J.W. Mason and Arjun Jayadev have a paper making a new (to me) point about the assignment of targets to instruments.
First I'm going to present an (over-?) simplified version of their model, to explain the gist of it. [I think I've got the gist of it, but I'm not 100% sure, and I know my simplification isn't 100% accurate.] Then I'm going to attack it (because it's a good paper and so worth attacking).
There are two instruments: M and F. There are two target variables: A and D, with targets A* and D* respectively. Which instrument do you assign to which target?
Suppose the picture looks like this:
You want to set M and F to get to the point where the red and green lines cross, so you hit both targets at the same time.