Iterative controls in real time

In this chapter, I discuss the use of iterative feed-back systems for economic control. Their appealing feature is that they combine into one process several steps in economic planning: data gathering, computation, and implementation. One drawback is obvious: they operate in real time, with whatever costs are consequent upon quantities (prices) being “wrong” during the process itself. A second drawback is less obvious, but will emerge as we consider examples (e.g. section 5.7 below). The most these processes can do is enforce, at the partial equilibrium level, qualitative rules or targets derived (although not quantitatively computed) from general equilibrium welfare theory. In their present form, at least, it is not clear how to adapt them in cases in which rules derived from general welfare theory are in some way wrong or inadequate. I can only point out the difficulties here, and leave it to the reader to judge whether real-time feed-back systems, as a method of control alternative to economic planning, deserve further consideration.

Not all the control mechanisms described in this book in fact rely on adaptive systems; those that do not are reserved to Part IV. It is required of them, too, that they economize on information and computation, and that, by providing suitable incentives, economize also on policemen.

The Criterion Function

It is thought that many organisms operate partially, or even entirely, as simple feed-back control processes.