The public will not deliberate in accordance with truth-sensitive principles; on the contrary, the public will err in accordance with definite patterns. The idea is that acquiring reliable knowledge about social theory (economics, pol sci, etc) is very costly to the average citizen, so he will rely on theories by default that are mostly false (for example: "we need to protect our industry against foreign competition", "higher crime results from lenient courts", etc, etc). The public, in short, is rationally ignorant. Reliable social science is hard because it is opaque and complex. Folk knowledge is easy to apprehend because it is vivid. Knowing this, politicians and others use, for electoral purposes, a rhetoric that feeds into these false theories. As a result, public deliberation does not bring us closer to the truth. On the contrary, deliberation increases error. We call this phenomenon discourse failure.
In comments, PrawfsBlawg's "resident deliberativist", Ethan Leib, argues that this is a rehash of the old "the people are too incompetent to govern" argument. I think that he's right that this problem may be partially rectifiable at an institutional level (or at least that we have to try), but in general I'm not optimistic.
I've hit on these themes before, most notably in my post attacking the concept of "persuasiveness" equaling "strong rhetoric" in debate, or my post of no-confidence in democracy. In general, I like the idea of "deliberative democracy," as long as "deliberative" has some depth to it--where "rational ignorance" no longer counts as deliberation. Engagement is the key. Unfortunately, I don't see much hope for salvation, and Professor Teson's argument about "rational ignorance" helps explain why.