INTRODUCTION

Agents are all the rage. ``Visioneering'' videos, such as Apple Computer's Knowledge Navigator, have helped to popularize the notion that programs endowed with agency, if not intelligence, are just around the corner. Soon, users need not themselves wade into the vast swamp of data in search of information, but rather the desired, or better yet, needed information will be presented to the user by an intelligent agent in the most comprehensible form, at just the right time.

Although such rosy scenarios are easy to come by, intelligent agents are considerably more difficult to obtain. Still, substantial progress is being made on a variety of aspects of the agent story. At least three general conceptions of agent-based software systems can be found in current thinking:

1. Agents are programs sent out over the network to be executed on a remote machine.
2. Agents are programs on a given machine that offer services to others.
3. Agents are programs that assist the user in performing a task.

Each of these models can be found to some extent in present-day software products, for example, in (1) General Magic's emerging TELESCRIPT interpreter, (2) Microsoft's OLE 2.0 and (3) Apple Computer's Newton and Hewlett Packard's New Wave desktop, respectively. Given this space of conceptualizations, we need to be specific about ours.

 
• Definitions and Objectives