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.

It is essential to reduce the cost of integrating information sources and to provide a path that allows for incremental integration that can be responsive to users' demands. This paper presents an approach to integrating disparate heterogeneous information sources that uses context logic. Our use of context logic reduces the up-front cost of integration, provides an incremental integration path, and allows semantic conflicts within a single information sources or between information sources to be expressed and resolved. Introduction The number of online network-accessible information sources grows daily. The information promises to provide tremendous value for individuals and corporations. The promise will remain unfulfilled, however, until it is possible to integrate and assimilate information from multiple heterogeneous sources. Because it is impossible to predict the users and patterns of usage in our changing information environment, information providers are not willing to pay a ...

Classical planners have traditionally made the closed world assumption --- facts absent from the planner's world model are false. Incompleteinformation planners make the open world assumption --- the truth value of a fact absent from the planner's model is unknown, and must be sensed. The open world assumption leads to two difficulties: (1) How can the planner determine the scope of a universally quantified goal? (2) When is a sensory action redundant, yielding information already known to the planner? This paper describes the fully-implemented xii planner, which solves both problems by representing and reasoning about local closed world information (LCW). We report on experiments utilizing our UNIX softbot (software robot) which demonstrate that LCW can substantially improve the softbot 's performance by eliminating redundant information gathering. Introduction Classical planners (e.g., (Chapman 1987)) presuppose correct and complete information about the world. Although recent wo...

) William P. Birmingham, Edmund H. Durfee, Tracy Mullen, and Michael P. Wellman Artificial Intelligence Laboratory University of Michigan 1101 Beal Avenue Ann Arbor, MI 48105-2110 fwpb,durfee,mullen,wellmang@eecs.umich.edu Abstract We provide an overview of a distributed-agent architecture design for the large-scale digital library environment currently under development at the University of Michigan. This paper presents some of the design principles and techniques underlying our approach, as well as some preliminary observations from our first steps in developing the system. Overview: Diversity in the UMDL One of the most interesting and technically challenging features of digital libraries is diversity---in users, in information sources, and in many other features. For example, user skills, information demands, and level of professional competence will vary greatly among users of digital libraries. Similarly, the collections will vary over many parameters, including breadth an...

In this paper a knowledge-level model of an individual designer as an agent is described, in which reflective reasoning about elements of situatedness, and reasoning from the point of view of other participants, are explicitly modelled. This model is based on existing models of single agent design. An individual designer in a specific distributed design process, namely website design, is used to illustrate the model.

This paper discusses an intelligent agent that learns to identify documents of interest to particular users, in a distributed and dynamic database environment with databases consisting of mail messages, news articles, technical articles, on-line discussions, client information, proposals, design documentation, and so on. The agent interacts with the user to categorize each liked or disliked document, uses significant-phrase extraction and inductive learning techniques to determine recognition criteria for each category, and routinely gathers new documents that match the user's interests. We present the models used to describe the databases and the user's interests, and discuss the importance of techniques for acquiring high-quality input for learning algorithms. 1. Heterogeneous document databases A growing number of businesses and institutions are using distributed information repositories to store large numbers of documents of various types. The growth of Internet services such as M...

Agent-Based Knowledge Discovery provides a new technique for performing data-mining over distributed databases. By combining techniques from Distributed AI and Machine Learning, software agents equipped with learning algorithms mine local databases. These agents then co-operate to integrate the knowledge obtained, before presenting the results to the user. We are currently exploring the use of a new software agent language, AgentK and the application of first order learning techniques to data-mining. However, the main area of investigation is how the agents should interact, and how the knowledge should be integrated. Introduction This paper describes our current research 1 which spans the fields of knowledge discovery and software agents. Knowledge discovery (or data-mining) is concerned with extracting knowledge from databases and/or knowledge bases (Piatetsky-Shapiro & Frawley, 1991). Most datamining systems employ one or more machine learning techniques to find previously unknown...

Information agents answer user queries using a large number of diverse information sources. The key issue in their perfomance is finding the set of information sources relevant to a query. Previous work has considered determining relevance soley based on compile-time analysis of the query. We argue that at compile-time, it is often not possible to significantly prune the set of sources relevant to a query, and that run-time information is needed. We make the following contributions. First, we identify the different types of information that can be obtained at run-time, and how they can be used to prune information sources. Second, we describe an algorithm which naturally extends query planning algorithms to exploit run-time information. Third, we describe the discrimination matrix, which is a data structure that identifies the information that can be used to help discriminate between different possible sources. Introduction Information gathering agents are programs that answer user q...

Agents need to be able to adapt to changes in their environment. One way to achieve this, is to provide agents with the ability of self-modification. Self-modification requires reflection and strategies with which new knowledge can be acquired, a necessary condition for creativity. This paper describes a knowledge-level model for the design of self-modifying agents and explores the feasibility of automatically designing self-modifying agents.

The paper presents the guidelines of a project of three italian Universities (Bologna, Siena, Trento) which aim is to integrate a multi-agent recommendation system that suggests educational resources to students into a mobile learning platform that supports mobile learning processes in a University context. The project covers three main areas. The first area is concerned with finding effective models for mobile learning. The second regards the evaluation of learning processes in mobile learning environments. The third focuses on the technological aspects of integrating the multi-agent recommendation system into mobile learning management system. The project has its foundations in the availability of significant experience on e-learning real processes, and on the availability of the source code of an elearning system developed in previous projects and currently used by different faculties. In this framework we will integrate InLinx, a multi-agent Web-based hybrid recommender system that provides a on-line bookmarking service in our m-learning architecture. 1.