Milestone 1 ams

From crowdresearch
Revision as of 00:59, 5 March 2015 by Maniksingh (Talk | contribs)

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

Experience the life of a Worker on Mechanical Turk

1. The life of a worker is not inspiring as the contribution of the worker is not appreciated.

2. The primary motivation is money, which is valid, as a worker is looking at increasing his/her average earning

3. The HITs listed are not curated or ordered according to the skill sets of the worker. Thus, the worker has to go through a lot of steps assessing whether a certain HIT will be successfully completed. The time and money trade off in this assessment test is significant.

4. There should be a significant rating system so that on the HITs search listing, there is a rating given along with the requester so that a worker can also immediately judge the reliability of the requester, in terms of payment.

5. Also, a suggestive dialog could indicate the average earning per time period, so that the worker can estimate the amount of money he/she will earn in the following X hours

Experience the life of a Requester on Mechanical Turk

1. The provision for embedding any kind of HTML/CSS input, and building your own template, in the question box was impressive. This allows the requester to be a lot more flexible in the kind of work he/she wants to provide to the worker.

2. There should be better classification of a HIT so that it is easier for the requester to target workers who are skilled and interested in that particular HIT.

3. The payment options were quite elaborate, and the HIT building wizard ensures that information about a HIT is detailed.

Explore alternative crowd-labor markets


1. Quite simplified in comparison to Amazon Turks.

2. The purpose and skill sets required of each task is well defined

3. The asynchronous payment method is still highly unreliable, although the rating system helps in weeding out clients whose past has been unreliable in terms of payment.

4. Allows payment by milestone, so that a large task is divided into small steps, and a partial payment is done at each step to keep the worker motivated and ensures the reliability of the client.




1. The jobs are to be completed on a low end mobile device; hence the work could be carried out anywhere anytime and by almost anyone with basic knowledge of the language.

2. The job targets low income households and provides them with a decent source of income without demanding too much from them.

3. The answers submitted by users are verified by a minimum of two users which eliminates all or most chances of error. Hence the high accuracy of 89%.


1. The OCR tasks are sourced via an online channel. They could be passed on to one person with mobile internet facility, who would then pass on the tasks to others offline (using Bluetooth for example) to others who cannot afford internet on their mobile phones.

2. Longer words must be verified by more than two users to ensure higher accuracy.

3. The errors could be looked upon again. And in case there is a trend in the mistakes being committed, it could be rectified in the next versions of the application.



1. The referral system is an outstanding method to bring in more number of people into the system.

2. The language verification by the Google transliteration solved a major conflict of different spellings for the same word of another language when converted to English.

3. The application targets low income workers which helps the job being done conveniently, efficiently and at a low price.

4. Images via SMS ensures/enables the tasks reach more number of workers.


1. The referral system could be extended to an offline mechanism. This would further strengthen the entire process and reach a wider audience.

2. Longer words should be verified by more number of users/workers to ensure better accuracy.

Flash Teams


1. Heuristic evaluation might lead to a series of complex paths that are not advisable to follow and will end up giving a huge population of solutions. In that particular case we try to improve the solutions as a whole rather than concentrating on one particular solution. Since Heuristics are mainly probability based, a vast expanse of solutions which is not feasible might feature in the entire range.

2. A balance between intensification and exploration should be maintained.