Milestone 4 pentagram Task Clarity: Coherent Tasks

From crowdresearch
Jump to: navigation, search

Problem Statement

Turkers spend a lot of time in finding suitable tasks. Either the tasks do not have clear instructions as to what is to be done, the payment for the task is too low or they may not find the tasks that interest them or applicable to their skills. The requesters perhaps do not understand clearly the tasks to be done themselves, or are unable to get the workers' perspective on the tasks.


This problem can be solved by creating a task-review system(for evaluation of clarity of tasks), a recommender system for tasks and possibly an artificially intelligent worker to automate the process in the future.

Description of solution

Flow Diagram

Coherent tasks.png


  • A set of workers ascertain if the task set by a requestor has clearly defined instructions, constraints and requirements. In return for rendering this service, these workers will get the first opportunity to take up that task, if it is certified to be lucid. If not, the requestor needs to redesign it based on their feedback and send it for review. However, once a worker takes up the task, he/she will have to complete the task as per the requestor's satisfaction otherwise he/she will not be paid.
  • The workers get points for every good review. A review can be considered good if their review agrees with the consensus review. These points will help them gain precedence as a professional turker and get more tasks to perform. The requestor also gets points for posting clear tasks. This can be implemented as a priority system.
  • A recommender system can be developed to suggest tasks for workers based on their interests and skills. It can also suggest the requestors qualified workers who have completed similar tasks in the past and the workers who the requester had been satisfied with.
  • Workers can form groups and perform the above tasks of reviewing the jobs, completing and validating them. Groups create more accountability, teamwork, dedication and bonding.
  • An algorithm can be developed to learn the skill of deciding if the task set is clear enough or not, based on the workers' reviews, which can act as the training set. After testing and validation, this algorithm can replace the workers in checking the task clarity.

Ideas that inspired our idea

  • Team Betzy's Dark horse idea was great!
  • Our own idea of Peer-Workers Review System, Priority System and Recommender system
  • Team Pumas' worker groups idea was good too!