I enjoyed reading David Lewis’ On the Plurality of Worlds. Of course, for me it was like entering a theater in the third act: my eclectic self-education has big holes in it which inhibit my comprehension of such a work.
The usefulness of expressing modal concepts in terms of possible worlds seems clear enough (here’s a blog post which outlines this well – hat tip to the latest philosophy carnival). David Lewis argues that the value and utility of possible worlds in modal thinking points to a metaphysical or ontological argument that the full set of possible worlds exists concretely. In the book he defends this thesis.
Most other philosophers would say possible worlds are abstract, like mathematical objects. Even setting aside that the ontological status of mathematical objects is highly controversial topic, Lewis moves from his usual modest tone toward vehemence in criticizing these views as insufficient in supplying the full structure needed for possible worlds; further the question of how the actual world is selected from the set of abstract possibilities is either inadequate or “magical” in competing accounts.
In other words a sufficient proposal for a modal metaphysics either needs to have Lewis’ ontological panoply of concrete worlds, or else we need a fuller account of the ontological status of the abstract world along with a detailed mechanism for selecting the actual world (in Lewis’ scheme, “actual” just means the world we happen to be in, but there is nothing else special about it relative to the other worlds). Note that this means a system with only one actual concrete world and no ‘abstract realm+selection’ endowment is simply an inadequate metaphysics. While Lewis takes this for granted, I think it's interesting that seemingly only a small subset of philosophers would endorse either the Lewisian system or an abstract alternative with full mechanisms for addressing the Lewisian critique. They are presumably content to deal with modality in epistemological and linguistic arenas unburdened by the metaphysical worries.
While Lewis doesn’t explicitly address it much in this book, the other way to parse the metaphysics is in terms of causality. Lewis’ approach enables him to be a Humean about causality. Everything in the world just happens. If we want to talk about what counterfactual situations could have possibly happened to us, say, then we are talking about things that happen to our counterparts in other worlds. If we want to say there is only one concrete world where things happen and these are caused in a way which could have happened differently (they are contingent things), then we need a model of “real” causation that can handle this selection process. These two alternatives exclude the common assumption many seem to share that there is only one world, but a simple billiard ball notion of causation is all we need. In such a world there is no possibility or contingency at all. Again, this would be an inadequate metaphysics.
So, what do I think of Lewis’ proposal? I don’t like the Humean/fatalistic aspect of it. I would like to think a selection process happens rather than think that every possible world exists in an even-handed way, and we just happen to be in one. But at this point this is a preference on my part, not an argument.
What I get out most out of this is what I’ve been emphasizing in this post: Lewis is an obstacle to those who think they can “get away with” a minimalist metaphysics like single-world physicalism. It’s not enough.