So what of language; if the author is establishing a meaning from the outset, then you know whenever you see X it has an applied technical implication, both consistent and clearly delineated. In fact, the whole purpose of the book is to define common sense for the consumption of the senseless. That's almost the prerequisite of a business books best seller – jargon replacing the thought that that must be absent in the reader, making them ill suited to executing any contained ideas in the first place.

Perhaps you are suggesting that 'paternalism' was an exceedingly poor choice of word to serve as their technical term for the 'Nudge' treatise. True dat.
Perhaps you're suggesting that their absolutes are lacking in defence and that moving directly to the application of an unsubstantiated and logically flawed postulation is an act without intellectual rigour. Hard to argue.
Perhaps you're lamenting that the text simply replaces concepts with terms, leaving the same challenge of execution as when one sets to tackle a problem in the first place.

Example: I propose that all choices are Xs, and all agents of choice are Y, and all choosers are Z. My amazing discovery is that Y has a position of power in presenting Xs in such a way as to steer Z's decision. So maybe Y should think about how they're presenting choices, and manage the expectations of Zs in the process. And maybe Y should also think about creative Xs related to where Z's heads are at.
I call it the XYZ Nudge Theorem. Patent pending.

Who is “we” in this sentence? If it is libertarians, then obviously. If it is anyone in politics, then I heartily disagree: motorcycle helmet laws, seatbelt laws, social security, etc. are all policies the creators/supporters of which deem helpful, and yet are choice-eliminating. And it is to exactly those in politics that T&S are appealing – those who would try to use paternalism in the well-worn sense that you’ve outlined above. And hey, combining an adjective with a noun to create something that means neither strictly the adjective nor the noun is called language (see “ant lion” or “compassionate conservative”).

But, it seems naive to think that the “choice architects” will limit their activity to things that will benefit all of the targets of their actions.

Their incentives are aligned with the wishes of people who will add to their power, not with people they affect. They will certainly end up “helping” in ways that are not actually helpful, low-cost, or even choice-preserving.

It requires ignoring mountains of evidence to think that this dynamic will change if we add adjectives like “soft” or “libertarian” to the activity.

They’re abusing “libertarian,” too.

They combine them and apply the combination to something that’s neither.

Also, pushmedia1, I agree with you that there are dynamic implications to making decisions for people, but it’s already being done and right know what people think the “right” decision is really isn’t (that is, the authors are pushing more for a change in the quality, not the quantity, of external choice architectures, since these architectures, whether we like them or not, are already there). Furthermore, recent psychological research points toward limited quantities of willpower and cognitive ability in a given day, suggesting that if people are currently subject to this willpower constraint, then an increase in the quantity of external choice architectures may actually increase the quality of internal choice architectures.