Alexander Pruss has put forward some very troubling counter-examples to the solution Frances Kamm gives to the 'Loop' case. Since I follow Kamm, at least in outline, these are problem cases for me too.
Rather than summarise all that here readers can see Pruss for themselves. Here is a response - though not, at the moment, a solution.
Case 1. Jim the railway enthusiast is keen to build a railway between two large cities. His motivation is simply that he loves railways. A safety expert tells him, however, that over 50 years there are likely to be 50 deaths on this railway, if built. Jim is sad because this appears to show that it would be wrong to build it: 50 deaths is a high price to pay for his dream of a gleaming new railway. However, the expert quickly adds that the railway will divert traffic from the roads, and since railways are generally safer than roads the number of lives saved will be, say, 55.
Jim is happy again: it seems that these two unintended side-effects of his project don't just cancel each other out, but leaves him with a modest credit balance.
Case 2. Benny the blackmailer tells Jill that unless she kills one person (Charlie) he, Benny, will kill two other people (all these victims are innocent with nothing much to distinguish them). Jill knows it would be wrong to intend the death of an innocent even to save two lives. Instead, she points her gun directly at Charlie's head, and, with the intention of giving her trigger finger some much-needed exercise, pulls the trigger. She knows, of course, that Charlie will be killed as a result of this, but whereas this would normally be an act of terrible recklessness, she knows that the unintended side-effect of Charlie's death will be more than off-set by the saving of the two other innocents.
What we need to bear in mind with Pruss's cases, which are like my Case 2, is that while it seems that in these cases something is going horribly wrong, the plausibility of the reasoning in Case 1 and others like it is perfectly ok. Not just ok, in fact, but it is essential that we are able to off-set side-effects if we are to engage in any large-scale action: government policies on transport, education, health and so on will invariably generate 'winners and losers' and we balance these out to see if the policy is permissible. If we are to stop Jill's line of reasoning we must show that it is different from that of Jim. Case 1 shows that Kamm's original insight ('triple effect') remains correct.
So how are they different?
1. Jim could intend the good results if he wishes; Jill could not, because if she did she would have to intend the evil means to them. Put another way, the good results come from the evil results for Jill, but not for Jim.
2. Jill is acting much more immediately than Jim, on persons she can identify. She is more closely involved in the evil of innocent deaths.
3. The trivial goods Jill and others like her are intending could be achieved in other ways, which do not bring about such drastic harms. We might ask: what is Jills' intention in doing it this way? The answer would be: to take advantage of the favourable balance of unintended side effects available in this exact situation.
I'm not happy with any of these three differences, as the basis for allowing Jim's reasoning and ruling out Jill's.
On 1: Jim could intend the good results, but as stipulated he doesn't. This will often be the case with the leaders of large-scale projects. Again, consider the 'butterfly effect': by turning over in bed, we may cause a hurricane in New Zealand. We don't need to worry, however, because we are just as likely to be preventing a hurricane in New Zealand. We don't need to go to the trouble of intending the good possible results of our careless actions in order to take advantage of their neutral overall effects; we just note (in response to objections) that we aren't making things any worse.
2. It has been noted before that having identifiable victims can have an effect on our intuitions. It surely can't have an effect on the truth of the matter. The identifiable victims in the Jill case partially explains our distaste for her reasoning, but I'm not ready to bite the bullet and say that, therefore, what she does is ok.
3. We might we say: Jill's intention in exercising her trigger finger with a loaded gun pointed at Charlie with the safety catch off is to capture the positive balance of side-effects, which would not be captured in any other way. Similarly, we always adapt our plans in order to avoid disastrous side-effects. If someone tells you that simply by travelling from A to B via C you may spread a deadly disease from C to B, then you don't go via C, you go via D instead. So Jill can exercise her finger with the gun pointed another way, unloaded etc. etc. but if she does that a disastrous result arises: two innocents are killed by Benny. Instead, she gains her trivial good in a way which won't have this disastrous result, but a less disastrous one, that of Charlie dying.
My feeling right now is that there is a moral difference between the cases, but I can't identify a principled explanation for it.