From crowdresearch
Revision as of 10:22, 14 February 2016 by Nalinchhibber (Talk | contribs)

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

This page is intended to serve for Winter Milestone 5 submission by user @nalinc[1].

Do not edit this directly


Name: nalinc, Type: individual, Category: Task-Feed


One of the most challenging problem in today's crowdsourcing marketplace is to let workers reach the right requesters and vice versa. However the prime motive should be not just to connect the right worker-requester pair, but also to enable workers access right set of tasks that fit their skills or domain of expertise. The premise behind this statement is:

Workers are humans, just like us and their preferences change:

  • Workers want to choose tasks that pay more.
  • At-times, they target tasks that are less in number(or takes lesser time) and pay decent.
  • Sometimes workers want to work on tasks they are good at and not just earn money doing something they dont find interesting.

Critical problem(s) in the area

Workers demand the option to choose tasks based on their preferences. As an example, a worker from literature background, would like to work on audio/video transcription or maybe tasks that demand summarizing a material(paper/video) whereas someone with an inclination towards creativity would find 'image-tagging' pretty interesting. A single themed task-feed is pretty dull and less motivating. An ideal crowdsourcing platform tends to address different concerns of workers(make money, minimize rejection, maximize hourly wage, or just to have fun), and a monotonous task-feed is insufficient to cover all such edge-cases. Moreover, workers periodically change their priorities to choose HITs. There are times when they would prefer choosing high paying tasks(or tasks that maximize their hourly wage), but there are times when they prefer choosing tasks that don't need serious attentiveness. Some example of tasks that do/don't require intense focus are as follows:

Time-intensive tasks

  • audio/video transcription, where one need to listen/watch the whole N minutes(sometimes hours) audio/video.
  • Summarizing a video or research paper or article ?
  • Tasks that require one to open a link(of-course in a new browser tab), do 'X' task and verify if 'Y' works properly or not.

Not-so time-intensive tasks

  • Image tagging*
  • Sentiment analysis*
  • Classification tasks to choose the odd one out, or identify 'X'*

(provided it doesn't require additional work and text is provided inline)

Suggestive improvement

Allow requesters to 'tag' their tasks while authoring, so worker can categorize their task-feed based on task-type. This should be implemented keeping the boomerang model in mind. So that, if I filter the task-feed by 'image-tagging' HITs, I should see the HITs with specific tag AND the list should follow the 'favorite-first' ordering i.e. in decreasing order of rating I provided to requesters.

Few metrics for filtering the task-feed are as follows:

  • Task Category: transcription, content writing, object classification, image tagging, website feedback, content generation.
  • Performance factors: avg hourly wage, avg time completion.


Workers are the soul of any crowdsourcing platform and it is important to engage them and pamper their needs. We need to enable workers to filter their taskfeed allowing them to choose tasks that they can perform while doing secondary work and even on low-end devices(mobile phones/tablets). This would also allow them to choose tasks they're actually interested in. As a worker, I would be more interested to see tasks that matters to me and not just ones by specific requesters. Filtering the task-feed must not compromise the benefits of boomerang.

Milestone Contributors

Author: Nalin Chhibber @nalinc