In this respect, what Pearl seems to have accomplished sometimes looks like a formalism in search of an interpretation without which the truth or the falsity of his claims is often impossible to assess. If the conceptions upon which his view is based do indeed conform to one or another of the traditional Bayesian models, moreover, then the very idea of a probability-based heuristic confronts a number of difficult problems of its own with respect to the distribution of probabilities to sets of alternative hypotheses, paths, or solutions, relative to the proposed refinements of those alternative hypotheses, paths, or solutions.6 These considerations suggest that traditional conceptions should not be taken for granted, especially if we assume that this is what Pearl intends by his observation that "Probability theory is today our primary (if not the only) language for formalizing concepts such as "average" and "likely," and therefore it is the most natural language for describing those aspects of (heuristic) performance that we seek to improve" (p. On general theoretical grounds, I think, there are excellent reasons to suppose that (a)-(f) are fundamental problems in AI science and that an extensional probabilistic analysis of this sort simply cannot lead to their effective solutions. In order to understand the traditional approach, however, this book is recommended with the reservations implied above, namely, that the author has omitted basic definitions that might not be familiar to some readers, and that serious difficulties seem to confront the theoretical framework he apparently endorses, where these difficulties are especially severe from an epistemological perspective.

The recent advances in computer speed and algorithms for probabilistic inference have led to a resurgence of work on planning under uncertainty. The aim is to design AI planners for environments where there might be incomplete or faulty information, where actions might not always have the same results, and where there might be tradeoffs between the different possible outcomes of a plan. Addressing uncertainty in AI, planning algorithms will greatly increase the range of potential applications, but there is plenty of work to be done before we see practical decision-theoretic planning systems. This article outlines some of the challenges that need to be overcome and surveys some of the recent work in the area. In problems where actions can lead to a number of different possible outcomes, or where the benefits of executing a plan must be weighed against the costs, the framework of decision theory can be used to compare alternative plans.

This article describes a methodology for programming robots known as probabilistic robotics. The probabilistic paradigm pays tribute to the inherent uncertainty in robot perception, relying on explicit representations of uncertainty when determining what to do. This article surveys some of the progress in the field, using in-depth examples to illustrate some of the nuts and bolts of the basic approach.

The Fourth Uncertainty in Artificial Intelligence workshop was held 19-21 August 1988. The workshop featured significant developments in application of theories of representation and reasoning under uncertainty. A recurring idea at the workshop was the need to examine uncertainty calculi in the context of choosing representation, inference, and control methodologies. The effectiveness of these choices in AI systems tends to be best considered in terms of specific problem areas. These areas include automated planning, temporal reasoning, computer vision, medical diagnosis, fault detection, text analysis, distributed systems, and behavior of nonlinear systems.

To assess the state of the art and identify issues requiring further investigation, a workshop on qualitative and abstract probability was held during the third week of November 1993. This workshop brought together a mix of active researchers from academia, industry, and government interested in the practical and theoretical impact of these abstractions on techniques, methods, and tools for solving complex AI tasks. The result was a set of specific recommendations on the most promising and important avenues for future research. The workshop, entitled "Putting Qualitative and Abstract Probability to Work," gathered active researchers from university, industry, and government to assess the state of the art and make recommendations for future research. The event was sponsored by the Palo Alto Laboratory of Rockwell Science Center.

This article examines the direction in which knowledge bases are constructed for diagnosis and decision making When building an expert system, it is traditional to elicit knowledge from an expert in the direction in which the knowledge is to be applied, namely, from observable evidence toward unobservable hypotheses However, experts usually find it simpler to reason in the opposite direction-from hypotheses to unobservable evidence-because this direction reflects causal relationships Therefore, we argue that a knowledge base be constructed following the expert's natural reasoning direction, and then reverse the direction for use This choice of representation direction facilitates knowledge acquisition in deterministic domains and is essential when a problem involves uncertainty We illustrate this concept with influence diagrams, a methodology for graphically representing a joint probability distribution Influence diagrams provide a practical means by which an expert can characterize the qualitative and quantitative relationships among evidence and hypotheses in the appropriate direction Once constructed, the relationships can easily be reversed into the less intuitive direction in order to perform inference and diagnosis, In this way, knowledge acquisition is made cognitively simple; the machine carries the burden of translating the representation "OK," we replied, "If the tiger were present, what is the probability that you would see that image? On the other hand, if the tiger were not present, what is the probability you would see it?" Before we could say "what is the probability there is a tiger in the first place?" Since then, we have pondered this question. Why is it that we want to look at problems of evidential reasoning backward?

Department of Computer Science Carnegae-Mellon Unaverszty P&burg, PA 15213 Editors' Note: Many expert systems require some means of handling heuristic rules whose conclusions are less than certain Baysian techniques and other numerical scoring methods have been developed to combine and propagate certainty measures as the expert system draws inferences in solving different problems. Doyle's paper argues that it is difficult for a human expert to produce reliable probabilities or numerical scoring factors for an inference rule, and that a radically different approach to the problem should be considered He essentially suggests that the expert be encouraged to think in terms of specific instances which would conflict with the general rule and to encode this knowledge explicitly. Methodologically this seems to be very appealing, and helps to make both explicit and rigorous some of the techniques currently used by knowledge engineers whm they encode and refine the expert's knowledge We would welcome comments and criticisms of this approach from those steeped in the practical issues of constructing large rule-based expert systems. Probabilistic rules and their variants have recently supported several successful applications of expert systems, in spite of the difficulty of committing informants to particular conditional probabilities or "certainty factors," and in spite of the experimentally observed insensitivity of system performance to perturbations of the chosen values Here we survey recent developments concerning reasoned assumptions which offer hope for avoiding the practical elusiveness of probabilistic rules while retaining theoretical power, for basing systems on the information unhesitatingly gained from expert informants, and reconstructing the entailed degrees of belief later @

The system consists of an overall agent architecture and five components within the architecture. The five components are (1) goaldirected learning (GDL), a decisiontheoretic method for selecting learning goals; (2) probabilistic bias evaluation (PBE), a technique for using probabilistic background knowledge to select learning biases for the learning goals; (3) uniquely predictive theories (UPTs) and probability computation using independence (PCI), a probabilistic representation and Bayesian inference method for the agent's theories; (4) a probabilistic learning component, consisting of a heuristic search algorithm and a Bayesian method for evaluating proposed theories; and (5) a decision-theoretic probabilistic planner, which searches through the probability space defined by the agent's current theory to select the best action. PAGODA's initial learning goal is just An autonomous agent must be able to select biases (Mitchell 1980) for new learning tasks as they arise. PBE uses probabilistic background knowledge and a model of the system's expected learning performance to compute the expected value of learning biases for each learning goal. The resulting expected discounted future accuracy is used as the expected value of the bias.

Today, mobile robotics is an increasingly important bridge between the two areas. It is advancing the theory and practice of cooperative cognition, perception, and action and serving to reunite planning techniques with sensing and real-world performance. Further, developments in mobile robotics can have important practical economic and military consequences. For some time now, amateurs, hobbyists, students, and researchers have had access to how-to books on the low-level mechanical and electronic aspects of mobile-robot construction (Everett 1995; McComb 1987). The famous Massachusetts Institute of Technology (MIT) 6.270 robot-building course has contributed course notes and hardware kits that are now available commercially and in the form of an influential book (Jones 1998; Jones and Flynn 1993).

The First International Workshop on Rough Sets: State of the Art and Perspectives was held on 2-4 September 1992 in Kiekrz, Poland. To stimulate the discussion, the participation was limited to 40 researchers who are involved in fundamental research in rough set theory and its extensions, logic for approximate reasoning, machine learning, knowledge representation and transfer, and applications of rough set methodology. The workshop focused primarily on applications of the basic idea of the approximate definition of a set and its consequences in other areas of science and engineering. Applications discussed at the workshop included machine learning, medical diagnosis, fault detection, medical image processing, neural net training, database organization, drug research, and digital circuit design. The workshop was the first international meeting of researchers working in this relatively new area. The approximate definition of a set in terms of lower and upper bounds, as introduced in the ...