Milestone 3

From crowdresearch
Revision as of 11:28, 14 March 2015 by Geza (Talk | contribs) (Initial Brainstorm)

Jump to: navigation, search

This week, we will take the set of needs that we collectively identified in the previous milestone and use those insights to generate design ideas.

Needs from Milestone 2

We synthesized the main needs groups identified in Milestone 2 in the following table:

Worker Needs

Needs Evidence
Observations Interpretations
Workers need to be able to quickly find tasks they'd want to work on Monday evening panel workers cite the challenge in identifying or distinguishing tasks because of poor tagging, Reddit discussion also cited the exorbitant amount of time that they spend trying to find tasks and do the mental calculations to find the opportunities that match them best (e.g. time to complete the task on average, average $ per minute on task, requirements to complete the task). Finding a task that matches well with the worker's skillset and pays well takes a significant amount of (unpaid) time.
Workers need to feel they are being fairly compensated for their work. Reddit discussion cites that the payment system for HITs is not adaptive and does not take into account changing marketplace conditions (supply/demand) and pricing of tasks based on those changes. Monetary compensation is the primary motivator for many crowd workers.
Workers need to feel like they are treated fairly and respectfully, and have a voice in the platform Comment on Turkopticon: "Got a mass rejection from some hits I did for them! Talked to other turkers that I know in real life and the same thing happened to them. There rejection comments are also really demeaning. Definitely avoid!" Unreasonable rejections and low payments lead workers to feel disrespected. The implicit assumption on MTurk is that workers are unskilled and replaceable. They can do little if their work is rejected.
Workers need to be able to expose their skills so they can get work they are qualified for and advance their skills Monday evening panel workers from oDesk cite that most employers will not work with them until they have enough feedback or past work on the platform If users cannot get new work without feedback, this makes it difficult for new users to establish their reputation and get jobs that will help develop their skillsets.

Requester Needs

Needs Evidence
Observations Interpretations
Requesters need to get their HITs completed (quickly / correctly) Requester asking on forum why nobody is doing his HITs (7-minute, 25-cent surveys - a very low wage) Requesters want their HITs done, and when nobody's doing them, they do not know the reason why (e.g. it is because he is underpaying workers)
Requesters need to be able to trust the results they get Requesters will often rely on previous workers whose results they can trust, and add mechanisms to detect spammers, or manually verify some results. If spammers are not caught, this brings the correctness of results into question. If requesters are not sure the results are correct, they may need to discard the data.
Requesters need to have workers who have the appropriate skills and demographics do their tasks Requesters worry that they are not able to verify self-reported demographics for surveys. Workers' self-reported skills and demographics are often not viewed as trustworthy. This is a problem for surveys, which need to have correct demographic data to be useful.
Requesters need to be able to easily generate good tasks Companies hire full-time developers to deal with the complexities of posting microtasks on MTurk. Requesters often develop their own tools and workflow systems on top of Amazon's. The process of authoring HITs is currently difficult and makes crowd-work inaccessible to potential requesters
Requesters need to price their tasks appropriately Requesters asking on forums about the appropriate amount they should pay for their HITs Requesters often don't have a good intuition of what the appropriate wage for their task would be in terms of price per HIT.
Requesters need workers to trust them Requesters say they are reluctant to reject work, because they fear they might get bad reviews. Workers are more likely to do HITs if the requester seems trustworthy. Requesters do not want bad reviews, because they may result in workers ignoring the requester's HITs

Michael Bernstein's synthesis

These needs boil down to two main issues: 1) trust, and 2) power.


  • How do I trust who you say you are?
  • How do I trust that the results I get are results that will be good?
  • How do I trust that you’ll respect me as a worker, and pay me accordingly?


  • Who has the power to post work?
  • To edit other peoples’ posted work?
  • To return results to the requester? Can I, as a worker, send it back myself, or does someone else need to vet it?

As we brainstorm, we should be thinking about solutions that holistically address these issues of power and trust, not just surface fixes that get at micro-elements of the system.

Optional Readings

This week's readings are optional and don't have a deliverable. They are meant to help you advance as a researcher and get a better sense of ideas and visions that other researchers have thought of

Nikki Kittur the future of Crowd Work

This a long paper. Feel free to focus on just the parts that particularly interest you.

Initial Brainstorm

Now it's time for your team to brainstorm some ideas based on these needs.

Work with your team to brainstorm as many ideas as you can under two headings: trust and power. Here are some examples of " How might we" questions (a technique which can inspire specific brainstorms) which can drive the generation of your ideas.

  • How might we” enable workers to trust the requester’s intention to pay?”
  • “How might we” enable requesters to trust the results they get back?

Use whatever tools you need - if you can get together in-person, whiteboards and sketchbooks are great tools, while if you're a remote team, services like Google Docs and should help you with the brainstorming process.

Sketch out enough ideas until you find a set that you’re inspired to explore further. It should be at least 10 ideas for each of worker and requester. This brainstorm should be wild and broad. Focus not on usability patches, but deeper design innovations.


The 10 or more ideas you brainstormed, (note: it's 10 ideas for trust, and 10 ideas for power). Provide them in whatever format you want - diagrams, sketches, descriptions, or a combination (the wiki supports images, see here for instructions on uploading them).

Dive Deeper into 2 ideas

Dark Horse idea

The term "dark horse" means a

Optional Readings

We're including a few optional papers to help train you as a researcher