When trying to decide between many options, one often-tempting move is to exclude all options that are dominated by some other option, i.e., that is strictly inferior on each dimension. This is a good move if you capture all relevant dimensions, but in practice, I think it’s easy to either miss some dimensions or use the wrong option space.
Examples from economic policy
[This section would be significantly improved by finding real examples of real people making these mistakes.]
In economic policy, my sense is that it’s easy to mix-up pareto-optimality with respect to a totally open option-space (where a benevolent dictator with complete information could magically distribute resources without distorting any incentives, similar to Kaldor-Hicks efficiency) and pareto-optimality with respect to what options are actually available to policy-makers. For example, from this wikipedia page, consider the line “attempts to correct the distribution may introduce distortions, and so full optimality may not be attainable with redistribution.” But insofar as redistribution can give any individual more resources than they would have had under any realistic non-distributive policy, then that redistributive policy will be pareto-efficient with respect to the realistic policy space. In which case you’d be making a mistake to reject redistribution based on an argument that Pareto-optimality is an unobjectionable criterion that only excludes options that are dispreffered by everyone.
Here’s another examples, inspired by complaints I’ve heard about how some people supposedly seem to reason about climate policy: The efficient solution to climate change (among some large set of policies) would be a global CO2 tax, which might naively suggest that we can exclude all policies that are dominated by this. But of course, in practice, a global CO2 tax isn’t politically feasible. Which means means that some other solutions will be Pareto-efficient with respect to the realistic policy space, even if they’re not efficient with respect to the larger policy space.
This can also happen in my personal life
Sometimes I consider a list of options, and decide that one of them seem to be dominated by another option, but feel some reluctance to drop it. If so, that’s likely because there’s some factor that made that option intuitively appealing to me, but that I either fail to notice or intellectually don’t give much value to. But then dropping that option is quite plausibly a very bad move, either because (i) my intuition was pointing at some factor that was actually important, or because (ii) my intuition will switch to supporting some other option that does well on that factor, but does worse on most other axes (but isn’t quite dominated by anything else due to doing well on at least one of them).
How I’d recommend thinking about this
When applied correctly, Pareto-optimality is indeed a weak and unobjectionable criterion. Thus, whenever you appeal to a dominance argument, I think the mood should be one where you’ve helpfully pointed out an obvious way to improve the option space, where every party in the discussion (and each of your personal intuitions) are happy to accept it and move on. And typically, there should still be plenty of discussion left, because Pareto-optimality alone is unlikely to narrow down the option-space to just one point.
But whenever you appeal to a dominance argument and:
meet opposition (from others or from some internal part of yourself), or
feel like you made a clever move that cut out lots of options that you previously deemed plausible,
you should be very suspicious that you’ve actually missed important dimensions, or misunderstood what the option-space is.
(Also, maybe people should talk less about Pareto-efficiency (and just generic “efficiency”) and instead talk more about Kaldor-Hicks efficiency, which often seems to be closer to what is meant in practice, and is more obviously a reasonable thing to object to.)