Stuart E. Dreyfus

An Appraisal of Some Shortest-Path Algorithms

This paper treats five discrete shortest-path problems: 1 determining the shortest path between two specified nodes of a network; 2 determining the shortest paths between all pairs of nodes of a network; 3 determining the second, third, etc., shortest path; 4 determining the fastest path through a network with travel times depending on the departure time; and 5 finding the shortest path between specified endpoints that passes through specified intermediate nodes. Existing good algorithms are identified while some others are modified to yield efficient procedures. Also, certain misrepresentations and errors in the literature are demonstrated.

The Five-Stage Model of Adult Skill Acquisition.

The following is a summary of the author’s five-stage model of adult skill acquisition, developed in collaboration with Hubert L. Dreyfus. An earlier version of this article appeared in chapter 1 of Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer (1986, Free Press, New York).

Peripheral Vision Expertise in Real World Contexts

In this paper we describe a five-stage phenomenological model of skill acquisition, of which expertise is the highest stage. Contrary to the claims of knowledge engineers, we argue that expertise in general, and medical expertise in particular, cannot be captured in rule-based expert systems, since expertise is based on the making of immediate, unreflective situational responses; intuitive judgment is the hallmark of expertise. Deliberation is certainly used by experts, if time permits, but it is done for the purpose of improving intuition, not replacing it. The best way to avoid mistakes is to take responsibility for them when they occur, rather than try to prevent them by foolproof rules. In bureaucratic societies, however, there is the danger that expertise may be diminished through over-reliance on calculative rationality.

Functional approximations and dynamic programming

Abstract : This paper indicates some ways in which the theory of approximation can be used to increase the range of present day computers. Although the primary interest is in applying these techniques to the functional equations occurring in the theory of dynamic programming. These same methods are applicable, and even more readily, to the classical functional equations of mathematical physics. The objective of the paper is to trade additional computing time, which is expensive, for additional memory capacity, which does not exist.

Towards a phenomenology of ethical expertise

ogy, in both its transcendental and existential versions, has made immense contributions to metaphysics, epistemology and the philosophy of action and mind. The same cannot be said of its contribution to ethics. With the exception of Sartre, phenomenologists have had little to say about ethics, and what Sartre has said has had little effect on the course of the subject, perhaps because Sartre takes following moral principles to be a form of inauthenticity. Our hypothesis is that if one returned to the phenomenon and tried to give a description of ethical experience one might find that phenomenology has a great deal to contribute to contemporary debate, particularly since the focus of discussion has shifted from interest in meta-ethical issues to a debate between those who demand a detached critical morality based on principles that tells us what is right and those who defend an ethics based on involvement in a tradition that defines what is good. This new confronta? tion between Kant and Hegel, between Moralit?t and Sittlichkeit, has produced two camps which can be identified with J?rgen Habermas and John Rawls on the one hand, and Bernard Williams and Charles Taylor on the other. The same polarity appears in feminism where the Kohlberg scale, which defines the highest stage of moral maturity as the ability to stand

Richard Bellman on the Birth of Dynamic Programming

W follows concerns events from the summer of 1949, when Richard Bellman first became interested in multistage decision problems, until 1955. Although Bellman died on March 19, 1984, the story will be told in his own words since he left behind an entertaining and informative autobiography, Eye of the Hurricane (World Scientific Publishing Company, Singapore, 1984), whose publisher has generously approved extensive excerpting. During the summer of 1949 Bellman, a tenured associate professor of mathematics at Stanford University with a developing interest in analytic number theory, was consulting for the second summer at the RAND Corporation in Santa Monica. He had received his Ph.D. from Princeton in 1946 at the age of 25, despite various war-related activities during World War II—including being assigned by the Army to the Manhattan Project in Los Alamos. He had already exhibited outstanding ability both in pure mathematics and in solving applied problems arising from the physical world. Assured of a successful conventional academic career, Bellman, during the period under consideration, cast his lot instead with the kind of applied mathematics later to be known as operations research. In those days applied practitioners were regarded as distinctly second-class citizens of the mathematical fraternity. Always one to enjoy controversy, when invited to speak at various university mathematics department seminars, Bellman delighted in justifying his choice of applied over pure mathematics as being motivated by the real world’s greater challenges and mathematical demands. Following are excerpts, taken chronologically from Richard Bellman’s autobiography. The page numbers are given after each. The excerpt section titles are mine. These excerpts are far more serious than most of the book, which is full of entertaining anecdotes and outrageous behaviors by an exceptionally human being.

The Ethical Implications of the Five-Stage Skill-Acquisition Model

We assume that acting ethically is a skill. We then use a phenomenological description of five stages of skill acquisition to argue that an ethics based on principles corresponds to a beginner’s reliance on rules and so is developmentally inferior to an ethics based on expert response that claims that, after long experience, the ethical expert learns to respond appropriately to each unique situation. The skills model thus supports an ethics of situated involvement such as that of Aristotle, John Dewey, and Carol Gilligan against the detached, rationalist ethics of Kant, John Rawls, Lawrence Kohlberg, and Jürgen Habermas.

A Customized Neural Network for Sensor Fusion in On-Line Monitoring of Cutting Tool Wear

A customized neural network for sensor fusion of acoustic emission and force in on-line detection of tool wear is developed. Based on two critical concerns regarding practical and reliable tool-wear monitoring systems, the minimal utilization of «unsupervised» sensor data and the avoidance of off-line feature analysis, the neural network is trained by unsupervised Kohonens Feature Map procedure followed by an Input Feature Scaling algorithm. After levels of tool wear are topologically ordered by Kohonens Feature Map, input features of AE and force sensor signals are transformed via Input Feature Scaling so that the resulting decision boundaries of the neural network approximate those of error-minimizing Bayes classifier. In a machining experiment, the customized neural network achieved high accuracy rates in the classification of levels of tool wear. Also, the neural network shows several practical and reliable properties for the implementation of the monitoring system in manufacturing industries

From Socrates to expert systems: The limits of calculative rationality☆

This paper examines the general epistemological assumptions of artificial intelligence technology and recent work in the development of expert systems. These systems are limited because of a failure to recognize the real character of expert understanding, which is acquired as the fifth stage of a five-step process. A review of the successes and failures of various specific expert system programs confirms this analysis.

What artificial experts can and cannot do

Ones model of skill determines what one expects from neural network modelling and how one proposes to go about enhancing expertise. We view skill acquisition as a progression from acting on the basis of a rough theory of a domain in terms of facts and rules to being able to respond appropriately to the current situation on the basis of neuron connections changed by the results of responses to the relevant aspects of many past situations. Viewing skill acquisition in this ways suggests how one can avoid the problem currently facing AI of how to train a network to make human-like generalizations. In training a network one must progress, as the human learner does, from rules and facts to wholistic responses. As to future work, from our perspective one should not try to enhance expertise as in traditional AI by attempting to construct improved theories of a domain, but rather by improving the learners access to the relevant aspects of a domain so as to facilitate learning from experience.