Step one in normative economics is, I think, finding out what you want. Without a goal, positive economics is as useless as a coffee break (unless, of course, your goal was a cup of coffee). Is it odd, then, that economists don't seem to spend any time figuring out what people want?
Although I'm certainly no authority, it seems to me that psychologists and sociologists spend a lot more time on this question. For example, I recently came across a psychology article with the wonderful title "Not Having What You Want versus Having What You Do Not Want"(here's a link, but the full text is subscriber-only). I cannot resist quoting the first paragraph:
"No childhood passes without disappointment about a birthday present, no adolescence seems to be complete without a disappointing love affair, and hardly anyone is a stranger to the unpleasant feeling that stems from buying an expensive consumer product that turns out to be less than expected. All in all, a life without disappointment seems rare."
And they call economics the dismal science... I wish we could be so melancholy. The point, however, is that while behavioral economics might be trying to push the boundary of what the people in economic models care about by incorporating, for example, disappointment, I don't know of any economics literature that's trying to figure out what people actually care about.
Of course, it's difficult. How could we go about it? Economists are very distrustful of surveys as unscientific. One of my favorite normative economics results is hidden in a paper called "Economics of the Endangered Species Act" (link): US household surveys asked people what they'd be willing to pay to save each of a set of endangered species. Scaled up, the answers implied that the US population would be willing to pay one percent of its total income to save two percent of endangered species. It's a bit of a facetious point, but it's a neat way of showing that talk is cheap in answering surveys. Simply observing what people do can't really answer the question either, especially when we're talking about a bigger scale than the individual level.
Do we assume that democracy will elect leaders who represent the goals and ambitions of the people? I think the whole thing poses a serious problem for any economist bold enough to make a policy recommendation: how has the normative branch of economics tried to figure out what people want?
If someone asked you what you wanted, specifically or generally, for you or for the country, or the world, what might make the list? Money, a job, friends, lovers, health, the environment? How closely will the list match the things inside your head? What if you had to give up one thing on your list to get another? Again, a common recourse in economics is to turn our back on this whole sorry mess and just take the things in which we have relatively high confidence: most people like money to some degree.
No economics paper seems to be complete without the mysterious "welfare analysis", which is essentially a bolted-on normative exercise attached to a positive, descriptive theory. How valuable is such an exercise if we have simplified the motivation of people in the theory? Obviously the normative "welfare analysis" is equally dependent on the assumptions of our theory as the theory itself. It's a false dawn that is equally as unsuited to answering the question of "what should be" as positive economics itself: again, if we could come up with the metric that captured the quality of any conceivable thing, the magical normative criteria, we should pack up and start a technocracy.
Yet there is no reason to force the positive and normative analyses into the same box. Expanding the foundations of normative analysis, in particular, to include the hypothetical answer to the question of what people want to happen, can happily be done without affecting the quality of positive theory, whether or not it rests on the same foundations. The economist who claims to evaluate the quality of an outcome fails to see that what he calls normative economics is only as realistic as positive economics: not at all. Real normative economics would spend more time trying to figure out what people really want.