WinterMilestone 3 nalinc ReputationIdea:SynchrOnous Network of Advance Reputation system

From crowdresearch
Jump to: navigation, search

Reputation-related Idea

SONAR: [SynchrOnous Network of Advance Reputation system.]

The title might sound a bit icky, but this proposal is an attempt to address practical limitations of Boomerang: The reputation system we currently have in Daemo.

According to the current model, rating given by requester/worker to his counterpart, effects his own future experiences on platform. It uses ratings to determine who will appear at the top of the task feed or who will be given early access to future tasks. This method however, significantly ignores the users who are new to the platform and clueless regarding the quality of tasks/history of workers.

What if a new requester publish a HIT ?

Shouldn't he have privilege to set a filter(maybe with an additional operational cost) ? Filters could be based on geography of workers or % of accuracy. Setting a minor operational cost for each such filter based on number of Hits would enable him to ensure quality for important tasks.

What if a new worker joins the platform ?

If all tasks are available to a set of workers before anyone else(including new workers), then newbies might suffer *starvation* and *infinite waiting* since they might not be able to see most of the tasks. As the famous saying goes "something is better than nothing", we should atleast show the tasks to every user rather than not showing them at all. We can disable to "accept" buttons and show a timer saying "this task will be available in x seconds". While it solves the purpose of making new tasks available to only a set of workers, it also keeps the task feed populated. A rich tasks-feed with abundant tasks is important.

Problem:

  • While Boomerang does address the problem of inflated reputation, it still only affects individual users based on the feedback they get from their counterpart. It does not give a clue to sibling workers about how good a requester is. To better understand this issue, we can take the example of TurkOpticon which is a browser extension build entirely to address this issue- to let workers know how good a requester is.

Solution:

  • We should have a more intuitive system to rate requesters based on several metrics. The 3-scale rating offered by boomerang gives little to no clue about requesters quality of tasks, speed of payment, communication etc to other workers.

The question is: Why would a worker provide such feedback for requester instead of working on new HITS ?

The answer lies in implementing a simple badge system(much like stackoverflow). The more feedback a worker provides, more badges he earns. Such badges might be broadly classified into gold, silver and bronze categories. A badge system will not be used by requesters to set filter as it might have a negative impact and result in 'starvation' for new workers by limiting their chances to participate in new HITS. But this will allow workers to share few information mutually among themselves.


There are two types of reputation systems in practice these days: User-driven and Content-driven. Following image illustrates the difference.

ContentVsUserdrivenReputation.png

I believe we should have the best from both worlds. In SONAR(Synchronous Network of Advance Reputation), I propose rapid and immediate propagation of user feedback while at the same time, keep them easy to understand. Integrating such a system would also be aligned with future interests of Daemo community to include synchronous chat rather than static feedback.

Image courtesy[1]

Milestone Contributors

Nalin Chhibber @nalinc