Which economists have that point of view? Some have accused Smith of this, but it is plainly untrue, as made clear in his "Theory of Moral Sentiments".[1] Austrian economists such as Mises (who few would call socialist) criticized even the cardinal utility model, and argued that only lists of ordinal utilities are useful.[2]
[1] http://en.wikipedia.org/wiki/The_Theory_of_Moral_Sentiments
[2] http://en.wikipedia.org/wiki/The_Theory_of_Money_and_Credit
von Mises is unusual in expressing the view that human utility cannot meaningfully be known and understood except through preferences expressed by [economic] action but that's merely a stronger version of the same principle that the price system is a way of linking people's preferences [purely ordinal, in von Mises' case] with a [symbolically cardinal] monetary value.
for example, this is why an interaction with a large (inter-)national bank is often very different from the local credit union. a credit union may give you a break on fees or on check clearing times because they see you standing in front of them in need and they're sympathetic to your situation, but at a major bank, policy decisions are made in private offices far from contact with everyday customers and are simplified into a cold hard calculation about how much profit can be extracted from you.
also i'd posit that not only are humans sometimes not rational, but that humans are commonly irrational. rationality derives from a useful simplification of the world that helped us survive against bigger, faster & stronger animals. we've stretched rationality marvelously to make it a useful model of a wide range of relatively simpler macro-phenomena. but it's a poor model for complex actors like people, who exhibit a much wider degree of freedom and range of action. rationality necessarily assumes that a single best choice can be made in any situation (rarely, two or more choices might be equivalently good but that's not relevant here) and that all relevant factors can be incorporated into that single decision. this is almost never the case in real life, except maybe in the problem sets assigned in school.
it's as likely that a rational model is designed to coincide with the center of a distribution of decisions made by real people, as it is that the model has explanatory power about that distribution. people are incorporating an astonishingly vast array of information in their decision-making that is very unlikely to be considered by a rational economic model, so those "astonishing bullshit reasons" shouldn't be so surprising really if you don't conflate the modeled behavior of the population with that of the individuals.