How much did you try with mechanical turk? My experience is that when describing a HIT (Human Intelligence Task) it takes a few tries to get the results you wan

How much did you try with mechanical turk? My experience is that when describing a HIT (Human Intelligence Task) it takes a few tries to get the results you want. I suspect that if you asked for an entire quiz you'd get bad results. If I were setting up a HIT I would give the user three sample quiz topics and only have them create a single question for the quiz. I would then only use the top percentage of questions answered yet still accept people's responses even if they were not used but still within the HIT's requirements. Also, I think you can still benefit from seeding some more on your own.