FTW Milestone 6

From crowdresearch
Jump to: navigation, search

Hourly Rate Estimation

An accurate estimation of the number of hours completing a task would take and the calculated hourly pay based on this information can be a feature that distinguishes Daemo from other crowdsourcing platforms. This feature would not only bring transparency to the platform but also it would give the worker a better judgment of his/her earnings based on the hours he/she needs to put in.


Model

N = {1,2,..., n} are workers who estimate the task O = {o>=0} are the possible value estimates given by N

Each worker in N has a private preference for the value estimates e, whereby the preferences might be $7 for 3 hours, $6 for 4 hours and $5.50 for 5 hours. ei fully represents the preferences of the worker. For example,

ui(3, ei) = 7 ui(4, ei) = 6 ui(5, ei) = 5.5

ei is the same for all i and is {êi,ëi,éi...} where utility values are known. All types are independent from the other types (silly examples: volunteer, gamer, full-time worker, part-time worker).

Plurality Estimating Mechanism Given that A = A1 X… An, where Ai = {ei…},

M: A → Rho (O) where the value estimate is the one provided by the most candidates.

M(o1,o1,o3) = o1

In case of tie, M is chosen at random. The number of workers participating in estimates are always odd.

After types are identified, then a N can be asked their type and serve as default for automated value estimates and task-worker fit.

Dictator Mechanisms (to prevent gaming) Since the value estimates of the worker are two domains (monetary value on y, time estimate on x), a couple of options may be applied including median estimates of circular values, take the max of peaks, or the left/upper/lower/right edges of the circle. Additionally, requestors may suggest a fixed value of time before payment or declare levels at which they are willing to engage workers in tasks. At any given time any of these methods may be executed to override the plurality mechanism and thwart gaming attempts. Additionally, loosing or tightening the variation acceptance offers a way to further shift values fairly.


In This Model

While uploading a task the requester is asked to report a maximum estimation of the task completion time. Workers are encouraged to report the duration it took them to complete the task. The workers’ completion time would then be calculated by estimating the number of the hours the first five workers who complete the task report. Followingly the system proceeds to drawing an average and estimates the actual task completion hours. As new workers complete the task and report the hours he/she spent on it, the average hours are calculated (Trimmed mean) and updated in realtime. There is a filter to sort the task feed by 1)The Total Earning or 2)Hourly Rate Value.

HourlyRateEstimationModel.jpg

Incentive

The requester If the number of hours that the requester initially estimated, equals to the average duration that the workers report, the requester receives an Honesty Badge.

The worker -Reporting the hours is optional and it's not mandatory. -The first five workers to report the number of hours the task completion took them would receive points. -The report process is simple and inviting. It encourages the workers to participate and be a part of the community. This is the same approach that Netflix takes in asking a user to rate a movie after watching it. It appeals to the worker because by participating they become a part of the community.


The point system

Each one hour spent on a task completion equals to one point. If a worker is among the first five to report he/she will get 10 bonus points. These points would then accumulate. Collecting a 1000 points either reduces Daemo’s commission fee, or will result in more earning like a dollar added to the worker’s account.


Solution for Overreporting and Underreporting

The workers who report the lowest and the highest hours and fall in the trimmed zone (the system can also use Standard Deviation to set up a range) would get notified by receiving a friendly message. The intent of this message is to create a learning experience by letting the worker know that whether they took the longest or the shortest time to accomplish this task.


Estimation Methods

Trimmed mean: A trimmed mean is a method of averaging that removes a small percentage of the largest and smallest values before calculating the mean. Standard deviation: Standard Deviation is a measure of how spread out numbers are. Its symbol is σ (the greek letter sigma) and the formula is the square root of the variance.


contributors to this page

@reneelin @anotherhuman @mahsa