I want to know that any target is selected because there is verifiable intelligence that he’s actively planning a terrorist attack on the United States or its allies; that the danger is pressing; that arrest is impossible; and that civilian lives are not wantonly risked.I find the latter two restrictions are a bit more important and should be tighter constraints. If Bin Laden is found but we cannot tell if he is actually planning anything, does that mean he is not fair game? Does one use a standard of reasonable doubt--what one needs to get a conviction in court? Or would the standard be something like what it takes to be indicted? I am no lawyer nor do I pretend to be one, so I have no clue about standards of innocent.
What I do know is that the intel requirements that Cohen suggests might mean no drones strikes at all. Which may not be a bad thing. Active planning is a tough threshold because you would need to know more info than if the standard was: we are fairly certain that person x (she?) is a senior leader of Al Qaeda. For me, that would be good enough. I think drone strikes to kill the lower level guys do not make sense unless they are active--moving to a location to engage in a violent act or are in the midst of a violent act. Otherwise, these guys are pretty replaceable which makes the tradeoffs tip the other way.
Cohen is clearly right that we need to have a process that is not capricious, that is accountable and where oversight does actually exist. My work on restrictions in Afghanistan have led me to the conclusion that the right combo for delegation is to delegate a great deal but provide clear instructions (commander's intent) and strong oversight. Combining micromanagement with either strict or loose oversight seems to be a bad idea as is delegating and forgetting.*
*Thanks to Scott Gates for pushing me at the ISA into thinking more about the discretion contribution of the current project.