To a certain degree one could test instrumental rationality indirectly. Perhaps have them set a goal they haven't made much progress on (dieting? writing a novel? reducing existential risk?) and see if instrumental rationality training leads to more progress on the goal. Or give people happiness tests before and a year after completing the training (i.e. when enough time has passed that the hedonic treadmill has had time to work). Admittedly, these indirect methods are incredibly prone to confounding variables, but if averaged over a large enough sample size the trend should be clear.