I sometimes think about robot controllers this way. Every arm corresponds to an action, meaning a control feedback loop applied to the robot for a short time.

I sometimes think about robot controllers this way. Every arm corresponds to an action, meaning a control feedback loop applied to the robot for a short time. If you find that two actions are both fairly good in a given state, the system might try automatically generating a new action that's some synthesis of the two, either an average or some kind of genetic algorithm crossover. I don't know if you can have a useful falsifiable hypothesis about individual arms. But you can about the entire system. The hypothesis you want to fail to falsify is that the system gradually improves its performance (measured by some reward function) over time.