Isn't this an example of a reflection problem? We induce this change in a system, in this case an evaluation metric, and now we must predict not only the next iteration but the stable equilibria of this system.