All About Computers

Research on computers as assistants means looking for new ways of dividing the labor between humans and computers. On the one hand, future systems should take on more tasks than existing systems do, especially those that seem tedious or difficult for humans. On the other hand, they should not automate tasks completely. The basic paradigm is that of assistance. In many fields of application the problems are either too complex or simply too numerous for any attempt to develop a machine with complete problem-solving competence to succeed. What is called for instead is a set of calibrated tools that the user can combine, adapt, and employ as he or she sees fit. Exhaustive treatment and coverage of a problem is in fact not the goal of computers as assistants. There are many characteristics of effective assistance. A human assistant, for instance, is expected to be competent in his or her domain of expertise, to know his or her limitations, to be able to process inexact instructions, to adjust to a client and to learn from the client, and to be able to explain his or her behavior and suggestions. The facilitation of communication and cooperation is a central function of an assistant. If computer systems are to offer assistant capabilities, they must be supplied with domain knowledge and knowledge about the user. There is an additional requirement: Computer systems need knowledge about themselves, that is, they need to know about the way they function. Only if a system can observe its own behavior, and reflect on it, will it be able to correctly evaluate its own competence and explain its behavior.