The concept of an agent has become important in both Artificial Intelligence (AI) and mainstream computer science. Our aim in this paper is to point the reader at what we perceive to be the most important theoretical and practical issues associated with the design and construction of intelligent agents. For convenience, we divide these issues into three areas (though as the reader will see, the divisions are at times somewhat arbitrary). Agent theory is concerned with the question of what an agent is, and the use of mathematical formalisms for representing and reasoning about the properties of agents. Agent architectures can be thought of as software engineering models of agents; researchers in this area are primarily concerned with the problem of designing software or hardware systems that will satisfy the prop-erties specified by agent theorists. Finally, agent languages are software systems for programming and experimenting with agents; these languages may embody principles proposed by theorists. The paper is not intended to serve as a tutorial introduction to all the issues mentioned; we hope instead simply to identify the most important issues, and point to work that elaborates on them. The article includes a short review of current and potential applications of agent technology.

this paper, I will argue that, contrary to these challenges, planning deserves its central place on the AI map. I will claim that intelligent agents are planning agents, and that philosophical and commonsense psychological theorizing about the process of planning can provide useful insights into the question of agent design. The theories I have in mind are not restricted to The Uses of Plans 3 how agents can form plans. Much of my research has concerned the ways in which intelligent agents use their plans. I will describe some of that research, and will argue that plans are used not only to guide action, but also to control reasoning and to enable inter-agent coordination. These uses of plans make possible intelligent behavior in complex, dynamic, multiagent environments. 2 Planning We can begin by asking what exactly we mean by "planning". For many years, planning had a quite specific meaning in AI: it was the process of formulating a program of action to achieve some specified goal. You gave a planning system a description of initial conditions and a goal, and it produced a plan of action whose execution in a state satisfying the initial conditions was guaranteed to result in a state satisfying the goal. These plans were akin to recipes for achieving the goal. Your goal might be to have a chocolate cake. In the initial state, you might have eggs, milk, and chocolate, a pan and a working oven. In these conditions, a valid plan might be to go the store to buy some flour, return home, preheat the oven, mix the ingredients, pour the mixture into the pan, and put it in the oven for 45 minutes. Traditional AI planning systems like STRIPS [22], NOAH [63], and SIPE [71], were designed to construct just this kind of plan---except usually the goal was something like a tower o...

The congeries of theoretical views collectively referred to as "situated action" (SA) claim that humans and their interactions with the world cannot be understood using symbol-system models and methodology, but only by observing them within real-world contexts or building nonsymbolic models of them. SA claims also that rapid, real-time interaction with a dynamically changing environment is not amenable to symbolic interpretation of the sort espoused by the cognitive science of recent decades. Planning and representation, central to symbolic theories, are claimed to be irrelevant in everyday human activity. We will contest these claims, as well as their proponents ' characterizations of the symbol-system viewpoint. We will show that a number of existing symbolic systems perform well in temporally demanding tasks embedded in complex environments, whereas the systems usually regarded as exemplifying SA are thoroughly symbolic (and representational), and, to the extent that they are limited in these respects, have doubtful prospects for extension to complex tasks. As our title suggests, we propose that the goals set forth by the proponents of SA can be attained only within the framework of symbolic systems. The main body of empirical evidence supporting our view resides in the numerous symbol systems constructed in the past 35 years that have successfully simulated broad areas of human cognition. During the past few years a point of view has emerged in artificial intelligence, often under the label of "situated action " (henceforth, SA), that denies that intelligent systems are correctly characterized as physical symbol systems, and especially denies that symbolic processing lies at the heart of

There are many formal approaches to specifying how the mental state of an agent entails that it perform particular actions. These approaches put the agent at the center of analysis. For some questions and purposes, it is more realistic and convenient for the center of analysis to be the task environment, domain, or society of which agents will be a part. This paper presents such a task environment-oriented modeling framework that can work hand-in-hand with more agent-centered approaches. Our approach features careful attention to the quantitative computational interrelationships between tasks, to what information is available (and when) to update an agent's mental state, and to the general structure of the task environment rather than single-instance examples. A task environment model can be used for both analysis and simulation, it avoids the methodologicalproblems of relying solely on single-instance examples, and provides concrete, meaningful characterizations with which ...

actually gets done, especially how people involve each other in their work. In particular, a model of practice reveals how people accomplish a collaboration through multiple and alternative means of communication, such as meetings, computer tools, and written documents. Choices of what and how to communicate are dependent upon social beliefs and behaviors---what people know about each other's activities, intentions, and capabilities and their understanding of the norms of the group. As a result, Brahms models can help human---computer system designers to understand how tasks and information actually flow between people and machines, what work is required to synchronize individual contributions, and how tools hinder or help this process. In particular, workflow diagrams generated by Brahms are the emergent product of local interactions between agents and representational artifacts, not pre-ordained, end-to-end p

Coordination, as the act of managing interdependencies between activities, is one of the central research issues in Distributed Artificial Intelligence. Many researchers have shown that there is no single best organization or coordination mechanism for all environments. Problems in coordinating the activities of distributed intelligent agents appear in many domains: the control of distributed sensor networks; multi-agent scheduling of people and/or machines; distributed diagnosis of errors in local-area or telephone networks; concurrent engineering; `software agents' for information gathering. The design of coordination mechanisms for group...

. Adaptation in open, multi-agent information gathering systems is important for several reasons. These reasons include the inability to accurately predict future problem-solving workloads, future changes in existing information requests, future failures and additions of agents and data supply resources, and other future task environment characteristic changes that require system reorganization. We have developed a multi-agent distributed system infrastructure, Retsina (REusable Task Structure-based Intelligent Network Agents) that handles adaptation in an open Internet environment. Adaptation occurs both at the individual agent level as well as at the overall agent organization level. The Retsina system has three types of agents. Interface agents interact with the user receiving user specifications and delivering results. They acquire, model, and utilize user preferences to guide system coordination in support of the user's tasks. Task agents help users perform tasks by formulating p...

We introduce a technique for analyzing the behavior of sophisticated A.I. search programs working on realistic, large-scale problems. This approach allows us to predict where, in a space of problem instances, the hardest problems are to be found and where the fluctuations in difficulty are greatest. Our key insight is to shift emphasis from modelling sophisticated algorithms directly to modelling a search space that captures their principal effects. We compare our model’s predictions with actual data on real problems obtained independently and show that the agreement is quite good. By systematically relaxing our underlying modelling assumptions we identify their relative contribution to the remaining error and then remedy it. We also discuss further applications of our model and suggest how this type of analysis can be generalized to other kinds of A.I. problems. Chapter 1

ion-Partitioned Evaluator (APE) architecture which has been tested in a simulated, single-agent, indoor navigation domain [SH90]. The APE architecture is composed of a number of concurrent, hierarchically abstract action control layers, each representing and reasoning about some particular aspect of the agent's task domain. Implemented as a parallel blackboard-based planner, the five layers --- sensor/motor, spatial, temporal, causal, and conventional (general knowledge) --- effectively partition the agent's data processing duties along a number of dimensions including temporal granularity, information/resource use, and functional abstraction. Perceptual information flows strictly from the agent sensors (connected to the sensor /motor level) toward the higher levels, while command or goal-achievement information flows strictly downward towards the agent's effectors (also connected to the sensor/motor level). Besides mechanisms for communicating with other layers, each layer in the AP...

There are many formal approaches to specifying how the mental state of an agent entails the particular actions it will perform. These approaches put the agent at the center of analysis. For some questions and purposes, it is more realistic and convenient for the center of analysis to be the task environment, domain, or society of which agents will be a part. This paper presents such a task environment-oriented modeling framework that can work hand-in-hand with more agent-centered approaches. Our approach features careful attention to the quantitative computational interrelationships between tasks, to what information is available (and when) to update an agent's mental state, and to the general structure of the task environment rather than single-instance examples. A task environment model can be used for both analysis and simulation, it avoids the methodological problems of relying solely on single-instance examples, and provides concrete, meaningful characterizations with which to sta...