usually they dont, based on the feedback we received from them.

They also cannot comb through all of the data and ensure that results are accurate and quality is delivered. Secondary markets solve this problem. Requesters post a problem and attach quality scores and accuracy rates. Workers can take up the task and hire other workers to do the job. They would then be responsible for ensuring that the quality is delivered. Then the main requester can glance through the data to ensure it meets the standard of quality.

This process ships the work out to people that know workers best - other workers. It empowers workers and gives them a path to promotion and ability to grow with MTurk. There could be limits that say only after 6 months and 1000 tasks on MTurk can you become a manager. These constraints make it possible to imitate real life and give workers more control over the system while giving requesters an easier path to getting work done.


This has a couple of drawbacks, mainly it costs more money since now you have several levels of management. There has to be a fair rating system for workers who act as managers because their incentives may not be aligned. They might want to give work only to their friends, instead of people that should be taking up the task. While these are valid considerations, the net benefit of not having to verify every worker and work outweighs the risks from having bad managers because for every manager you will have 10 or more workers (which could be implemented as a rule).

The flow could look something like this:

Hire requesters.png