A year ago, we published a special issue of The Knowledge Engineering Review entitled "Perspectives on Recent Intelligent Agents Research as Viewed through Two Conferences". Our intention was to provide a snapshot of current research on intelligent agents - an area which continues to grow apace - by looking at the work presented at the 1997 instantiations of two of the major events in the Agent calendar. These were the International Conference on Autonomous Agents (Agents '97) and the International Conference and Exhibition on the Practical Application of Intelligent Agents and Multi-Agent Technology (PAAM '97). The result was a group of six papers which together spanned the main areas covered by the conferences as well as identifying some of the challenges faced by the agents community as a whole. In particular, we had summaries by the organisers of both of the conferences, speculation on the future of robotics, surveys of expert assistants and electronic commerce, and a discussion of how agents could be used in workflow management.

Constraint satisfaction is a fundamental artificial intelligence technique offering a simple yet powerful representation. An increasing amount of attention has recently been paid to the development of constraint satisfaction techniques, and it has become clear that the original formulation of a static Constraint Satisfaction Problem (CSP) with hard, imperative constraints is insufficient to model many real problems. Two important extensions to the classical CSP framework which address some of these deficiencies are flexible and dynamic constraint satisfaction. This paper examines in detail classical, flexible and dynamic CSP. It reviews the motivations behind both extensions, and describes the techniques used to solve each type of problem. The paper employs a running example throughout to illustrate the ideas presented.

The toxicity prediction problem lies squarely at the interface of biology, chemistry and computational domains and will require the integrated application of knowledge and approaches from each domain for its solution. It is not a single modeling problem, but rather multiple levels of modeling compartments derived from the goals of toxicity risk assessment. These compartments include different categories and characteristics of toxicity (e.g., cancer vs. non-cancer, acute vs. delayed) and, therein, different levels of biofunctional organization and multiple mechanisms of toxicity extending to the level of individual chemical structures. A toxicity prediction model should strive to resolve the global toxicity prediction problem to local modeling compartments that reflect coherent biofunctional mechanisms and common modes of action while retaining some level of useful generalizations. This paper attempts to use these general concepts to frame the challenges and opportunities for application of Artificial Intelligence (AI) and computer-based methods to the goal of chemical toxicity prediction.

In the software and knowledge engineering literature we can find maintenance strategies offered to maintain seven main types of knowledge: words; sentences; behavioral knowledge; and meta-knowledge. Meta-knowledge divides into problem solving methods, quality knowledge, fix knowledge, social knowledge and processing activities. There are five main ways in which these seven knowledge types are processed: acquire, operationalise, fault, fix and preserve. We review systems that contribute to these 7*5 = 35 types of knowledge maintenance to make the following conclusions. First, open issues with the current maintenance research are identified. These include (a) areas that are not being addressed by any researcher; (b) the recursive maintenance problem; and (c) drawbacks with rapid acquire systems and the operationalisation KM assumption. Secondly, a process is described for commissioning a new maintenance tool. Thirdly, a general common principle for maintenance (search-space reflection) is isolated.

Model-based simulation systems have been created in various fields of engineering to train personnel or students in operation, maintenance and troubleshooting of complex devices and systems. A review of literature indicates a lack of good overviews of the approaches to system design of model-based training simulations in Intelligent Computer Assisted Instruction (ICAI). Though single systems have to some extent been evaluated with regard to their performance, an organised evaluation, especially a comparative evaluation of the systems that have been created within the field is lacking. To be able to successfully conduct an in-depth review under these conditions, we concentrate and centre our review around the design of the domain simulator component and its implications for the other parts of instructional system. For this purpose we have identified features that are in our opinion suitable to characterise the design approach of a model-based simulation in ICAI for training. We have organised these design features in a framework to enable an effective review. Within this framework, we investigate each feature separately and illustrate it by giving examples from existing applications where appropriate.

Often, the problems encountered by applications and developments of software agents require further investigation and improvements of agents' social capacities, that is, of the properties enabling them to act adaptively in multiagent domains. This is so for at least three reasons. First, quite often software agents are designed for reasoning upon and interacting with other agents (both human and artificial), as is the case in client-server and personal assistants. Secondly, to accomplish their primary task, these systems ought to reason upon, predict and eventually coordinate with, or at least prevent obstacles from, other systems accomplishing the same or different tasks in a concurrent way (think of electronic commerce, meeting agents, etc.). Thirdly, many applications are actually distributed over a set of agents with their specific roles and tasks, which are expected to participate in a joint activity, aimed at achieving a given global outcome.

Predicting or learning numeric features is called regression in the statistical literature, and it is the subject of research in both machine learning and statistics. This paper reviews the important techniques and algorithms for regression developed by both communities. Regression is important for many applications, since lots of real life problems can be modeled as regression problems. The review includes Locally Weighted Regression (LWR), rule-based regression, Projection Pursuit Regression (PPR), instance-based regression, Multivariate Adaptive Regression Splines (MARS) and recursive partitioning regression methods that induce regression trees (CART, RETIS and M5).

Dealing with context seems to be a major challenge which the field of Artificial Intelligence (AI) will have to face in the next few years. The importance of the challenge is borne out by the various scientific events focusing on context which have been held since 1995. This does not mean that context has not been studied before—it has in such domains as Natural Language Processing—but it has yet to be studied in depth since previous research has only covered few aspects of the problem. We present in this paper a survey of the literature dealing directly and explicitly with context whatever the domain. This makes it possible for us to have a clear view of context in AI. One of the conclusions of this survey is to point out the existence of different types of context in areas such as the representation of knowledge in a computer system, the reasoning that the system carries out using the knowledge, and the interaction the system has with people.

Telecommunications infrastructures are a natural application domain for the distributed software agent paradigm. The authors clarify the potential application of software agent technology in legacy and future communications systems, and provide an overview of publicly available research on software agents used for communications management. The authors focus on the intelligent agent type of software agent, although the paper also reviews the reasons why mobile agents have made an impact in this domain. The author's objective is to describe some of the intricacies of using the software agent approach for the management of communications systems. The paper is in four main sections. The first section provides a brief introduction to software agent technology. The second section considers general problems of network management and the reasons why software agents may provide a suitable solution. The third section reviews some selected research on agents in a telecommunications management framework. The final section concludes the paper by discussing some of the problems encountered and some future directions for further research.

The rapid development of the field of agent-based systems offers a new and exciting paradigm for the development of sophisticated programs in dynamic and open environments, particularly in distributed domains such as web-based systems of various kinds, and electronic commerce. However, the speed of progress has been such that it has also brought with it a new set of problems. This paper reviews the current state of research into agent-based systems, considering reasons for the way the field has grown, and pointing at the way it might continue to progress. It pays particular attention to problems with defining the nature of agents, the technologies that have enabled the rapid progress to date, and ways in which work can be consolidated through the development of large-scale applications, and the integration with theoretical foundations.

Both the advent of large communication networks and the development of agent models designing more and more sophisticated (in some cases autonomous) behaviours, are producing what can be called an Agent Society. Agent societies are constituted by both artificial and human agents (in perspective indistinguishable from each other) communicating with each other (not necessarily each with all the others), with individual or collective tasks, different resources, different skills, and so on. In other words, these agent societies will become more and more similar to the human ones. As in real societies, in these virtual societies it will be necessary to consider all the problems connected with secure and reliable interaction, with trusting the other agents, with the possibility to be frauded, cheated, deceived in some exchange or relationship.

The new technologies for Knowledge Discovery from Databases (KDD) and data mining promise to bring new insights into a voluminous growing amount of biological data. KDD technology is complementary to laboratory experimentation and helps speed up biological research. This article contains an introduction to KDD, a review of data mining tools, and their biological applications. We discuss the domain concepts related to biological data and databases, as well as current KDD and data mining developments in biology.

This paper sets out, ambitiously, to present a brief reappraisal of software agents research. Evidently, software agent technology has promised much. However, some five years after the word ‘agent’ came into vogue in the popular computing press, it is perhaps time the efforts in this fledgling area are thoroughly evaluated with a view to refocusing future efforts. We do not pretend to have done this in this paper—but we hope we have sown the first seeds towards a thorough first five-year report on the software agents area. The paper contains some strong views not necessarily widely accepted by the agent community.The views expressed in this paper are those of the authors only, and do not necessarily reflect the views of our organisation.

For a number of years, researchers in AI and robotics have been sharing organisation principles for software development called “architectures” (Arkin, 1998; Hexmoor et al., 1997; Kortenkamp et al., 1998). Most recently, there has been an interest in extending the software architectures originally designed for robotic applications to accommodate autonomous operation beyond robotics (Pell et al., 1998). These architectures attempt to state principles for organising software for intelligent systems that exhibit cognitive functionalities such as natural language understanding, planning and learning, as well as organism-like functionalities to cope in the world such as reflexive and reactive behaviours.

The 1998 Workshop of the UK Special Interest Group on Multi-Agent Systems was held in Manchester in December, chaired and organised by Michael Fisher of Manchester Metropolitan University, continuing the series of focussed and constructive meetings in this field. After two very successful workshops on the Foundations of Multi-Agent Systems at the University of Warwick in 1996 (Luck, 1997; Doran et al., 1997; d'Inverno et al., 1997; Fisher et al., 1997) and 1997 (Luck et al., 1998; Aylett et al., 1998; Binmore et al., 1998), the scope was broadened for 1998 to a wider range of issues concerning all aspects of multi-agent systems.

How free are our software agents to take the best possible care of our interests? How free can we make them? In what sense and to what extent do currently proposed mechanisms and agent behaviors consider self-interest? What current research addresses these issues? What needs to be done? We address these issues in the context of electronic markets, such as consumer goods markets and (future, more fine-grained) markets for electric power or communication bandwidth.

In systems composed of multiple autonomous agents, negotiation is a key form of interaction that enables groups of agents to arrive at a mutual agreement regarding some belief, goal or plan, for example. Particularly because the agents are autonomous and cannot be assumed to be benevolent, agents must influence others to convince them to act in certain ways, and negotiation is thus critical for managing such inter-agent dependencies. The process of negotiation may be of many different forms, such as auctions, protocols in the style of the contract net, and argumentation, but it is unclear just how sophisticated the agents or the protocols for interaction must be for successful negotiation in different contexts. All these issues were raised in the panel session on negotiation.