WinterMilestone 1 Team Vanilla

From crowdresearch
Revision as of 11:10, 18 January 2016 by Namitjuneja (Talk | contribs) (Daemo)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Experience the life of a Worker on Mechanical Turk

Positives

  • Clear identification of the level of worker to ensure better quality of output
  • The dashboard is elaborate and helpful in tracking progress

Negatives

  • The interface is not very intuitive or motivating
  • There is no clear classification of hits

Experience the life of a Requester on Mechanical Turk

Positives

  • Number of workers per task can be easily mentioned
  • Has a requester api which is one more step towards reducing manual working.( But again supports limited programming languages)
  • Batches progress page is helpful in terms of keeping track of progress

Negatives

  • Type of tasks to be given out is very limited in terms of type of tasks

Research Engineering (Test Flight)

Successfully deployed Daemo and learnt more about the implementation.

Readings

MobileWorks

India is a big source of crowd resource but what is difficult is having a mechanism which shows maximum efficiency in tapping this potential. MobileWorks makes an attempt to utilize as much of the unused work resources as it can.

Observations

  • Each task is divided into multiple smaller pieces
  • The correct solution is decided by comparing the solutions of two workers
  • Pay is a function of quality of work and the task


Strengths / Positives

  • Source of additional income.
  • A more populated country means cheaper resources, hence trivial jobs can be done for cheaper rates.


Negatives

  • Most of the tasks are generated are english based but english literacy is low especially among st the target population (i.e workers at the bottom of the pyramaid)
  • The type of data delivered is limited and also restricted as the low end browsers have limited support
  • Non - continuous data connectivity (e.g for a worker moving in a bus) can mean wastage of resources.


Proposed Solutions / Improvements

  • Tasks can be downloaded in batches as it solves two problems
    • Continuous data consumption is expensive as the mobile browser is not the only app utilizing the data.
    • Consistent connectivity is not possible with the current infrastructure available


Daemo

Strengths / Positive Observations

  • A lot of times information flows from requester to doer (in terms of instructions) and doer to requester ( in terms of work done) but very few platforms lay emphasis on the 3rd communication which is feedback as it creates valuable data to further keep the model relevant further.
Daemo's focus on that (Boomerang) is very important.
  • Prototype tasks eliminates the need of trusting an individual based on his/her ratings/qualifications.

Flash Teams

Strengths / Positive Observations

  • It organizes experts with minimum manual effort (creating structured collaborations).
  • The mechanism / model needs to remain
    • Generic
    • Scalable
    • Replicable


Negatives

  • This model isolates one team from another and hence the complexity of the task has a glass ceiling since more complex problems require more inter communication between different teams of different domains.
  • Time synchronization can be a problem
    • between two modular teams
    • within a modular team ( different time zones)


Possible Improvements

  • Just by investing more resources one can have >1 modular teams working on the same task and at the end the the team with a better output can be chosen by the admin to be pushed forward through the pipeline.
This is analogous to the MobileWorks system but instead of comparing outputs of two different workers to check the correctness, this compares different solutions to the same task as most of the jobs are subjective in this scenario and can have better solutions.

Milestone Contributors

@namitjuneja