I worry that metaphysics is a sham -- or at least, that it's a sham if it's thought of as an a priori discipline whereby one discovers a special class of truths while reflecting from an armchair. When you tilt back in your armchair and reflect, there's only one kind of thing you can discover, it seems to me: Facts about your own psychology. What gets called metaphysics, then, is really just a certain kind of self-study -- typically the study of the metaphysician's own concepts.
You can learn about the world by learning about your concepts, if your concepts contain in them important information about the world. Often they do contain such information: My concept of a bird as a type of biological organism, warm-blooded, bipedal, winged, feathered, egg-laying, etc. contains packed in it information about a certain cluster of traits that tend to travel together. Because of this, reflecting on my concept of "bird" can bring forward facts I may not have explicitly considered in the past.
How do my concepts come to contain information about the world? They must do so by contact with the world -- either my own contact (directly by empirical observation or indirectly by hearing the reports of others) or my ancestors' contact if the concepts are innate. I don't see, then, how studying my own concepts could yield a different kind of information than studying the world directly; nor do I see how our concepts could provide an independent source of information about the world orthogonal to empirical observation or immune to empirical refutation. (For a parody of the idea that the truths suggested by conceptual reflection are immune to empirical refutation, see here.)
The weird, science fiction cases that philosophers tend to dwell on in metaphysical discussions are exactly the kinds of cases where we should expect our concepts to be least in touch with the world, aren't they? Consider disputes in the metaphysics of consciousness: Would a silicon robot that behaves just like a human being be genuinely conscious, or would it have no more real consciousness than the computer on which I'm writing this post? Is it "metaphysically possible" (even if in practice unlikely) that my conscious experience is radically different from yours (e.g., red-green inverted or worse) despite all the similarities in our behavior?
If we construe these questions as questions about our concepts or pre-existing ideas, then it's not unreasonable to think we can make progress on them from the philosopher's armchair (though other methods for studying our concepts may be equally or more illuminating, in the spirit of recent "experimental philosophy"). Some people apparently find it impossible to conceive that a robot that behaved like a human being wouldn't be conscious; others apparently find it impossible to conceive that such a robot would have human consciousness. That shows something about their concepts or background assumptions. But how could it be (as Searle and Block and Putnam and Lewis and many others seem to think) that armchair reflection could reveal whether robots really would be conscious? Our concept of "bird" works well for near-home cases but tells us nothing about life on other planets; so also our concept of "consciousness" works well enough for distinguishing waking from dreamless sleep, mundane red experiences from mundane yellow experiences, but how could it cast useful light on robots or inverted spectra?
Metaphysicians often respond to such concerns by pointing to mathematics: In math, it seems, we discover substantive facts about the universe from the armchair, so why not also in philosophy? But is it clear that in studying math we do discover substantive facts about the universe? Not every philosophy of math grants this assumption. Maybe what we do, in studying math, is simply invent and apply rules for symbol manipulation. Maybe we discover facts about the structure of our concepts and invent new concepts. So irrefutable seems to me the view (empirically grounded!) that from the armchair we can discover nothing beyond the circuit of our own minds, that a conservative philosophy of math is mandatory. I find that considering alternative rules of logic (e.g., intuitionist logic or dialethism) and alternative rules of arithmetic (e.g. Boolean algebra) helps me feel the pull of the idea that mathematics is more an invention than a discovery of mind-independent facts.