08/31 - Striking Saturday!

dot dot dot, dash dash dash, dot dot dot

 

$8,987.00 worth of cat beds and Stripes is......

 

And Blacky

 

I've been waiting for Carrie Underwood and Pink to realize the same thing about their husbands for years....don't hold your breath

 

Title: Find the webpage link | PANDA
Requester: Dorothy D. Breitegam [A2L6YFN0Q57U4R]
TurkerView: [ $ / hour ]
Description: Find the webpage link
Duration: 60 Min
Available: 497
Reward: $0.05
Qualifications:

• Link EqualTo 10,

[3TSW5DH5LC1406A4MPIJEOIAPWX2TR]​

 

Title: Find the webpage link | PANDA
Requester: Dorothy D. Breitegam [A2L6YFN0Q57U4R]
TurkerView: [ $ / hour ]
Description: Find the webpage link
Duration: 60 Min
Available: 497
Reward: $0.05
Qualifications:

• Link EqualTo 10,

[3TSW5DH5LC1406A4MPIJEOIAPWX2TR]​

 

Title: Write Stereotypes about People [MANY HITS INCOMING]! | PANDA
Requester: MIT [A1WZ5548AAD4H5]
TurkerView: [ $3.00 / hour ]
Description: Can you write a sentence about a group of people? There will be many HITs uploaded soon!
Duration: 20 Min
Available: 84
Reward: $0.40
Qualifications:

• HIT approval rate (%) GreaterThan 95,
• Location EqualTo US,

[3NJNK03V5JHWTOO8UPA4548W9C5WWY]​

 

So my biggest issue with this is where are we finding a technology on this planet to determine fairness?

I mean, it's a cool thought experiment, but it fails in practical application as I'm seeing it...

Let's say we ignore the fact there is no system to determine "fairness" and just leave it down to RNG.

So, the biggest problem is that the largest % of HITs will still go to those who are already HIT-hawking the platform, because you cannot sub to a HIT you don't yet know exists (the same problem in panda management now). Alrighty, so everyone else can sub to the leftovers, cool. Or maybe we just eliminate manual acceptance of HITs and only allow subs then a distribution after some publish time.

We just RNG assign it and.. we're still in the same position. It's already RNG assigned to whoever happens to have a network request go through the moment the HIT actually becomes available. Introducing a pub/sub client doesn't really improve too much on that process, and introduces its own pitfalls/shortcomings (either you are assigning to clients who may have subscribed forever ago and are possibly no longer around, or you are interrupting a client in the middle of work [introducing unfairness to the requester relying on attention to their task] to confirm a subscription => assignment effect). If we try to assign by latest sub we 1) introduce unfair allocation and 2) produce the same "gaming" of the system that already exists, those who continually re-sub will gain an advantage that produces more unfairness. Even a system which limits the # of subscription requests can be gamed by intelligent systems, so we're not in a great standing here.

I do kind of like the idea of a pub/sub panda management system where a client can select some N of HITs and receive an RNG ticket to them. It still introduces some intelligent decision making (choosing valuable, but not overly competitive HITs and ones you hopefully qualify for) but I can't picture a working model of the "guaranteed fair allocation" part of your idea. I don't think it's possible to do, but would be interested to hear if you legitimately think it's possible vs just musing on how to stop "power" workers from vacuuming up HITs (I think this is less a problem in practice than others, tbh).

 

It's a Mountain Goats cover, but Laura Jane Grace is a badass

 

Also if you introduce this system on MTurk as an Amazon feature you're abandoning one of their strongest claims to the IC system: they do not control the flow of work (looks @ Prolific, currently operating in Cali, which currently has pretty clear cut rules around this).

If we throw this onto the requesters.. why not just have them implement a system w/ qualifications as some already choose to do?

 

Title: Decision Making Study $0.25-$1.25 approx. 8 minutes | PANDA
Requester: BTOF42 [AQM225XDL35GZ]
TurkerView: [ $13.70 / hour ]
Description: Slip into the role of an R&D manager in a short experiment, earn $0.25 for participation + of up to $1 ~8 minutes
Duration: 30 Min
Available: 5
Reward: $0.25
Qualifications:

[35MNBRJU8A31N1T7CVWO3H4OUQD7TK]​

 

Social Perception Research Lab [A262SMAQH7D4SE] Online Opinions(~ 6 minutes) - $0.75 | PANDA

Good

Unrated

Unrated

$12.80 / hour
00:03:31 / completion time

​

Pros:
pretty quick

Cons:

Qualifications: Exc: [1795151158-167993] DoesNotExist ; Location In US;

To read Social Perception Research Lab's full profile check out TurkerView!​

 

Have you seen this one lol

 

Serg, why you gotta be like that.

 

Title: Nature park hiking game (~25 minutes and bonus) | PANDA
Requester: Niv Lab [A3RR9T5E327UK4]
TurkerView: [ $16.93 / hour ]
Description: You will be hiking in a nature park, and your goal is to hike every trail in it while repeating as few trails as possible. BONUS: You will be paid according to your performance. The experiment takes ~ 25 mins, and you are paid a baseline of $0.50, and a performance bonus up to $3.50.
Duration: 120 Min
Available: 0
Reward: $0.50
Qualifications:

• Total approved HITs GreaterThanOrEqualTo 1,
• HIT approval rate (%) GreaterThanOrEqualTo 95,
• Location EqualTo US,

[3S82SNMR1Z20OA3GIQZX4LHVRIVFA8]​