This post which attacks utility maximization reminded me of the argument in defence of utility maximization models: While we do not actually believe that consumers equate their marginal rates of substitution across goods we think that they act as if they do. We certainly do not believe that compute their MRS as they are staring at the grocery store shelves.
The analogy that is usually made is that of a game of billiards or pool. The player certainly does not make momentum and angle computations as he is about to take a shot. The outcome of the shot is that he acts as though he does. Sometimes he makes mistakes. This is the flaw in the theory of utility maximization - it does not account for deviations or mistakes. Why does the consumer not maximize utility - is it the cost or complexity of the calculations. What situations is she most likely to make mistakes - consumer choice or career choice? What are the consequences of these deviations for social welfare.
Unfortunately, my reading of the evidence in favor of utility maximization is that it is weak. Empirical research does not focus on MRS directly but on the implications of utility maximization, i.e. demand curves and demand systems. Alternatives are tests of whether the Generalized Axiom of Revealed Preferences are violated.
P.S. What I find a little contradictory is that economists who have no problems with utility maximization and models of consumer behavior criticize representative agent models and DSGE in general. While I agree that the latter are not the best models, it is also the case that utility maximization may not be the best paradigm in economics.