Thank you for the many excellent commments.
I would like to address one theme in particular. Some have suggested that public deliberation can still be successful because many political issues are moral in nature, so ignorance about social theory (e.g., economics) doesn't matter as much as I claim it does. This is a very popular view among political philosophers such as Rawls and Gutmann-Thompson, for whom the main political disagreements are moral in nature.
This view is also wrong, I believe, and we devote a chapter in the book to show why. Most (but not all) social disagreements are empirical, not moral. In fact, a major form of discourse failure is the attempt to overmoralize politics. Many people, especially politicians, try to take the high moral ground in order to conceal the fact that their proposal is counterproductive. For example, when you point to supporters of the minimum wage that it causes umemployment, they reply that the minimum wage is required for moral reasons (say, to avoid exploitation).
Similarly, someone who publicly defends protectionism often treats it as a moral matter: "we must save the jobs of our workers," thus concealing the fact that millions of consumers and poor foreign producers will be hurt. Unfortunately, scholars are not above this either. We call this move The Moral Turn, and it is a major form of discourse failure. It is Kantianism gone awry. It is also lazy, because the supposed morality of the position exempts the speaker from complex causal inquiry.
Yet some political disagreements are genuinely moral (abortion for example.) We propose The Display Test: a position is genuinely moral when the speaker can accept, without embarrassment, its bad consequences. For example, I am prepared to publicly accept, without embarrassment, that criminal defendants should have a number of rights, even if someone shows me that implementing those rights increases crime. If the speaker cannot concede without embarrassment the bad consequences of her proposal, then she is perpetrating discourse failure.