Current AI Research Articles

An approach to assessment of plant man-machine systems by computer simulation of an operator's cognitive behavior

Computer simulation of an operator's cognitive behavior is a promising approach for the purpose of human factors study and man-machine systems assessment. In this paper an architecture of the simulation is proposed, based on the current AI technologies. The simulation system has been constructed as a knowledge-based system of the blackboard model, which can represent well the revisable nature of human thought, by use of the Truth Maintenance System. Assessment of cognitive workload imposed on an operator was attempted by scoring information-processing activities of this system. The scoring was done separately for each type of activity. A test simulation was performed on some cases of nuclear power plant operation under abnormal plant conditions. In the case study, investigation of the mechanism of mistakes and assessment of the cognitive workload were performed.

AI research and applications in Digital's service organization

The Digital Services Research Group and its predecessor groups and offshoots in Digital Equipment Corporation have been mobilizing leading-edge AI research to bear on real-life problems that face the corporation and its customers. The general strategy of the group is to explore emerging techniques relevant to service and support needs through developing rapid prototypes, deploying these prototypes, and incorporating feedback from users. With over 32 major projects undertaken during the past decade, we have worked on broad spectrum of problems and explored a variety of advanced AI techniques. This article describes the current AI activities in five areas: (1) enterprise advisory systems, (2) natural language processing and textual information retrieval, (3) largescale knowledge base management and access, (4) software configuration management, and (5) intrusion detection. We also list some future research directions.

Emergent functionality among intelligent systems: Cooperation within and without minds

In this paper, the current AI view that emergent functionalities apply only to the study of subcognitive agents is questioned; a hypercognitive view of autonomous agents as proposed in some AI subareas is also rejected. As an alternative view, a unified theory of social interaction is proposed which allows for the consideration of both cognitive and extracognitive social relations. A notion of functional effect is proposed, and the application of a formal model of cooperation is illustrated. Functional cooperation shows the role of extracognitive phenomena in the interaction of intelligent agents, thus representing a typical example of emergent functionality.

An experimental study of criteria for hypothesis plausibility

Abstract Abductive diagnostic problem-solving systems use causal relations to infer plausible diagnostic hypotheses. An important but controversial issue for such models is what characteristics should define the most plausible hypotheses. While there are theoretical predictions relevant to this issue, there are almost no empirical data on which to base rational decisions. Accordingly, this study examines four different criteria of hypothesis plausibility in diagnosing the site of brain damage in 100 medical patients. The criteria examined are (1) naive minimal cardinality, (2) irredundancy, (3) most probable (Bayesian), and (4) minimal cardinality when adjacency relations are taken into account. Model performance when these different hypothesis plausibility criteria are used confirms the previously predicted inadequacy of minimal cardinality. It also indicates that irredundancy (‘minimality’), the criterion most widely used in current AI models, is not useful in this setting because of the large number of...

Human intelligence and AI: how close are we to bridging the gap? (open line)

A few issues that the author considers of central importance are identified, and their implications for the central question of how relevant current AI studies are to understanding the architecture of human intelligence are analyzed. A plausible strategy for inferring the architecture of human intelligence is outlined and used as a framework for discussing the issues raised by the author. It is concluded that we are far from bridging the gap between human intelligence and AI, because we have not been asking the right questions or tackling the real issues.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Self-explanations: How students study and use examples in learning to solve problems

The present paper analyzes the self-generated explanations (from talk-aloud protocols) that “Good” and “Poor” students produce while studying worked-out examples of mechanics problems, and their subsequent reliance on examples during problem solving. We find that “Good” students learn with understanding: They generate many explanations which refine and expand the conditions for the action parts of the example solutions, and relate these actions to principles in the text. These self-explanations are guided by accurate monitoring of their own understanding and misunderstanding. Such learning results in example-independent knowledge and in a better understanding of the principles presented in the text. “Poor” students do not generate sufficient self-explanations, monitor their learning inaccurately, and subsequently rely heavily on examples. We then discuss the role of self-explanations in facilitating problem solving, as well as the adequacy of current AI models of explanation-based learning to account for these psychological findings.

Self‐Explanations: How Students Study and Use Examples in Learning to Solve Problems

The present paper analyzes the self‐generated explanations (from talk‐aloud protocols) that “Good” and “Poor” students produce while studying worked‐out examples of mechanics problems, and their subsequent reliance on examples during problem solving. We find that “Good” students learn with understanding: They generate many explanations which refine and expand the conditions for the action parts of the example solutions, and relate these actions to principles in the text. These self‐explanations are guided by accurate monitoring of their own understanding and misunderstanding. Such learning results in example‐independent knowledge and in a better understanding of the principles presented in the text. “Poor” students do not generate sufficient self‐explanations, monitor their learning inaccurately, and subsequently rely heavily on examples. We then discuss the role of self‐explanations in facilitating problem solving, as well as the adequacy of current AI models of explanation‐based learning to account for these psychological findings.

Logical and decision-theoretic methods for planning under uncertainty

Decision theory and nonmonotonic logics are formalisms that can be employed to represent and solve problems of planning under uncertainty. We analyze the usefulness of these two approaches by establishing a simple correspondence between the two formalisms. The analysis indicates that planning using nonmonotonic logic comprises two decision-theoretic concepts: probabilities (degrees of belief in planning hypotheses) and utilities (degrees of preference for planning outcomes). We present and discuss examples of the following lessons from this decision-theoretic view of nonmonotonic reasoning: (1) decision theory and nonmonotonic logics are intended to solve different components of the planning problem; (2) when considered in the context of planning under uncertainty, nonmonotonic logics do not retain the domain-independent characteristics of classical (monotonic) logic; and (3) because certain nonmonotonic programming paradigms (for example, frame-based inheritance, nonmonotonic logics) are inherently problem specific, they might be inappropriate for use in solving certain types of planning problems. We discuss how these conclusions affect several current AI research issues.

Progress and challenges in the application of artificial intelligence to computational fluid dynamics

An approach to analyzing CFD knowledge-based systems is proposed which is based, in part, on the concept of knowledge-level analysis. Consideration is given to the expert cooling fan design system, the PAN AIR knowledge system, grid adaptation, and expert zonal grid generation. These AI/CFD systems demonstrate that current AI technology can be successfully applied to well-formulated problems that are solved by means of classification or selection of preenumerated solutions.

AI in the German Democratic Republic

Artificial Intelligence has not only experienced a dramatic growth in West Germany, but also to a certain extent in the German Democratic Republic. Some research institutions, such as the Academy of Sciences of the GDR and VEB Robotron, have been working in the area of AI since as long ago as the early or mid-seventies. Many of their results are also known in the West, e.g. through the Automated Language Analysis congresses that are commonly held every two years. In the last few years, a great number of AI activities at other places have joined these “classic” research institutions. In addition, a form of research on a GDR-wide basis has recently come to the fore. During a lecture tour to most of the relevant AI institutions, I had the opportunity to learn about the state-of-the-art of AI research in East Germany. The following report is certainly not exhaustive, but it covers the major current AI activities in this country. The details of individual research projects cannot be given, but the individual scientists to whom I refer will probably supply more information.

“How”—beyond the “what”, towards the “why”: A rule-assessment approach to achievement testing

The advantages of a rule assessment approach to the interpretation of achievement test results have been demonstrated using an S-P chart with coded error types. The problems of similar total test scores resulting from completely different misapprehensions, as well as correct answers resulting from incorrect rules of operation, were addressed using a simulated data-set. Although the overall quality of the test used here as measured by conventional psychometric indices proved satisfactory, it was shown that the traditional interpretation, which refers to total test scores, can be misleading, especially when adaptive remediation is sought. It is well known in medical sciences that a disease has several symptoms yet several diseases can share the same symptoms (i.e. high fever). Consequently, no responsible physician would prescribe the same medicine for two patients suffering from different diseases just because they both share high fever as one of their symptoms. Similarly, when two students with different misapprehensions get the same total test score, should the teacher prescribe the same remediation for correcting their misapprehension? Although the method for diagnostic test construction was out of the scope of this paper, it should be noted that test design is a crucial matter which eventually determines the quality of the diagnosis. One has to, therefore, carefully choose the items for the diagnosis in order to maximize the information about the rules of operation underlying the students' responses. A task specification chart (Birenbaum & Shaw, 1985) may serve as a useful tool in the process of test construction. As was illustrated in the chart, when an item yields the same results as a result of various “bugs”, its contribution to rule assessment is in question. Although in reality test results are contaminated by noise resulting from careless errors or strategy changes during the test, the overall identification rate achieved by diagnostic tests ranges between 70%–80% (Tatsuoka, 1984). Similarly, current AI diagnostic systems such as DEBUGGY and DPF are reported as being capable of identifying 80%–90% of student errors (VanLehn, 1981; Ohlesson & Langley, 1985). It seems that such a rate justifies the tedious work involved in constructing a diagnostic tool.

Toward Better Models of the Design Process

What are the powerful new ideas in knowledge based design? What important research issues require further investigation? Perhaps the key research problem in AI-based design for the 1980's is to develop better models of the design process. A comprehensive model of design should address the following aspects of the design process:the state of the design ; the goal structure of the design process;design decisions; rationales for design decisions; control of the design process; and the role of learning in design. This article presents some of the most important ideas emerging from current AI research on design especially ideas for better models design. It is organized into sections dealing with each of the aspects of design listed above.

Artificial Intelligence Research in Statistics

The initial results from a few AI research projects in statistics have been quite interesting to statisticians: Feasibility demonstration systems have been built at Stanford University, AT-T bell Laboratories, and the University of Edinburgh. Several more design studies have been completed. A conference devoted to expert systems in statistics was sponsored by the Royal Statistical Society. On the other hand, statistic as a domain may be of particular interest to AI researchers, for it offers both tasks well suited to current AI capabilities and tasks requiring development of new AI techniques.

Reviews of "Artificial Intelligence, A New Tool for Industry and Business by Richard K. Miller", SEAI Institute, ISBN 0-89671-051-3.

A measure of the business interest in AI is the number of sources of information geared for management. This report is a review of the capabilities of current AI systems especially in how they an be used for industrail automation and management. Most of the information is readily available elsewhere, but not in a single source. The discussion of varous AI research projects concentrates on what the system did rather than on how the system did it. Actual products are discussedd, with some details, and sources for more information are given. Since the report is meant for managers rather than researchers, systems which produce tangible results and those in actual sue (e.g., DELTA at GE, XCON at DEC, DENDRAL) are given more discussion than pure research efforts.

Partial Evaluation, Programming Methodology, and Artificial Intelligence

This article presents a dual dependency between AI and programming methodologies. AI is an important source of ideas and tools for building sophisticated support facilities which make possible certain programming methodologies. These advanced programming methodologies in turn can have profound effects upon the methodology of AI research. Both of these dependencies are illustrated by the example of anew experimental programming methodology which is based upon current AI ideas about reasoning, representation and control. The manner in which AI systems are designed, developed and tested can be significantly improved in the programming is supported by a sufficiently powerful partial evaluator. In particular, the process of building levels of interpreters and of intertwining generate and test can be partially automated. Finally speculations about a more direct connection between AI and partial evaluation are presented.

Artificial Intelligence, Psychology, and the Philosophy of Discovery

In their contributions to this volume, Bruce Buchanan (1983) and Lindley Darden (1983) have provided compelling reasons why philosophers of science concerned with the nature of scientific discovery should be aware of current work in artificial intelligence. Buchanan presents abundant examples of the mechanization of generation of empirical hypotheses in current AI programs, while Darden, at a more speculative level, discusses the use of analogy in theoretical discovery. The implication of both these studies is that the philosophy of discovery can profitably borrow notions from work in artificial intelligence, since useful analogies can be found between problems in that field and the problems facing philosophers studying scientific discovery.I have no quarrel with this claim, except that it is too weak. I shall argue that work in AI is not only useful but essential for investigations in philosophy of science concerning discovery. The connecting link between the two fields is empirical cognitive psychology.

A computer model of conversation

This project is concerned with the modelling of language use in a conversation. The aim is to develop techniques for making a machine a more helpful conversant. We are examining problems that have essentially been ignored in current AI natural language research. For instance, our model will have strategies for deciding what to say, for inferring a speaker's intentions, and for being helpful. In order to do this, the system maintains an explicit model of the user's beliefs which includes the user's model of the system's beliefs, etc. This technique is implemented by using Hendrix' partitioned semantic nets.

The QP Triangle: A Graphical Method for Bias Design

A graphical design procedure is presented based upon a QP triangle for the design of bipolar transistor (BJT) bias circuits. The design technique stresses quiescent point (QP) location on the IC-VCE characteristics and considers the effects of simultaneous variations in the BJT parameters hFE, VBE, and ICO upon QP location. The QP triangle method is developed for the standard one-battery BJT CE stage discussed in many introductory electronic circuits textbooks. The QP triangle method is applied to a specific CE stage which has to meet certain design specifications. One important specification is that the circuit must operate over the temperature range 25-100°C with silicon N-P-N BJT's having values of h<FE (25°C) > 40 and hFE 100°C) < 200. The available tradeoffs between the peak-to-peak voltage Vpp and the current gain AI are stressed and the best available design is selected. The performance of the selected design was simulated on a digital computer and measured in the laboratory. Both the computer simulation and the experiment are in good agreement with the design. The QP triangle method has been used in an introductory electronic circuits course with success for several years. Students understand this graphical design procedure and are able to apply it. It is recommended for beginning electronics students. An interactive computer program AMPDSN to aid students and instructors design the standard one-battery BJT CE stage is also described. An algorithm based upon the QP triangle is used. The program language is Basic.

The Role Of Perception And Action In Intelligent Systems

Robotics plays an important role in the applications of Artificial Intelligence in Engineering, since it deals with the interaction of an intelligent system with the world by means of perception and action. However, Robotics has been traditionaly considered as just a mere application area of AI. In this paper, the role of perception and action in current AI systems is analyzed and some recent fundamental results concerning the use of Robotic Intelligence for applications of AI in Engineering are discussed, in order to build artificial systems that behave in an intelligent way in the real world. Finally, two implemented systems based on this methodology are described to show the relevance of the proposed strategy to Engineering applications, one using visual perception, and another based on force/torque sensing.