But the robot _might_ be being controlled by some guy in another room who is also controlling the lightbulb (alternatively the person could be following instructions given by an earpiece). Or maybe they're both hooked up to a geiger counter. I admit it seems pretty unlikely that anybody would wire a robot or person up to flip a switch at exactly the same time a lightbulb comes on, but it only seems unlikely because of our pre-existing knowledge about robots and lightbulbs. If we take away that knowledge and simply make it a statement about how variable A changes when variable B does, it's much harder to infer which way, if either, the causal link goes.

(I suppose if you really wanted to be paranoid you could suppose that your _own_ actions in flipping the lightswitch might be subconsciously being controlled by someone else, but this is a whole different level of skepticism)