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.

In this paper, we bring techniques from operations research to bear on the problem of choosing optimal actions in partially observable stochastic domains. We begin by introducing the theory of Markov decision processes (mdps) and partially observable mdps (pomdps). We then outline a novel algorithm for solving pomdps offline and show how, in some cases, a finite-memory controller can be extracted from the solution to a pomdp. We conclude with a discussion of how our approach relates to previous work, the complexity of finding exact solutions to pomdps, and of some possibilities for finding approximate solutions.

In this paper, we describe the partially observable Markov decision process (pomdp) approach to finding optimal or near-optimal control strategies for partially observable stochastic environments, given a complete model of the environment. The pomdp approach was originally developed in the operations research community and provides a formal basis for planning problems that have been of interest to the AI community. We found the existing algorithms for computing optimal control strategies to be highly computationally inefficient and have developed a new algorithm that is empirically more efficient. We sketch this algorithm and present preliminary results on several small problems that illustrate important properties of the pomdp approach.

The concept of an agent has recently become important in Artificial Intelligence (AI), and its relatively youthful subfield, Distributed AI (DAI). 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 the area into three themes (though as the reader will see, these 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 constructing software or hardware systems that will satisfy the properties specified by agent theorists. Finally, agent languages are software systems for programming and experimenting with agents; these languages typically 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 key issues, and point to work that elaborates on them. The paper closes with a detailed bibliography, and some bibliographical remarks. 1

We define the probabilistic planning problem in terms of a probability distribution over initial world states, a boolean combination of propositions representing the goal, a probability threshold, and actions whose effects depend on the execution-time state of the world and on random chance. Adopting a probabilistic model complicates the definition of plan success: instead of demanding a plan that provably achieves the goal, we seek plans whose probability of success exceeds the threshold. In this paper, we present buridan, an implemented least-commitment planner that solves problems of this form. We prove that the algorithm is both sound and complete. We then explore buridan's efficiency by contrasting four algorithms for plan evaluation, using a combination of analytic methods and empirical experiments. We also describe the interplay between generating plans and evaluating them, and discuss the role of search control in probabilistic planning. 3 We gratefully acknowledge the comment...

An agent's ability to understand an utterance depends upon its ability to relate that utterance to the preceding discourse. The agent must determine whether the utterance begins a new segment of the discourse, completes the current segment, or contributes to it. The intentional structure of the discourse, comprised of discourse segment purposes and their interrelationships, plays a central role in this process (Grosz and Sidner, 1986). In this thesis, we provide a computational model for recognizing intentional structure and utilizing it in discourse processing. The model specifies how an agent's beliefs about the intentions underlying a discourse affects and are affected by its subsequent discourse. We characterize this process for both interpretation and generation and then provide specific algorithms for modeling the interpretation process. The collaborative planning framework of SharedPlans (Lochbaum, Grosz, and Sidner, 1990; Grosz and Kraus, 1993) provides the basis for our model ...

: We provide a model for reasoning about knowledge and probability together. We allow explicit mention of probabilities in formulas, so that our language has formulas that essentially say "according to agent i, formula ' holds with probability at least b." The language is powerful enough to allow reasoning about higher-order probabilities, as well as allowing explicit comparisons of the probabilities an agent places on distinct events. We present a general framework for interpreting such formulas, and consider various properties that might hold of the interrelationship between agents' probability assignments at different states. We provide a complete axiomatization for reasoning about knowledge and probability, prove a small model property, and obtain decision procedures. We then consider the effects of adding common knowledge and a probabilistic variant of common knowledge to the language. A preliminary version of this paper appeared in the Proceedings of the Second Conference on T...

This paper presents a formal account of how to determine the discourse relations between sentences in a text, and the relations between the events they describe. The distinct natural interpretations of texts with similar syntax are explained in terms of defeasible rules. These characterise the effects of causal knowledge and knowledge of language use on interpretation. Patterns of defeasible entailment that are supported by the logic in which the theory is expressed are shown to underly temporal interpretation. 1 The Problem of Temporal Relations An essential part of text interpretation involves calculating the relations between the events described. But sentential syntax and compositional semantics alone don't provide the basis for doing this. The sentences in (1) and (2) have the same syntax, and so using compositional semantics one would predict that the events stand in similar temporal relations. (1) Max stood up. John greeted him. (2) Max fell. John pushed him. But in (1...

Most known computational approaches to reasoning have problems when facing inconsistency, so they assume that a given logical system is consistent. Unfortunately, the latter is difficult to verify and very often is not true. It may happen that addition of data to a large system makes it inconsistent, and hence destroys the vast amount of meaningful information. We present a logic, called APC (annotated predicate calculus; cf. annotated logic programs of [3], that treats any set of clauses, either consistent or not, in a uniform way. In this logic, consequences of a contradiction are not nearly as damaging as in the standard predicate calculus, and meaningful information can still be extracted from an inconsistent set of formulae. APC has a resolution-based sound and complete proof procedure. We also introduce a novel notion of "epistemic entailment" and show its importance for investigating inconsistency in predicate calculus as well as its application to nonmonotonic reasoning. Most importantly, our claim that a logical theory is an adequate model of human perception of inconsistency, is actually backed by rigorous arguments.

A fundamental assumption made by classical AI planners is that there is no uncertainty in the world: the planner has full knowledge of the conditions under which the plan will be executed and the outcome of every action is fully predictable. These planners cannot therefore construct contingency plans, i.e., plans in which different actions are performed in different circumstances. In this paper we discuss some issues that arise in the representation and construction of contingency plans and describe Cassandra, a partial-order contingency planner. Cassandra uses explicit decision-steps that enable the agent executing the plan to decide which plan branch to follow. The decision-steps in a plan result in subgoals to acquire knowledge, which are planned for in the same way as any other subgoals. Cassandra thus distinguishes the process of gathering information from the process of making decisions. The explicit representation of decisions in Cassandra allows a coherent approach to...