It's not by any means a full answer, but I think that when it comes to economics and politics, we may have a predilection for systems that are more efficient and bring better results for the highest number of people. It sounds consequentialist, but ir's only close to that: the thing is that when we are discussing the State or a society, we have to think of the common good; no longer are we simply thinking of individual ethics and rules of action, we are, instead, thinking about what best favors the community in general. So I think this gives considerable strength to policies which are more efficient and bring the best results for the highest number of people or the biggest sectors of society. These policies should, however, respect intrinsic natural rights, and as such there's really no consequentialism involved.

To me the best ethical theory for this is is still natural law; politics would be something like ethics applied to the community, when we are seeking to further society's natural ends: peace, prosperity, and to better allow every individual to also fulfill his own natural ends. The common good takes precedence over the individual good, but at the same time individual rights are to be protected and there shouldn't be unjust laws.

Just my 2 cents anyway, nothing too fancy.

I suppose it depends if one is using the term descriptive economics in a general sense, or one means specifically what is current economic orthodoxy (i.e., more or less neoclassical economics). It doesn't seem to me that the latter should be treated without caution. Otherwise, I would say that descriptive economics, so far as it is purely descriptive and can be trusted as accurate, should be treated like any other technical or scientific knowledge is treated. Any normative or prescriptive platform has to take onboard relevant technical or scientific knowledge: if true, these  scientific fact are just how the world works. 

​That said, it is important to discern what are normative elements being sold as descriptive ones, just as it is important to make sure the latter are an entirely accurate picture of reality. 
 

I haven't read Freakonomics, but if I recall, it is one of those attempts to apply neoclassical microeconomics to wide facets of life? An obvious issue is that neoclassical microeconomics is a neat and tidy shambles, with absurd assumptions (SMD-conditions, the Walrasian auctioneer, no reswitching, etc.), dubious methodology (static equilibrium analysis, for example), and internal contradictions and inconsistency. I think any conclusion about the real-world based simply on neoclassical microeconomics must be treated with utmost suspicion. 

My understanding is that rational choice theory generally assumes an individual has perfect information and the unlimited ability to rank one's preferences, not to mention that it is possible to aggregate individual choices to derive a smooth, downward sloping demamd curve that intersects the supply curve once.

I am not sure what good such an approach would do to examining any social phenomena (unfortunately, mainstream macroeconomics now incorporates rational choice theory, as well as rational expectations).

That is not correct. The basic rational choice theory does assume people have perfect information about the alternatives, that is, about the rankings between two choices they face (it is also true that the same basic foundation, including perfect knowledge, is used in much of the rest of neoclassical economics, from general equilibrium theory to the efficient markets hypothesis). This is implicit in the precision of preference ranking.

​These are Samuelson's four principles of consumer rationality:

1. Completeness: When given two combinations of goods A and B, a consumer can decide which one he prefers. So A > B, B > A, or A = B (he can get the same degree of satisfaction from both combinations, in which case he is indifferent).

2. Transitivity: If A is preferred to B, and B is preferred to C, then A is preferred to C.\

3. Non-satiation: More is always preferred to less

4: Convexity: The marginal utility a consumer receives from each commodity falls with additional units consumed.

Leaving aside the obviously questionable notion of being able to rank one's preferences ad infinitum*, ​my understanding is that there is empirical evidence against this model of consumer rationality, especially transivity, such as Sippel (1997). Also, if the desire is to aggregate individual preferences and demand to form a market demand curve, this has been shown, by neoclassical economists themselves, not to lead to a smooth, downward sloping demand curve that will intersect the supply curve (which we can assume - perhaps incorrectly, as Sraffa showed - will slope upwards) just once (without the patently unrealistic SMD conditions). 

​I agree that the general notion of rational choice in some sense captures important parts of human behaviour. People will make what seems to them rational trade offs in an attempt to secure the best outcomes for themselves.  The basic issue with rational choice theory (at least the one at the centre of neoclassical economics), though, is the degree of precision, knowledge, and stability it assumes in the consumer, not to mention its methodological individualism. I see no reason why this variety of rationality should be the basis of all economics. It doesn't seem in tune, for example, with the Post-Keynesian emphasis on uncertainty, especially in financial markets and macroeconomics, nor with the empirical evidence. Nor does trying to plot informational costs seem like it will improve matters too much. For a start, that seems to fall victim to the unfortunate neoclassical tendency to equate risk and uncertainty (which seems to follow from any attempt to give a functional relationship to such costs).

​* Steve Keen has a good illustration of how questionable this assumption is:

Consider, for example, your regular visit to a supermarket. The typical supermarket has between 10,000 and 50,000 items, but let’s segment them into just 100 different groups. How many different shopping trolleys could you fill if you limited your decision to simply whether to buy or not buy one item from each group? You would be able to fill two to the power of one hundred shopping trolleys with different combinations of these goods: that’s 1,267,650,600,228,229,401,496,703,205,376 trolleys in total, or in words over 1,000 million trillion trillion shopping trolleys. If you could work out the utility you gained from each trolley at a rate of 10 trillion trolleys per second, it would take you 100 billion years to locate the optimal one. Obviously you don’t do that when you go shopping.