Showing papers in "International Journal of Human-computer Studies \/ International Journal of Man-machine Studies in 1972"

Learning Strategies and Individual Competence.

Abstract: In a free-learning task, individuals are assigned to different categories of cognitive competence on the basis of the teaching strategy they adopt. Two major types are distinguished: holists or global learners and serialists or step by step learners, on the basis of analysis of the type of hypothesis that learners test in carrying out the task and a content analysis of the protocols produced when learners are asked to teach back what they have learned. For other tasks, using the same subjects, it is shown that teaching is most effective when the teaching materials are structured so as to match an individual's competence and the converse holds when there is mismatch. The results are interpreted for a theory of learning/teaching and the need for a language suitable for talking about strategies and subject matter structures is discussed.

From cohomology in physics to q-connectivity in social science

Abstract: If we wish to be “scientific” in our approach to the Social Sciences then, in some way, we must want to make these disciplines more like the Physical Sciences. But here we meet the immediate snag that our intuition runs away from the thought that, as individual people, we can be adequately described as so many electrons and protons. This also leads to the fear that perhaps Physics is not formulated in such a manner as to show us the heart of its methods, as opposed to the fruits of its labours. This paper is a review of a personal search over many years to find a formulation for physical science which would not do too much violence to accepted theories and yet show a way for its extension and generalization into fields of social science. Since the latter seems to require a language which mathematicians would call “combinatorial”, being concerned with finite sets, it became a search for a mathematical language which would describe certain key properties of the familiar continuum but which would carry these over when that continuum should be replaced by a finite set of points. The language which exhibits these properties is the twentieth-century development of algebraic topology (what used to be called combinatorial topology), and some of its basic concepts are referred to in the first half of the paper. From this point of view an indication of the role of the cocycle in physics is first developed (although many of the intriguing details are given elsewhere), and this is replaced by the same idea but referred to a relation between finite sets. The notion of the simplicial complex is then developed as the vehicle for that sense of structure which is inherent in either the laws of physics or the behaviour of social systems. The result is that one finds, when applying it to some specific field of enquiry, that it is peculiarly powerful for the representation of social or human activities. It leads too to a view of data which is structural (in a multi-dimensional space) in a way which the statistical view is unable to penetrate. Hopefully the method will help to develop new techniques for understanding, the data of relations.

Test Retest Reliability and Cost Analyses of Automated and Face to Face Intelligence Testing

Abstract: Two replication studies were reported and costs were compared for automated WAIS testing versus face to face WAIS testing. Experiment I replicated the Elwood & Griffin (in press) study. Two sessions of automated WAIS testing indicated high test retest reliabilities for intelligence measures (VIQ 0.94, PIQ 0.95 and FSIQ 0.97), and moderately high rs for some non-IQ measures of WAIS responses. The study was interpreted as a successful replication of the Elwood & Griffin (in press) experiment. Experiment II was interpreted as a successful replication of the Coons & Peacock (1959) study. Two sessions of face to face WAIS testing resulted in these test retest rs: VIQ 0·98, PIQ 0·94 and FSIQ 0·97. An analysis of the salary levels and manhours required for testing indicated that for each dollar spent for face to face WAIS testing by a Phi) psychologist, the same testing could be performed by a BA level technician for $0·51 and that a technician and clerk typist working together could administer and score the automated WAIS for $0·42.

Axioms for adaptive behaviour

Abstract: The study of adaptive behaviour, both natural and artificial, has developed without any generally accepted definitions of the terms “adaptive” or “adapted”, and it has been argued that a formal definition of these terms is not possible. However, if only for purposes of meaningful communication, some form of accepted definition would be useful, and it is possible that an investigation of the “impossibility” of definition might throw some light on fundamental problems concerning adaption and learning processes. This paper commences with a critical discussion of previous attempts to define terms concerned with adaption and learning. It is concluded that these terms cannot meaningfully be defined absolutely, and do not have a single connotation, but that rigorous definitions of a variety of phenomena with connotations of adaption may be given relative to (arbitrarily) defined rules for describing behaviour. An extensive set of definitions of behavioural phenomena in these terms is given. Finally a machine, the adaption automaton, is defined whose behaviour is equivalent to that of the adaptive system and the various definitions are re-phrased in terms of the structure of the adaption automaton.

Multi-dimensional structure in the game of chess

Abstract: This paper is an immediate follow-up of the author's previous “From Cohomology in Physics to q -connectivity in Social Science”, published in the previous issue of this journal. It is an illustration of the theory of structure contained in the first paper, being an application to the game of chess. The relations which exist between the chess-men and the squares of the chessboard, together with the rules of the game, are used to define mathematical relations Γ W and Γ B . The connectivity patterns in multi-dimensional spaces, which these relations define, are then obtained by the use of computer programs written in Basys. The accepted theories of positional chess are traced out and expressed in terms of the structure in Q -space; in particular, checkmate can be defined precisely in terms of the intersection of appropriate simplicial complexes. Finally, a detailed analysis of the so-called Immortal Game (that between Anderssen and Kieseritsky—London, 1851) is given in terms of connectivity patterns: the subordination of tactical play to positional judgement (the structure in Q -space) is illustrated by the variation of the Structure Vector. If chess masters base their profound positional judgements on an unconscious appraisal of the connectivity structure in Q -space—as the author believes—then this analysis holds out the prospect of a chess-playing computer using the same methods of assessment in its play.

Formants in automatic speech recognition

Abstract: This paper concerns the use of formant frequency information in automatic speech recognition. The discussion is addressed to the physical significance of the formant and to how this relates to the phonetic concepts of segment and equivalence that are needed for the recognition of phonetic types. Specifically, the definition of the phone in terms of articulatory dynamics can be interpreted acoustically in terms of formant dynamics. Hence formant :transition information can aid segmentation. Also, formant frequencies for given utterances by single speakers display remarkable interrepetition stability, while the speaker identity, phonetic type, and the phonetic, prosodic, and linguistic contexts are sources of nonrandom variability that should be included in a complete acoustic phonetic description of formant behavior.

Automated WAIS testing correlated with face-to-face WAIS testing: A validity study

Abstract: Automated WAIS results were correlated with criterion measures that consisted of face-to-face WAIS results. A counter-balanced procedure was used in which one group of 20 subjects received automated testing in the first session and then faceto-face testing in the second session while another group of 20 subjects received face-to-face testing first and then automated testing afterwards. The validity coefficients for IQ measures for the automated-first group and the face-to-face-first group were verbal IQ 0-92, 0-95; performance IQ 0-93, 0-88; and full scale IQ 0-95, 0-96. It was concluded that more research needed to be focused into this area to hasten the time when automated procedures would become the clinicalstandard methods for performing certain psychological testing functions.

Automated versus face-to-face intelligence testing: comparison of test-retest reliabilities

Abstract: It was hypothesized that automated WAIS testing would yield higher test-retest correlations ( rs ) than face-to-face WAIS testing. Two samples were obtained: 24 subjects received automated WAIS testing, and 29 subjects received the usual face-to-face WAIS testing. The test-retest rs for the two groups of subjects were compared for significance of differences. The results of the experiment were that the rs of the automated and face-to-face groups were not significantly different except on the Digit Span subtest on which the face-to-face group r was higher. It was concluded that, at this point in research, the strongest arguments for continued development of automated intelligence testing systems rested on considerations such as savings of professional man-hours, reduced costs of testing, and increases in the number of objective behavioral measurements that automated testing could make available for clinical decision making.

An Abbreviated Guide to Planning for Speech Interaction with Machines: The State of the Art.

Abstract: The paper provides an overview of the field of speech interaction with machines, both at the level of speech pattern discrimination and at the level of understanding and generating natural language. It is not an exhaustive report, space considerations among other factors rendering complete coverage impractical, though the references given will allow the interested reader to dig deeper. What is provided is a signposting of major recent highlights, to allow realistic appraisal of the current and probable future possibilities in voice interaction with machines. Those planning man-machine systems can no longer afford to ignore even the current possibilities.

Controlling the Learning of Diagnostic Tasks.

Abstract: A cardinal aim of computer-assisted learning should be to devise methods which optimize its techniques of instruction, and these require a precise representation of tasks. and learners, a set of teaching operations and means-ends guidance rules. An analysis of diagnostic tasks analogous to those carried out in medicine, points to ways in which computer programs can generate tasks to fit a particular learner's competence. Such methods can also give adaptive help in teaching, but there is a need for much experiment before satisfactory means-ends guidance rules can be formulated. For this purpose a computer-controlled diagnostic game was developed in which students attempted to classify patterns of symbols which represented sets of “attribute states” of various “diseases”. The characteristics of this game, and some analyses of student performances and some measures of task difficulty are described. Preliminary results of having the computer match student performances to its own diagnostic strategies are also outlined, and the development of these to control adaptive teaching is discussed.

Basic directions in automatic speech recognition

Abstract: This paper represents a view of basic problems in automatic speech recognition drawn from the vantage point of applied linguistics. The basic areas of speech production, articulatory phonetics, acoustic analysis, acoustic phonetics, and phonetic sequences of natural speech are crucial to flexible automatic speech recognition, especially for the long-range goal of recognizing continuous-flow large-vocabulary natural speech of different speakers. Articulatory phonetics provides a link between the physical events of speech and the elements of the phonological code. The processes of speech production are basic to articulatory phonetics while the acoustic theory of speech production provides a basis for the phonetic interpretation of the acoustic. speech waveform. Formant frequencies are particularly important acoustic parameters and recent improvements in formant tracking are encouraging for speech recognition. Even if complete phonetic recognition is realized, however, there remains the conversion of phonetic strings to lexical strings. This problem requires extensive phonetic analysis of actual spoken language.

Machine Intelligence: The Best of Both Worlds?

Abstract: Man and machine can be conveniently characterized by “slow, sloppy but brilliant” and “rapid, rigorous but stupid” respectively. This essay examines some of the available evidence to determine if there are any good reasons for presuming that an intelligent machine will, or will not, be able to exhibit “the best of both worlds”.

Spline function interpolation in interactive hemodynamic simulation

Abstract: A computer graphics system is being employed to simulate blood movement in several hemodynamic studies. Of particular interest is the representation of flow velocity profiles in an arterial channel past a lesion. The profiles are represented by smoothed curves as formed by spline function approximations. Resulting graphics displays show the advantages of spline functions in comparison to other approximation methods for fitting smooth curves to data. In turn, the interpolation of surfaces to boundary values by spline functions is studied. Such interpolation, in combination with certain advanced simulation equipment, is shown to be useful for studying the interim phases in the initiation and formation of atherosclerotic lesions.

Structural models of simple sensory motor co-ordination

Abstract: Current views of simple sensory-motor functions have it that sensory inputs serve to select (trigger) one of some collection of pre-wired motor responses. This paper compares several schemes which not only can select an appropriate class of prewired responses (say, “approach”), but also make a selective response directed at only one of several objects present. A classification of layered spatially arrayed nerve networks is used to constrain possible implementations of this selection function.

Developing Measures to Reveal Individual Styles in the Use of an Acceleration Control

Abstract: Three limited studies are described in which attempts were made to quantify some of the differences in style between individual operators of a simulated aircraft control. Two-dimensional tracking tasks (height and heading control) were presented on a CRT display. Even a straight-forward integral of control displacement over time can show differences between individuals within a group (for example R.A.F. squadron pilots). Other, less gross analyses may be required, however, if inter-operator differences are to be explained in terms less coarse than, say, a pilot-test pilot distinction.

A Criticism of Adaptive Neural Nets as Models of Perception

Abstract: Adaptive neural nets are collections of threshold logic units (TLU) connected by variable weights. They are often compared to biological systems, serving as models for memory and perception, and are frequently used for Optical Character Recognition (OCR). The Perceptron is the most commonly used model among these nets. Taking it as an example we will show it to be inadequate to account for perception. The following notation is used: “ φ,φ ....“ denotes individual characters φ,φ 2....; denotes the class of all characters φ, φ2.

Human behaviour in an interactive environment using a simple spoken word recognizer

Abstract: In 1970 a trial interactive computer program using a simple word recognizer was demonstrated to visitors at the Physics Exhibition. This program took the form of a speech-operated teaching machine where the visitors were instructed by the machine on how to use it. This paper describes a simple contextual model ( a priori heuristics) used by the teaching program. This model was used to detect two types of behaviour which modified the response of the machine. From analysis of the behaviour of 134 visitors who used the machine, modifications to these rules are given ( a posteriori heuristics).

Descriptions and plans in an interactive robot simulation system

Abstract: The problem of designing a robot-controller is approached by taking a simplified computer-simulated, model of a robot in an environment, and writing programs to enable the robot to move around its environment in a reasonably intelligent manner. At no point is mathematical logic used. The problems of concept representation and the creation and execution of plans are dealt with in this simple system, and some attention is paid to the problem of exploration. ROSS, an interactive computer program which simulates the robot-environment model, is described. A command language allows the user to specify tasks for the robot at various conceptual levels. The main contribution is showing how, in a rectanguloid two-dimensional environment, a robot's model of his world may be represented as a graph, and how the robot should use this graph to create and execute plans of action. Also, two new algorithms are presented for handling two-dimensional rectanguloid objects (shapes for short), and the concept of a “maximal subrectangle” of such a shape is introduced. The algorithm DECOMP decomposes or “parses” a shape into its maximal subrectangles, while the algorithm CONTAIN compares a pair of shapes and decides whether one of the shapes could be moved to fit inside the other. Several problems are listed concerning the ways in which a robot might explore, represent, and make plans about its environment, most of which are amenable to direct attack in this simplified model. Finally, theoretical questions concerning two-dimensional rectanguloid shapes are raised.

On a class of automata models of learning machines

Abstract: A class of machines modelled as deterministic Moore automata is proposed. Particular attention is given to investigating their capability of exhibiting learning behaviour assessed by an associated index of performance. This is done by assuming that the automata are operating in a random environment about which they have no a priori knowledge.

An iterative array which can represent the rotation and translation of objects

Abstract: An iterative array of sequential logic circuits (cells) is described, which can represent the motions of finite-sized rigid objects. Images of the objects are formed within the array, for example by a lens system and photocell matrix. Subsequently the images can be moved, together or independently, in response to external command signals. Any combination of rotation and combination can be induced, continuously and at any speed, so that all possible motions of objects in a particular plane can be represented. Depth-information can be attached to each image-point, so that three-dimensional space can also be represented. Each imagepoint can be moved in any direction, and its location is defined not only by the cell it occupies, but also within the cell: the array space is truly Euclidean, in contrast with the tessellation spaces of earlier arrays. Each image-point can be labelled: images can therefore be continually monitored, by treating the array as a randomaccess store with the cell labels as contents addresses. An application in production-line automation, is proposed, and relevance to the investigation of human and machine intelligence is suggested.

Prominence Patterns in Air Traffic Control Messages

Abstract: Recordings of messages spoken by Air Traffic Controllers have been analysed subjectively and objectively in an attempt to obtain acoustic correlates of the subjective measures “syllabic prominence” and “word prominence”. A computer program has been written that assigns syllabic prominence at two levels with fair success. The repercussions of this on the recognition of continuous speech by machine are discussed. Word prominence depends on the complex inter-relationship of many speech features; is difficult to assign by rule and of less value to speech recognition by machine.

An interactive speech processing system using a large computer

Abstract: A speech processing system which includes facilities for the analysis and synthesis of speech and on-line input and output of speech waveforms has been developed at the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Division of Computing Research. The system is thoroughly interactive and works within the DAD operating system of the Control Data 3600 using a POP-8 computer for real-time on-line interaction with the external world. This paper describes the design and implementation of the system and puts it in the context of the literature on computer processing of speech. The interactive control of processing together with audio and graphic monitors of processed speech data give great potential for a wide range of research on speech and man-machine communication using speech.