Paul De Grauwe is critical of the macroeconomic models used by central banks:
Cherished myths fall victim to economic reality, by Paul De Grauwe, Commentary, Financial Times: ...But that is not the world of the macroeconomic models that are now in use in central banks. The world of these models is one of supernatural and God-like creatures for which the world has few secrets. These creatures can perfectly compute the risks they take and estimate with great precision how an oil price shock will affect their present and future production and consumption plans. They may not be able to predict each shock, but they know the probability distribution of these shocks. Thus the risk involved in financial instruments is correctly evaluated by individuals populating these models.
These superbly informed individuals want the central bank to keep prices stable so that as consumers they can optimally set their consumption plans with minimal uncertainty, and as producers they can set prices equal to marginal costs (plus a mark-up). If the central banks keep prices stable, these individuals, helped by well-functioning markets, will take care of all the rest and ensure that the outcome is the best possible one. This is a world in which free and unfettered markets are always efficient.
This is also a world where individual agents cannot make systematic mistakes. Their consumption and production plans are optimal. They will never build up unsustainable debts. In the world of these macroeconomic models financial crises should not occur. And if they do, it cannot be because of malfunctioning markets. Governments that impose silly constraints on rational individuals are messing things up, and central banks that do not keep their promises to maintain price stability are the source of macroeconomic instability. ...
There is a danger that the macroeconomic models now in use in central banks operate like a Maginot line. They have been constructed in the past as part of the war against inflation. The central banks are prepared to fight the last war. But are they prepared to fight the new one against financial upheavals and recession? The macroeconomic models they have today certainly do not provide them with the right tools to be successful.
There is quite a bit of research on monetary policy that is devoted to the issues he is worried about, particularly the literature on adaptive learning. See here, and some of his publications here for and example from a European central bank. These ideas are also well-known and of considerable interest to the US Fed, e.g. one example is here, but there are many more. Some of the recent work of Chris Sims is also of interest in this regard:
...Most recently, theories that postulate deviations from the assumption of rational, computationally unconstrained agents have drawn attention. One branch of such thinking is in the behavioral economics literature (Laibson, 1997; Benabou and Tirole, 2001; Gul and Pesendorfer, 2001, e.g.), another in the learning literature (Sargent, 1993; Evans and Honkapohja, 2001, e.g.), another in the robust control literature (Giannoni, 1999; Hansen and Sargent, 2001; Onatski and Stock, 1999, e.g.).
This paper suggests yet another direction for deviation from the seamless model, based on the idea that individual people have limited capacity for processing information. That people have limited information-processing capacity should not be controversial. It accords with ordinary experience, as do the basic ideas of the behavioral, learning, and robust control literatures. The limited information-processing capacity idea is particularly appealing, though, for two reasons. It accounts for a wide range of observations with a relatively simple single mechanism. And, by exploiting ideas from the engineering theory of coding, it arrives at predictions that do not depend on the details of how information is processed.
Returning to the article above criticizing macroeconomic models, the author goes on to explain why he thinks the reliance of central banks on macroeconomic models is a problem:
This intellectual framework helps to explain the single-minded focus of many central bankers on inflation. Clearly, inflation is important and maintaining price stability is an important task of the central bank. It is not the only task, though. Financial stability is equally important. But this dimension is completely absent from the macroeconomic models now in use.
Perhaps more attention to financial market instability is warranted, there's a lot of work on that currently underway, but as this overview of the learning literature makes clear, if you drop the rational expectations assumption and assume agents must learn about their economic environment (one means of generating financial market instability), it may still be that the an aggressive response to inflation is optimal:
Expectations, Learning and Monetary Policy: An Overview of Recent Research, by George W. Evans and Seppo Honkapohja, July 16, 2008: ...contemporaneous Taylor-type interest-rate rules should respond to the inflation rate more than one for one in order to ensure determinacy and stability under learning...
So central banks' focus on inflation comes solely from examination of standard macroeconomic models.
The point is that, unlike the implication in the article, central banks are quite anxious to explore the implications of agents that are less than fully informed or fully rational, how that impacts behaviors such as risk assessment, and to examine the implications of financial market instability. They are particularly interested in how these factors impact the conduct of stabilization policy. The models aren't perfect, and standard models do miss a lot of these elements, but standard models are not all we have and central bankers are quite aware of, and actively engaged in exploring the policy implications of alternative theoretical structures that can tell us more about these issues.