Writing about global development, Owen Barder criticizes the 'do no harm principle'. He says that in medicine and development it makes no sense to insist on doing no harm, because you wouldn't be able, then, to do anything at all. Sound policies can have harmful side effects: 'Lorry drivers are harmed when the railway is built'; a medicine may save many lives but harm a few. He sums up what he takes to be the sound moral position on this question as follows:
We should pursue policies which do (much) more good than harm, which take account of the broader consequences, which take proportionate steps to limit the harm and risks, and which are careful and responsible about looking after those who are likely to lose from change.
As a statement for the context in which he formulates it, one can see the point of his argument: risk-free and cost-free policies are hard to come by, and some harms may have to be countenanced in order to secure greater all-round benefits. But if we take this in a more general way as laying down a moral standpoint for all contexts, it's in need of qualification. More good than harm isn't always enough to swing it.
First, there are some harms which are not acceptable whatever the benefits they might be thought to secure. This is so for those who think that there are fundamental human rights that may not be violated. That's why, for example, the right against torture in international humanitarian law is non-derogable: it is to be respected in all circumstances, including war, crisis, and national emergency. Nothing, no putative benefit, justifies the use of torture. Second, benefits which are very long-range and uncertain don't necessarily outweigh costs that are immediate and to existing persons, even if in abstract the former appear to be greater than the latter. Real costs can't be defeated by very speculative benefits.
There are doubtless more qualifications necessary than just these two, but the point is that the 'Do no harm principle' can't be adequately replaced by a purely quantitative measure of goods against harms. Some harms should not be done in any circumstamces. Some benefits are too uncertain to justify definite harms. (Via.)