Compensation is difficult. As most of you probably know there are many pressures to reduce cost in academia, while simultaneously paying fair wages to workers. There is also movement (due to the higher cost) to move to "opt-in" surveys where the "payment" is some kind of personally interesting information. For instance,
this site lets you take many different surveys. That is very useful for research and is potentially an interesting (almost fun) use of time. Because there is no money involved, questions of "fair wage", "exploitation" etc are largely irrelevant. It is obvious that the people are doing it for the intrinsic value of getting personal information from a psychological lens. We may consider this model in the future. We probably will get a different sample using such an approach.
Of course, sites like that are not Mturk. As you probably know, a few years ago "$2-$3/hour" was supposedly the norm. This self-propagating poor payment was a product of mturk's infancy. Since then there are new norms saying $6/hour or some will say pay "minimum wage". I've seen people arbitrarily defined the "minimum wage" as 7.25/hour. To me this seems as arbitrary as any other number, but whatever - that's what people are staking their morality on currently.
So we created a payment scheme that fell in that range. We guarantee $6/hour if you finish the entire battery, and will pay a bonus that could bring your earnings up to 7.50/hour. There is some wiggle room due to time spent, and the bonuses aren't guaranteed, but we wanted to keep some aspect of the payment performance based as a motivational force.
Finally, we pilot these tasks. We ran this study a month or so ago, and that helped us become more confident in our payment and the time taken for the whole HIT.
This are just my thoughts on the matter, but I hope they give some clarity!