ABSTRACT

This chapter is concerned with the processes by which procedural knowledge is acquired. There are two aspects to this acquisition process. First, the system must identify the procedures that are required to succeed in a particular task. This is not a trivial problem because the procedural knowledge is not usually directly stated to the learner. He must induce the procedures by observing examples of the behavior the procedures are supposed to generate. This has the appearance of being a very difficult logical task for problem domains like language acquisition. It has seemed incredible to some that humans are able to succeed at induction tasks like language learning. This induction problem has no real parallel in the acquisition of declarative knowledge. Usually, we are directly told the propositions we have to commit to memory, whereas there seldom is a direct statement of procedures. 1 Of course there are some cases where we must induce declarative knowledge. For example, this is just the task of a scientist. However, it seems that we go about that induction task much more awkwardly and with much less success than when we induce procedural knowledge. Witness the fact that a generation of scientists has had so little success in identifying the principles of language understanding, but that almost all children learn to understand a language.