Turkopticon
turkopticon.ucsd.edu
turkopticon.ucsd.edu
Here's a rebuttal: https://blog.turkerview.com/writer-who-never-learned-to-driv...
1) The $9/hour figure is based on making ~$0.15 per minute. This means if a survey on the platform took 5 minutes it should pay out ~$0.75. But then there's the question of who sets that time, and it's ultimately the requester. If they aim too low (or, conversely, high) this skews the accuracy. And given they're the ones paying it's in their interest to assume low. We were fond of using calibration-like surveys that would inform roughly how long a survey took before launching more to reduce this bias.
2) Even if you got (1) right, the $9/h figure assumes you're able to do those back-to-back with zero delay. That's damn near impossible without smart tools. There's often a multi-second (if not multi-minute) delay between surveys and all of those add up. Although we did generate some surveys long enough to meet the $9/h figure (and Turkers did get paid for them accordingly) most of the time we saw many small surveys where the cost of latency adds up. Ultimately, I don't think this was a problem of our approach so much as it was just a reality for any HIT or similar "microtask".
3) Even if you could have a perfect (1) and (2), another reality is that there is a finite volume of tasks being requested. Assuming an average of 5 minute HITs we'd need to be producing 12 units of work for an individual Turker per hour. That's a lot of outgoing work and really depends on what you're looking to attain. As a platform we couldn't really force more work to come out of it, so there was a natural limit.
4) Lastly, even if you managed (1) (2) and (3), it assumes everyone was actually getting all of their work from us. Fact of the matter is that there are a lot of companies that pay differently. Perhaps more, perhaps less, perhaps keyed on other more meaningful metrics than $/hour. But that all changes the calculus.
5) Lastly, even if you somehow managed everything, there's still the effect of locale. The value of $9USD/h varies wildly depending on the context. For this reason, it wasn't uncommon to see many requests coming from regions in the world where the dollar is comparatively strong.
Ultimately, it's really hard to say how much you could make in an hour as a Turker just because of the inherently volatile nature of the work. In instances where you do make an outsized amount of money it's more likely that you were filling an urgent need at that time. If I were to guess, chances are you'd see more work closer to the middle/end of business accounting periods as that's when large-n research would likely answer important questions, with higher pay likely towards the end to fulfill any last-minute research gaps.