Milestone 8 PixelPerfect 1

From crowdresearch
Jump to: navigation, search

Foundation 1: Micro + Macro Task Market

What would the tasks look like? How are they submitted? Does this look like AMT where any expert certified in an area can accept the task? Or like an oDesk negotiation?

The tasks should be given a certain template based on the task itself. The basic tasks can be classified as expert or non-skilled based on the required prerequisites. Non-skilled tasks would not require so much of a description as an expert task would. Expert tasks would require proper description. The suggestion about releasing the tasks to a certain crowd of high-rated workers does seem inviting but the new workers would probably never see the face of the task itself because high-level workers would start working on it beforehand. There could be equally skilled novice workers who would be at a disadvantage due to the system.

The tasks could be released to the worker community as a whole, where any expert certified in the area can apply for the task, but acceptance can be done following a process of negotiation, like that of ODesk. There can be an Upvote/Downvote button and Comments section. (like the Quora model) This can be something of a psychological trick where the community feels like it is on a social networking site but it is instead made to rate the given task template. Suggestions and Edits can also be enabled.

We can create standardized task templates for certain tasks. But the problem would be that there are just so many tasks to create a template for and that would a gargantuan task by itself. Some objective questions can be asked to the requesters while they post a new task to clear ambiguities.

How do we ensure high-quality results? Do you let an expert work for hours and submit? That seems risky. Should there be intermediate feedback mechanisms?

Letting an expert work for hours is indeed risky. The Flash Teams model sounds really interesting and the results were positive too, despite the fact that the platform was used for expert tasks. Such a model could also be incorporated into our platform, where the team structure is loose and floating. A large task can be broken down into smaller tasks with short, specified time durations. The requester can have an interface to guide and track the progress of the task. In case a worker needs to drop out, an alternate worker with similar ratings and expertise can be picked from the worker community in the platform. This makes sure the experts don't spend too much time on an expert task and the quality doesn't deteriorate. At each level of progress the requester can evaluate the work before passing the baton to the next level. To ease communication, we can enable comments from the workers and requesters involved in the task to bridge any ambiguities.

How do you trust that someone is an expert?

To judge an expert is an extremely difficult task. Some ways to do this are:

  • Making workers upload certificates/resumes.
  • Links to Github profile or something similar to showcase previous experience.
  • In-built tests on platform to determine expertise.
  • Requesters can themselves pay the workers a small fee to take the test to know the quality and expertise of the worker before hiring.
  • Requesters can hire a worker under a mentorship program by paying a smaller fee than he would if he were to hire a high rated worker. Initially this could be non-paid too. The worker learns and gains expertise by working with the mentor over a certain period, a certain level of trust and camaraderie is established, and the worker is also guaranteed tasks and a nominal fee for the certain period. It also becomes easy for the requester to find a trusted worker to work with. This is mutually beneficial. Further, recommendations from the mentors can be used to find new jobs too. This could also be extended to high-level workers. High-rated workers who get a lot of tasks can hire a novice worker as a mentee to help them with the tasks by paying a small fee.

A combination of all the five might seem like a redundant task but will ensure that experts are not unjustly chosen/rejected. All the novice workers can be categorized as non-skilled. If a worker wants to work in an expert task he can be made to apply by uploading certificates and adding details about previous experiences. The worker can then apply to the mentorship programs.