How to be a good doctor
The following will not convince anyone deeply opposed to consequentialist lines of thinking. But for those who waver, and for consequentialists who regret that they are committed to this, I hope to provide some encouragement.
Imagine you invent an amazing luck changing machine: the serious illness concentration machine (TSICM). The TSICM doesn't affect the total number of serious illnesses that affect the population of the group it targets, but does change their distribution.
Let's imagine that 1/10 people die of heart disease, 1/10 of lung cancer, 1/10 of liver cancer, 1/10 of gender specific cancers, and 1/10 of stomach cancer. Let's further assume that (a) a person cured of one such illness will live an equal number of extra healthy years as a person cured of any other such illness, (b) that none of these is, in fact, treatable, and (c) that all of these illnesses are not encouraged by any lifestyle choices their sufferers have made. (These assumptions are untrue of the above, but will be true for some other diseases)
So as it is, 5/10 people die of these illnesses. The TSICM does something quite peculiar about this. It recognises when a patient gets a gender specific cancer, and automatically gives them the other four illnesses. They were doomed anyway, so the extra illnesses make little difference to their fate. But since the TSICM doesn't affect the total number of serious illnesses in the population, this means that four other people will now avoid death from these things.
So the overall outcome is that the death rate from these illnesses drops by 4/5. One patient still dies, but four fewer than would otherwise.
If we were in possession of such a machine, should we turn it on? Of course! It is, under my assumptions, pareto optimal. No-one is worse off, and four people in ten are better off.
---
Now a brief twist: Your engineers can't quite make a TSICM. But they can manage a TSICM-beta. The TSICM-beta doesn't assign four illnesses to one person who already has one. Instead, it assigns all five randomly to one healthy individual. Should we still turn it on?
Previously, someone was unlucky enough to receive a gender specific cancer, through no fault of their own. That's pretty arbitrary, and effectively random. So previously, under the TSICM regime, one person died at random. This is equally true under the TSICM-beta regime. It also leads arbitrarily to a random death. So it seems to me that the TSICM-beta is not worse than the TSICM. We should also turn it on.
---
The final twist: Now imagine you are a doctor, and have just had five patients admitted to your hospital, each in need of a different organ without which they will die. You also have a perfectly healthy patient down the corridor. You could kill her, and distribute her organs amongst the newcomers, saving all their lives.
Many forms of consequentialism commit you to endorsing such redistribution (at least under various conditions that rarely hold in the real world: e.g. that no-one should know what you've done). Is this unpalatable?
Well, how does it differ from the previous machine? If it does not differ, then I can see no reason for rejecting it but accepting use of the TSICM-beta.
