There’s a view in philosophy on which you can be doing one of two things, conceptual analysis or talking about the world. These things are held to be quite distinct, none of this is to say that it’s impossible to get from one to the other, but in particular the road from conceptual analysis to talking about the world is thought to be long and arduous.
I think this view is wrong, at least for some problems. I think that often conceptual analysis can tell you things about the world with very few premises required.
Suppose that the correct analysis of the concept of knowledge had been “true and justified belief.”
Now suppose that you accept two premises:
1. At least for a great variety of concepts, what that concept refers to is the property most similar to it .
2. Properties are very plentiful. For pretty much any list of features there’s a property that is a conjunction or disjunction or whatever else of those features.
We have the list of features specified {Is a belief AND Is justified AND Is true}, and we know that knowledge refers to whatever is most similar to that list of features, and we have reason to think that there exists a property which is exactly similar to that list of features. We’ve moved from the concept knowledge to the property of being knowledge in record time. What’s more, those premises can be relaxed in various ways to fit in with different theories of properties, and a variety of different theories of content and reference. I’ll leave it up to you to see whether some variant of those premises fit your own view.
Now what I really want to know is, am I going crazy? That argument seems pretty elementary to me, but it also seems like there’s a number of papers where the authors worry about the relation between metaphysics and concepts but don’t even acknowledge the possibility of that tactic. Are there obvious and widely known objections to it that I somehow have never ran into? Please advise!
Blogging the End 
It seems to me that this tactic would require a theory of properties on which they can be uninstantiated (at least at the actual world). To ‘discover properties’ in this way doesn’t even require that we have analysed some pre-existing concept, we can just concoct new ones. It seems to me that we should admit that we can know about the existence of lots of properties a priori in this way (so I don’t think there will be any ultimately good objections, but perhaps there are some sophistical apparent ones).
I suggest that the intuition that conceptual analysis can’t tell us anything about the world, if it is to be saved, should be saved by refining what we mean by ‘something about the world’, for example to mean ‘some contingent fact’, in which case your tactic gives no clear counterexample, since we can hold that these properties exist necessarily.
This is an interesting strategy for a lot of purposes but won’t actually save the defender of the gap. It’s fine to say that everything we learn about the world from conceptual analysis is necessarily true, and therefore we cannot learn anything contingent from conceptual analysis. But the defender of the gap wants to say we can’t learn anything about metaphysics through conceptual analysis. Unless you want to say that metaphysics is contingent then, this counter argument does nothing.
I agree. The last paragraph was just meant to help explain away the intuition motivating the defender of the gap. It wasn’t meant to be a counterargument for the gapper. It strengthens your refutation, by helping explain why the gap seemed plausible, instead of just leaving that mysterious.
Analysis of a concept A gives you a number of other concepts that are in sum, equivalent to it. But nothing there tells you that there is anything that answers to the concept A. It is not difficult to take a concept (say Santa Claus) and decompose it into subsidiary concepts. But there may not be anything that answers to the subsidiary concepts either (such as fat man able to defy gravity). The fundamental difference is modal. To falsify a conceptual analysis all you need is a possible case in which it fails. For a theory of the world you need an actual instance where it is false and possibilities do not matter.
So yes, I think you are way off on this one.