I suspect this works best if you avoid priming the test subjects on what they're going to be tested on, otherwise I think they will expend effort to seem extra-reasonable contra their natural impulse.

But, yes, good idea, I was way too quick to call it an insurmountable problem.