WinterMilestone 1 WilliamDai

From crowdresearch
Jump to: navigation, search

The life of a Worker on Mechanical Turk

Feels awfully tedious. Extremely frustrating to know that all the work I put in might not be accepted, or I might not be recompensed. Sometimes encounter vague/confusing instructions.

The life of a Requester on Mechanical Turk

Tried out the Sandbox. Found it fairly straight-forward to create projects. Wasn't sure how to merge different actions i.e. Categorization, Tagging of Images, Transcription from an Image ... into a single project.



  • What do you like about the system / what are its strengths?

- Customized to the circumstances unique to India: ubiquity of mobile devices, access to internet coupled with lack of widespread desktop platforms - Positive social impact for those in difficult economic conditions - Worker ratings based on accuracy in an Optical Character Recognition system makes sense; reputation is directly linked to quality of work

  • What do you think can be improved about the system?

- System looks usable, but not too polished/commercial (aesthetics lacking, minor issue) - Questionable - should simple repetitive tasks provide the same level of pay as perhaps more engaging, typical work? Of course, if MobileWorks becomes the sole source of income, then income equity is a necessity - but as noted by study participants, the OCR work can be done nearly anywhere, anytime. In this case it is supplementary income. The purpose of the system needs further defining.


  • What do you like about the system / what are its strengths?

- Address problems with crowdsourcing from the start, improving efficiency and productivity, instead of sorting out poor quality work with filters. - Acknowledges prevalent reputation inflation. Merely sorting and partnering highly ranked workers with highly ranked requesters is not sufficient. [introduce Boomerang system] - Takes into account discrepancies between desired results and stated benchmarks. [prototyping tasks reconcile desires and needs of requesters with the impressions workers get from project descriptions]

  • What do you think can be improved about the system?

- Needing to prototype tasks incurs additional costs for requesters. Perhaps a necessary sacrifice for higher quality work. - From worker feedback, highly suggest requesters include an example of what they want. i.e. "Put the handle" should also include "The handle looks like @helloworld" to avoid all worker confusion. - As noted, it is not possible to force workers and requesters to rate individuals, especially under high-volume conditions.

Flash Teams

  • What do you like about the system / what are its strengths?

- The idea of having teams that can dynamically assemble as needed sounds great for productivity and targeted tasks - Such teams will allow crowdsourcing to address more intricate, less trivial/repetitive tasks. Who knows, maybe crowdsourcing can create the next crowdsourcing platform. But that's what we're doing, isn't it. :simplesmile: - Acknowledges that most people are neither managers nor natural programmers - Combines the best of MTurk and Elance or Fiverr - both small tasks and large projects like websites more typical of individual freelancers can be undertaken by the crowd - Can result in the rise of specialized workers who only focus on specific elements of backend or frontend, depending on the degree of modularity in the tasks

  • What do you think can be improved about the system?

- In handing off modular sections, great documentation will probably be essential. Creating such (internal) documentation might take greater time than otherwise needed. - There will need to be great coordination and a solid overall understanding. It will be pretty difficult for individual requesters of the common sort to provide this level of detail, and even more so if different teams are to know only the specifics of their part (to maintain secrecy of the overall project).

Milestone Contributor(s)