'Bednets would be a more efficient use of those resources' is a nearly-fully-generalizable argument against the vast majority of spending within the United States.
You are correct on many of your objections, though I disagree on this point.
Levels of action, talks about how we can either do something, or increase the rate at which we do something, or increase the rate at which we increase the rate... ad infinitum. It's basically the difference from increasing a number directly, or increasing it's 1st/2nd/3rd/etc derivative.
To do lives saved, we can give children bed nets to directly save lives. Or we can build factories to build bed net to save lives faster. Or we can invest in automated technologies to build faster factories. Or we can invest in general AI research to build automated technologies faster (eg, the singularity). By devoting resources to bed nets directly we save lives now, but by going up each level we save way more lives later. Thus, things like donating to SIAI, are saving more lives eventually than bed nets directly. But if you're here, you probably already agree with SIAI above bed nets.
For example with the government's case, scientific research eventually speeds up the rate at which we do everything, and education eventually speeds up the rate of the rate at which we do everything. In this manner, I wouldn't consider bed nets better than sci research or education. And there are a lot of other programs that provide similar benefits. However, I don't think imprisonment is one of those things where more funding gets us better meta-improvements.
You are vastly overstating the value of those savings, by valuing them based on the best possible use of any dollar, rather than on their actual likely use.
These inefficient programs should also get cut until their marginal utility is comparable. If they start wasting the money that was saved to imprison other people, then we fight that inefficient practice too. There are millions of ways in which our government is wasteful and inefficient. It's the job of us voters to try and constantly push it to be less so.
Or is the convention against discussing politics here silly?
I propose a test. I'm going to try to lay down some rules on voting on comments for the test here (not that I can force anybody to abide by them):
1.) Top-level comments should introduce arguments (or ridicule me and/or this test); responses should be responses to those arguments.
2.) Upvote and downvote based on whether or not you find an argument convincing in the context in which it was raised. This means if it's a good argument against the argument it is responding to, not whether or not there's a good/obvious counterargument to it; if you have a good counterargument, raise it. If it's a convincing argument, and the counterargument is also convincing, upvote both. If both arguments are unconvincing, downvote both.
3.) Try not to downvote particular comments excessively, if they're legitimate lines of argument. A faulty line of argument provides opportunity for rebuttal, and so for our test has value even then; that is, I want some faulty lines of argument here. If you disagree, please downvote me, instead of the faulty comments, because this post is what you want less of, not those comments. This necessarily implies, for balance, that we not excessively upvote comments. I'd suggest fairly arbitrary limits of 3/-3?
Edit: 4.) A single argument per comment would be ideal; as MixedNuts points out here, it's otherwise hard to distinguish between one good and one bad argument, which makes the upvoting/downvoting difficult to evaluate. (My apologies about missing this, folks.)
I'm going to try really hard not to get personally involved, except to lay down a leading comment posing an argument against abortion, a position I don't hold, for the record. The core of the argument isn't disingenuous, and I hold that this argument is true, it just doesn't lead to my opposing abortion. I do not hold the moral axiom by which I extend the basic argument to argue against abortion, however; I'm playing the devil's advocate to try to help me from getting sucked into the argument while providing an initial point of discussion.
Which leads me to the next point: If you see a hole in an argument, even if it's an argument for a perspective you agree with, poke through it. The goal is to see whether we can have a constructive political argument here.
The fact that this is a test, and known to be a test, means this isn't a blind study. Uh, try to act as if you're not being tested?
After it's gone on a little while, if this post hasn't been hopelessly downvoted and ridiculed (and thus the premise and test discarded as undesirable to begin with), we can put up a poll to see whether people found the political debates helpful, not helpful, and so on.