HCI Bibliography Home | HCI Journals | About IJMMS | Journal Info | IJMMS Journal Volumes | Detailed Records | RefWorks | EndNote | Hide Abstracts
IJMMS Tables of Contents: 2324252627282930313233343536373839

International Journal of Man-Machine Studies 33

Editors:B. R. Gaines; D. R. Hill
Dates:1990
Volume:33
Publisher:Academic Press
Standard No:ISSN 0020-7373; TA 167 A1 I5
Papers:38
Links:Table of Contents
  1. IJMMS 1990 Volume 33 Issue 1
  2. IJMMS 1990 Volume 33 Issue 2
  3. IJMMS 1990 Volume 33 Issue 3
  4. IJMMS 1990 Volume 33 Issue 4
  5. IJMMS 1990 Volume 33 Issue 5
  6. IJMMS 1990 Volume 33 Issue 6

IJMMS 1990 Volume 33 Issue 1

Listening Typewriter Simulation Studies BIBA 1-19
  A. F. Newell; J. L. Arnott; K. Carter; G. Cruickshank
In order to investigate the acceptability of automatic speech recognition systems in a creative writing task, experiments have been performed using a simulation in which a human operator is used to convert the speech signal into orthography. Pilot experiments investigated the appropriateness of natural dialogue in such a task. The major experiment was a partial replication of Gould, Conti and Hovanyecz (1983) "Listening Typewriter" simulation experiment. In contrast to Gould et al., however, a machine shorthand transcription system was used rather than a QWERTY keyboard. This ensured that the simulation of the speech recognizer was not restricted by the speed at which the operator could enter text. Also an important variable in the experiment was whether or not the subjects had been made aware that they were using a simulation and not a fully automatic speech recognition machine.
Supervised Learning in N-Tuple Neural Networks BIBA 21-40
  J. R. Doyle
An N-tuple Neural Network (NNN) is described in which each node fires selectively to its own table of binary trigger patterns. Each node receives input from k input terminals. Supervised learning is used with specially constructed problems: the system is taught to map specific instances of an input set onto specific instances of an output set. Learning is achieved by: (1) calculating a global error term (how far the set of actual outputs differs from the desired set of outputs); (2) either changing the connections between input terminals and N-tuple nodes, or by changing the trigger patterns that the node fires to; (3) re-calculating the global error term, and retaining the changes to the network if the error is less than in (1). Steepest descent optimisation described in (3), is compared with simulated annealing optimisation. Simulated annealing gives better solutions. Other results are that as connectivity k increases the number of possible solutions increases, but the number of possible non-solutions increases even faster. Simulated annealing is particularly helpful when the relative difficulty (ratio of search to solution) increases. In randomly chosen network configurations there is less entropy in the output than there is in the input to the system. When output is re-cycled as input, NNN either cycles or reaches an end-point. When solving complex I/O maps the system counteracts this trend by systematically increasing its sensitivity. Predicativity can be improved by combining the results of two or more independent NNN models.
Font Recognition by a Neural Network BIBA 41-61
  Ming-Chih Lee; William J. B. Oldham
Two neural network models, labelled Model H-H1 and model H-H2 by Hogg and Huberman have been successfully applied to recognize 26 English capital letters, each with six font representations. These two models are very similar, but Model H-H2 has the capability for modification of the basins of attraction during the training phase, whereas Model H-H1 does not. This appears to be a desirable feature for a neural network. It is shown in this work that this is indeed true. In either model, it is difficult to find a single set of parameters for one network or memory that can distinguish all of the characters. Therefore, a cascade of memories is utilized. Thus, in the training phase, a decision tree is built by cascading the memory matrices that represent the models. That is successive layers of refinement in selection of basins of attraction are used to generate output patterns unique to each input pattern. In the recognition phase, the subject characters are recognized by searching in the tree. Model parameters such as memory array size, Smin Smax, and Mmin Mmax were varied to elucidate the models' behavior. It is shown that there exist parameter values for both models to achieve a 100% recognition rate when all six fonts are used both as the training and the recognition set. Model H-H2 significantly outperformed Model H-H1 in terms of recognition rate, use of memory space, and learning speed when all six fonts were used as the training set.
An Algorithm for Segmenting Handwritten Postal Codes BIBA 63-80
  Michel Cesar; Rajjan Shinghal
Postal codes, although known by different names, are used in many countries to uniquely identify a location. In automated mail sorting, these postal codes are recognized by optical character readers. One problem faced for the recognition of handwritten postal codes is that of segmentation; each character in the postal code should ideally be assigned a different segment. Difficulties arise in segmentation because of the writing habits of people: characters may be broken, may touch one another, or may overlap one another. In this paper we present an algorithm to segment handwritten postal codes. The heuristic components of the algorithm were developed after analysing the writing habits of the people. Experimental results with handwritten Canadian postal codes show the algorithm to be effective, robust, and general enough that it can be suitably adapted for the postal codes of many other countries.
The Impact of Pascal Education on Debugging Skill BIBA 81-95
  Dan N. Stone; Eleanor W. Jordan; M. Keith Wright
Education in the Pascal programming language has been touted as a means of learning structured programming principles. This paper reports the results of two experiments that tested the effect of Pascal education on the debugging skills of novice programmers under timed test conditions. Results of both experiments indicate (1) that Pascal education is a better predictor of debugging performance than major, previous COBOL education, number of computer courses taken, or professional programming experience, and (2) that novice programmers who have studied Pascal demonstrate superior debugging performance regardless of program structure. Measures of program comprehension used in Experiment Two suggest that Pascal education may improve debugging performance by increasing the comprehension of program goals and plans. These results suggest that the value of structured programming techniques may be realized more in the programmer's way of thinking about a program than in the creation of a structured program per se.
Individual Differences and Effective Learning Procedures: The Case of Statistical Computing BIBA 97-119
  Alison J. K. Green; Kenneth J. Gilhooly
This paper reports two experiments examining individual differences in procedures for learning to use MINITAB (1982 edition). In Experiment 1, ten novices provided think aloud protocols over five sessions of learning to use MINITAB. On the basis of overall performance, novices were divided into two groups of five faster and five slower learners. The protocols suggested that subjects used two major classes of learning strategy: learning by doing and learning by thinking. Each class of learning strategy comprised a set of learning procedures. Differences in procedure usage were confined primarily to the learning by doing procedures. Faster learners used the mapping and exploratory procedures more frequently than slower learners, paid more attention to prompts and error messages and acted appropriately on evaluation feedback. In contrast, slower learners used the trial and error and repetition procedures more frequently than faster learners. In Experiments 2, 26 novices were allocated to one of three experimental groups. Group 1 received no instruction, Group 2 received instruction in the use of ineffective procedures (those procedures that did not serve to differentiate between fast and slow learners in Experiment 1) and Group 3 received instruction in the use of effective procedures (those procedures that did differentiate between fast and slow learners in Experiment 1). The overall performance of the effective procedures groups was significantly better than either the control or the ineffective procedures groups. Effective procedure usage ratings correlated significantly with overall performance. The negative correlations between effective procedure usage ratings and both requests for help and mean time to complete the MINITAB tasks were significant. Finally, on a free recall task, the effective procedures group remembered significantly more of the procedures they had been taught than the ineffective procedure groups.

IJMMS 1990 Volume 33 Issue 2

An Introduction to Hypermedia Issues, Systems and Application Areas BIBA 121-147
  John A. Begoray
This article is intended to be a general introduction to the new information representation technology: hypermedia. There is a lack of consensus as to a specific definition of hypermedia. Several general characteristics and specific terms are, however, emerging and are presented here as an introduction to the area.
   A number of design issues associated with hypermedia systems have been identified. This survey presents these issues and discusses a number of the more pre-eminent hypermedia systems within the context of the issues they address.
   From one perspective, hypermedia is not an application. It is, instead, a technology which can be used to develop and enhance many application areas. Hypermedia's potential contribution to some of these areas is discussed and some general conclusions presented.
A Database Interface Based on Montague's Approach to the Interpretation of Natural Language BIBA 149-176
  R. A. Frost; W. S. Saba
In this paper we describe a database interface that is loosely based upon some of the concepts proposed by Richard Montague in his approach to the interpretation of natural language. The system is implemented as an executable attribute grammar specified in a higher order, lazy, pure functional programming language. The attribute grammar formalism provides a simple means of implementing Montague's notion of "semantic rule to syntactic rule correspondence" and the higher order functional language in which the attributes grammar is constructed provides an appropriate vehicle for implementing Montague's higher order semantics. The purpose of the paper is two-fold: (i) to demonstrate that many of Montague's ideas can be used to advantage in creating natural language interfaces to databases, and (ii) to introduce a method for implementing attribute grammars in functional languages that is suitable for investigating both grammars and semantic theories of language.
Validation of Intent Inferencing by a Model-Based Operator's Association BIBA 177-202
  Patricia M. Jones; Christine M. Mitchell; Kenneth S. Rubin
OFMspert (Operator Function Model expert system) is an architecture for an intelligent operator's associate. The function of such an associate is to provide intelligent assistance for the human operator of a complex dynamic system. The basis for intelligent, context-sensitive advice and reminders is the ability of the associate to infer likely operator intentions in real time. This paper describes the implementation and validation of OFMspert's intent inferencing capability. In particular, a two-stage methodology for validation is proposed. This methodology is then used in the experimental validation of OFMspert's intent inferencing.
Utilizing a Process Model During the Acquisition of Redesign Knowledge BIBA 203-226
  Evangelos Simoudis
The paper presents REKL, a learning component which interacts with TLTS, a knowledge-based tool that uses a model for redesign in the domain of printed circuit boards. The model has been implemented around a blackboard architecture. REKL acquires four types of redesign knowledge: (meta-knowledge, eligibility demons, redesign operator scheduling rules, and clusters of redesign operators). The last three types of knowledge are organized by REKL into redesign knowledge sources. REKL acquires new redesign knowledge sources by compiling the knowledge that is contained in redesign plans that the user creates in cooperation with TLTS. It acquires meta-knowledge for selecting among the redesign knowledge sources, by interacting with the user and TLTS.
A Comparison of Formal Knowledge Representation Schemes as Communication Tools: Predicate Logic vs Semantic Network BIBA 227-239
  John T. Nosek; Itzhak Roth
An experiment was conducted to test the effectiveness of two popular knowledge representation schemes as communication vehicles between the human expert and the knowledge engineer. Validation by the human expert of the knowledge encapsulated depends upon how well the expert understands and interprets a representation scheme. A between-group experiment was conducted. Each group received two treatments of the same representation technique, with the second treatment slightly more complex that the first. All the scores for the Semantic Network representations were higher than that obtained for the Predicate Logic representations; and the Semantic Networks were clearly better for comprehension and conceptualization tasks. The results demonstrate some of the weaknesses of Predicate Logic and some of the strengths of Semantic Networks as communication tools during the validation process.

IJMMS 1990 Volume 33 Issue 3

Categories of Programming Knowledge and Their Application BIB 241-246
  Ruven Brooks
More or Less Following a Plan During Design: Opportunistic Deviations in Specification BIBA 247-278
  Willemien Visser
An observational study was conducted on a mechanical engineer throughout his task of defining the functional specifications for the machining operations of a factory automation cell. The engineer described his activity as following a hierarchically structured plan. The actual activity is in fact opportunistically organized. The engineer follows his plan as long as it is cognitively cost-effective. As soon as other actions are more interesting, he abandons his plan to proceed to these actions. This paper analyses when and how these alternative-to-the-plan actions come up. Quantitative results are presented with regard to the degree of plan deviation, the design components and the definitional aspects which are most affected by these deviations, and the deviation patterns. Qualitative results concern their nature. An explanatory framework for plan deviation is proposed in the context of a blackboard model. Plan deviation is supposed to occur if the control, according to certain selection criteria, selects an alternative-to-the-planned-action proposal rather than the planned action proposal. Implications of these results for assistance tools are discussed briefly.
Knowledge Exploited by Experts during Software System Design BIBA 279-304
  Raymonde Guindon
High-level software design is characterized by incompletely specified requirements, no predetermined solution path, and by the integration of multiple domains of knowledge at various levels of abstraction. The application of data-driven knowledge rules characterizes expertise. A verbal protocol study describes these domains of knowledge and how experts exploit their rich knowledge during design. It documents how designers heavily rely on problem domain scenario simulations throughout solution development. These simulations trigger the inferences of new requirements and complete the requirement specification. Designers recognize partial solutions at various levels of abstraction in the design decomposition through the application of data-driven rules. Designers also rely heavily on simulations of their design solutions, but these are shallow, that is, limited to one level of design methods, notations, and specialized software design schemas. Finally, the study describes how designers exploit powerful heuristics and personalized evaluation criteria to constrain the design process and select a satisfactory solution. Studies, such as this one, help map the road to understanding expertise in complex tasks.
Variability in Program Design: The Interaction of Process with Knowledge BIBA 305-322
  Robert S. Rist
A model of program design is proposed to explain program variability, and is experimentally supported. Variability is shown to be the result of different decisions made by programmers during three stages in the design process. In the first stage, a solution is created based on a particular design approach. In the second stage, actions in the solution are organized by features they share. The actions may then be merged together to define a more concise solution in program code, the third stage of design. Different programs will be created depending on the approach taken to design the features selected to group actions in a solution, and the features used to merge actions to form program code. Each of the variants observed in the study was traced to the use of a specific piece of information by a programmer at one of the three stages of program design. Many different programs were created as the process of design interacted with the knowledge of the programmer.
An Empirically-Derived Control Structure for the Process of Program Understanding BIBA 323-342
  Francoise Detienne; Elliot Soloway
Various models of program understanding have been developed from the Schema Theory. To date, the authors have sought to identify the knowledge that programmers have and use in understanding programs, i.e. Programming Plans and Rules of Discourse. However, knowledge is only one aspect of program understanding. The other aspect is the cognitive mechanisms that use knowledge. The contribution of this study is the identification of different mechanisms involved in program understanding by experts, specifically the mechanisms which cope with novelty. An experiment was conducted to identify and describe the expert's strategies involved in understanding usual (plan-like) and unusual (unplan-like) programs. While performing a fill-in-a-blank task, subjects were asked to talk aloud. The analysis of verbal protocols allowed the identification of four different strategies of understanding. Under "normal" conditions the strategy of symbolic simulation is involved. But when failures occur additional strategies are required. The authors identified three types of understanding failures the subject may experience (no expectation, expectation clashes, insufficient expectations) and the additional strategies invoked in those cases: (1) reasoning according to rules of discourse and principles of the task domain; (2) reasoning with plan constraints; (3) concrete simulation. The authors develop an operational description of these strategies and discuss the control structure of program understanding in the framework of schema theory.
Common Cognitive Representations of Program Code Across Tasks and Languages BIBA 343-360
  Scott P. Robertson; Chiung-Chen Yu
Plans are underlying cognitive structures used by programmers to represent code. In two studies we examined the content of plan-based representations and sought to show that common representations are used for programs that instantiate the same plans, even when they perform different tasks and are written in different languages (Pascal or FORTRAN). Our results support plan-based models and show that the organizing structures for chunks of code are abstract programming goals. The same abstract structures are formed for programs that perform different tasks using the same plans and for programs written in different languages but using the same plans. While plans were the primary organizing structures for code representations, other task-related information also played a role suggesting that programmers really utilize multiple representations. We advocate viewing code comprehension more like a plan recognition process and less like a text comprehension process.

IJMMS 1990 Volume 33 Issue 4

A Methodology for Validating Large Knowledge Bases BIBA 361-371
  Rajiv Enand; Gary S. Kahn; Robert A. Mills
Knowledge acquisition is not complete until a knowledge base is fully verified and validated. During verification and validation, knowledge bases are substantially refined and corrected. This paper offers a methodology for verification and validation that focuses knowledge acquisition on a progressively deeper set of issues related to knowledge base correctness. These are knowledge base verification, domain validation, procedural validation, and procedural optimization. This methodology has been developed in the course of using the TESTBENCH diagnostic shell to build a large system.
Developing a Tool for Knowledge Integration: Initial Results BIBA 373-383
  Kenneth S. Murray; Bruce W. Porter
Knowledge integration is the task of incorporating new information into existing knowledge. The task is difficult because the consequences of an addition to an extensive knowledge base can be numerous and subtle. Current methods for automated knowledge acquisition largely ignore this task, although it is increasingly important with the move toward large scale, multifunctional knowledge bases. To study knowledge integration, we propose to develop and evaluate a knowledge-acquisition tool that helps with extending a knowledge base through interaction with a knowledge engineer. An initial prototype of this tool has been implemented and demonstrated on a complex extension to a large knowledge base.
Semi-Automatic Acquisition of Conceptual Structure from Technical Texts BIBA 385-397
  Stan Szpakowicz
We present a system which processes technical text semi-automatically and incrementally builds a conceptual model of the domain. Starting from an initial general model, knowledge-based text understanding is turned into knowledge acquisition. Incompletely understood text fragments may contain new information which should be integrated into the model under the control of an operator. The text is assumed to describe the domain fully. Typical problems in this domain are assumed to be solvable by indicating activities which manipulate objects. Activities, objects and their properties enter relationships that form a conceptual network. To test our representation, we have created a large hierarchy of concepts for PowerHouse Quiz. The system relies in its operation on the text and the growing network; it includes a parser with broad syntactic coverage, and a matcher retrieving subnetworks relevant to the current text fragment. The frequency of the operator's necessary interventions depends on the initial network's size which will be determined experimentally. We discuss the status of the system and outline further work.
Semantic Strings: A New Technique for Detecting and Correcting User Errors BIBA 399-407
  James H. Bradford
Modern software is used to control a spectrum of industrial applications ranging from banking networks to nuclear reactors. The design of an effective error-handling strategy for user interfaces is becoming a vital issue. Such a strategy must include techniques for handling lexical, grammatical and semantic errors. The handling of semantic errors is the most difficult of these problems. This paper introduces a new technique for detecting and correcting semantic errors in user dialogue.
Individualized Tutoring Using an Intelligent Fuzzy Temporal Relational Database BIBA 409-429
  L. W. Hawkes; S. J. Derry; E. A. Rundensteiner
The student record (SR) is a major source of input for any decision making done by an Intelligent Tutoring System (ITS) and is a basis of the individualization in such systems. However, most ITSs still have "generalized" student models which represent a type of student rather than a particular one. Until the SR becomes truly representative of each individual student, the goal of providing individualized tutoring cannot be attained. In this paper we describe an Intelligent Fuzzy Temporal Relational Database (IFTReD), and intelligent system-independent SR which allows for almost any degree of individualization the designer wishes to incorporate. It is anticipated that this IFTReD will provide a significant improvement over standard AI storage techniques for the SR. These improvements will be realized in terms of: (1) intelligence; (2) greater storage efficiency; (3) greater speed in retrieval and query; (4) ability to handle linguistic codes, ranges, fuzzy possibilities, and incomplete data in student models; (5) friendliness of query language; (6) availability of temporal knowledge to give a history of past performance; and (7) a more holistic view of the student, permitting greater individualization of the tutor.
EMCUD: A Knowledge Acquisition Method which Captures Embedded Meanings Under Uncertainty BIBA 431-451
  G. J. Hwang; S. S. Tseng
In this paper, we propose a knowledge acquisition method EMCUD which can elicit embedded meanings of the initial knowledge provided by domain experts. EMCUD also helps experts to decide uncertainty of the embedded meanings according to the relationships of the embedded meanings and the initial knowledge. The strategy of EMCUD could easily be added to the repertory grid-oriented methods or systems to enhance the knowledge in the prototype. We present a realistic example to show how EMCUD enriches the knowledge base constructed by a repertory grid-oriented method and hence ease the refinement processes.
Direct Manipulation as a Source of Cognitive Feedback: A Human-Computer Experiment with a Judgment Task BIBA 453-466
  Dov Te'eni
Correctly designed feedback can play a pivotal role in improving performance. Human-computer interaction can generate various forms of feedback, and it is important to examine the effectiveness of the different options. This work compares: (1) feedback that is presented as information which is distinct from the user's action, with (2) feedback that is generated by direct manipulation and is embedded in the same information which facilitates the user's action. The former is the traditional form of feedback in which the user acts and receives feedback information from a distinct source. The latter is information generated during the user's action, and it becomes effective feedback only when the individual uses this information as feedback.
   This paper explains why certain elements of direct manipulation can invoke the second form of feedback. An experiment demonstrates that feedback resulting from direct manipulation is more effective and time efficient than the distinct form of feedback in conditions of high task complexity. However, direct manipulation has its limits and must be complemented with traditional forms of feedback for complex cognitive tasks.
Two Perspectives of the Dempster-Shafer Theory of Belief Functions BIBA 467-487
  Pawan Lingras; S. K. M. Wong
This study presents two different perspectives of the Dempster-Shafer theory of belief functions. The first view of the theory, called the compatibility view, constructs the belief function for a frame of discernment by using the compatibility relationship of the frame with another frame called evidence frame for which the probability function is known. The second view referred to as the probability allocation view provides a generalization of the Bayesian theory by allocating the probability mass to the propositions which may not necessarily be singleton sets of possible answers, based on some vague body of evidence. Both these views are useful in the design and implementation of expert systems. The belief functions constructed using the evidence frames are useful in situations where limited information regarding the relationship between two frames of discernment is available. The belief functions based on the allocation of probability mass are useful when the evidence cannot be explicitly expressed in terms of propositions, but probability allocation based on evidence is possible. A successful implementation of these two views requires a rule for combining evidence. However, the Dempster rule of combination cannot incorporate dependencies among different bodies of evidence. This study illustrates a possible method for incorporating dependencies based on limited information.

IJMMS 1990 Volume 33 Issue 5

Learning Plans for an Intelligent Assistant by Observing User Behavior BIBA 489-503
  Keith R. Levi; Valerie L. Shalin; David L. Perschbacher
A critical requirement of intelligent automated assistants is a representation of actions and goals that is common to both the user and the automated assistant. Updating the intelligent system's knowledge base by observing user behavior is a convenient method for acquiring this common representation. We are developing an explanation based learning system to automate the acquisition of new plans for a large pilot-aiding expert system. We have developed a planning/learning shell that is based on the TWEAK planning system and DeJong's explanation based learning system. We are applying this shell to the pilot-aiding problem in a joint industry/university research effort involving Honeywell, Lockheed, ISX, Search Technology, and the Universities of Illinois and Michigan.
The Meaning Triangle as a Tool for the Acquisition of Abstract, Conceptual Knowledge BIBA 505-520
  Stephen Regoczei; Graeme Hirst
The meaning triangle is presented as a useful diagramming tool for organizing knowledge in the informant-analyst interaction-based, natural language-mediated knowledge acquisition process. In concepts-oriented knowledge acquisition, he knowledge explication phase dominates. During the conceptual analysis process, it is helpful to separate verbal, conceptual, and referent entities. Diagramming these entities on an agent-centered meaning triangle clarifies for both informant and analyst the ontological structure that underlies the discourse and the creation of domains of discourses.
Using Personal Construct Techniques for Collaborative Evaluation BIBA 521-536
  Douglas Schuler; Peter Russo; John Boose; Jeffrey Bradshaw
Efforts are under way to characterize better group processes at the Boeing Company as part of a project to design software for computer-supported collaboration. This paper describes work in progress to support multi-user, collaborative situations using Aquinas, a knowledge acquisition workbench. An experiment is described in which Aquinas is used to facilitate the collaborative evaluation of an in-house Boeing Advanced Technology Center course in knowledge engineering.
Functional Design of a Menu-Tree Interface within Structured System Development BIBA 537-556
  Peretz Shoval
A systematic method for designing a menu-tree interface is presented. The method is part of ADISSA, a comprehensive systems analysis and design methodology, which is based on the use of modified hierarchical data flow diagrams (DFD). Thus, the functional structure of the designed menu-tree is consistent with the functional structure and the user-model of the system. The method consists of several steps, beginning with an initial menu-tree derived automatically from DFDs, which is then improved and modified, taking into consideration factors other than functionality. Menu Tree Designer, one of the software tools of ADISSA methodology, supports the designer in all stages of the interface design.
An Intelligent Language Tutoring System BIBA 557-579
  Camilla B. Schwind
In this paper, we present the theoretical background and describe the design and implementation of an intelligent language tutoring system (ILTS). The most important properties of our system are: (1) The system is based on a very complete and "objective" grammar knowledge base; (2) Students can at any moment during an exercise ask the system questions about the grammar, and they are immediately answered without losing the exercise context. Thus the normal behaviour of a tutor is better simulated, which contributes to a user-friendly interface; and (3) It allows for individual correction of errors and reaction to errors. This is due to the fact that the system is firmly based on a linguistically well-founded analysis. The sentences formulated by the students are parsed and analysed. They are not simply matched against predefined answers as is still the case with many other more classically oriented systems.
An Evaluation of the Effectiveness and Efficiency of an Automobile Moving-Map Navigational Display BIBA 581-594
  Jonathan F. Antin; Thomas A. Dingus; Melissa C. Hulse; Walter W. Wierwille
This experiment was performed to evaluate the effectiveness and efficiency of navigating with an automobile moving-map display relative to navigating with a conventional paper map and along a memorized route, which served as a baseline for comparison. Results indicated that there were no differences in the quality of routes selected when using either the paper map or the moving map to navigate. However, the moving map significantly drew the driver's gaze away from the driving task relative to the norm established in the memorized route condition, as well as in comparison to the paper map. These findings are discussed in the context of the different navigation strategies evoked by use of the paper and moving-map methods of navigation.
Different Notions of Uncertainty in Quasi-Probabilistic Models BIBA 595-606
  L. C. van der Gaag
In the early years of the research into plausible reasoning several quasi-probabilistic models for handling uncertainty in rule-based expert systems have been proposed. These models were computationally feasible but could not be justified mathematically. Although current research in this sub-area of artificial intelligence concentrates on the development of mathematically sound models, the early quasi-probabilistic models are still employed frequently in present-day rule-based expert systems. In this paper we show that two of these models, the certainty factor model developed by E. H. Shortliffe and B. G. Buchanan, and the subjective Bayesian method developed by R. O. Duda, P. E. Hart and N. J. Nilsson, model different notions of uncertainty. We support this statement by pointing out the difference in the interpretation and application of production rules in the respective models.

IJMMS 1990 Volume 33 Issue 6

PRED: A Frame-Based Primitive Editor BIBA 607-621
  Shun-En Xie; David F. Dumaresq; Philip H. Winne
PRED allows domain experts to easily create frame-based windowed knowledge acquisition interfaces for knowledge systems development. These interfaces provide visibility for complex domain knowledge and speed up the entry of knowledge by allowing the domain expert to supply lists of default values. Domain experts first use PRED to schematize a knowledge unit, building a knowledge description. The knowledge description becomes a map to create a panel or the data structure to contain each instance of entered knowledge. A knowledge description can be customized by locking certain slots giving special views of the unit. This is useful for reducing choices in a generalized model and improves the time for entering data.
Towards a Classification of Text Types: A Repertory Grid Approach BIBA 623-636
  Andrew Dillon; Cliff McKnight
The advent of hypertext brings with it associated problems of how best to present non-linear texts. As yet, knowledge of readers' models of texts and their uses is limited. Repertory grid analysis offers an insightful method of examining these issues and gaining an understanding of the type of texts that exist in the readers' worlds. The present study investigates six researchers' perceptions of texts in terms of their use, content and structure. Results indicate that individuals construe texts in terms of three broad attributes: why read them, what type of information they contain, and how they are read. When applied to a variety of texts these attributes facilitate a classificatory system incorporating both individual and task differences and provide guidance on how their electronic versions could be designed.
Display-Based Competence: Towards User Models for Menu-Driven Interfaces BIBA 637-655
  Andrew Howes; Stephen J. Payne
This paper discusses the critical role played by aspects of the display in the use of many computer systems, especially those driven by menus. We outline a formal model of "display-based competence" by extending the Task-Action Grammar notation (Payne & Green, 1986). The model, D-TAG (for display-oriented task-action grammar) is illustrated with examples for the well-known Macintosh desk-top interface, and from a more deeply-nested menu interface to a device used for the remote testing of telephone lines (RATES).
   D-TAG exploits two extensions of TAG to address important aspects of interface consistency. The most important extension uses a featural description of the display to capture the role of the display in structuring task-action mappings; the second describes the "side-effects" of a task, i.e. those effects not described by the semantic attributes of a task. By embedding these extensions within the organizing framework of TAG's feature-grammar, we are able to develop descriptions of interfaces which highlight aspects of (display) design that are outside the scope of other formal user models.
Integration of Linguistic Probabilities BIBA 657-676
  David V. Budescu; Rami Zwick; Thomas S. Wallsten; Ido Erev
In a previous study, Zwick, Budescu and Wallsten (1988) found that the membership functions representing the subjective combinations of two independent linguistic probabilistic judgements could not be predicted by applying any dual t- and co-t-norm to the functions of the underlying terms. Their results showed further that judgements involving the "and" connective were best modelled as the fuzzy mean of the two separate components. The present experiment extended those results by manipulating the instructions regarding the "and" connective and also including an additional task in which subjects selected a third phrase to represent the integration of the two independent judgements. Again, no t-norm rule predicted subjects' responses, which were now best modelled by the point-wise arithmetic or geometric means of the functions. In addition, most subjects selected phrases and provided membership functions in response to two identical forecasts that were more extreme and more precise than the individual forecast, a result inconsistent with any t-norm or averaging model. A minority of subjects responded with the same phrase contained in the forecasts. The entire pattern of results in the Zwick et al.(1988) and the present study is used to argue against the indiscriminate application of mathematically prescribed, but empirically unsupported operations in computerized expert systems intended to represent and combine linguistic information.
Sorting-Based Menu Categories BIBA 677-705
  Douglas Hayhoe
Several researchers have conducted sorting experiments or pairwise comparisons with a database of menu items in order to form coherent menu categories. However, these experiments have all contained one or more of the following potential weaknesses: (1) they used only one particular database; (2) they used too few sorting subjects; (3) they uncritically used only one scaling technique to form the clusters; or (4) they did not conduct an experimental comparison of the categories formed. In the present research, sorting experiments were conducted with 48 subjects and four 48-item databases: clothes, furniture, occupations, and sports. Latent partition analysis and hierarchical clustering (Ward's method and group average linkage) were used to form menu categories. These were placed into a "pull-down" menu system in two conditions: (1) titles chosen by each individual subject; and (2) titles chosen by the investigator. Two other conditions were added: (3) categories and titles formed by software design experts; and (4) categories and titles formed by each subject for his or her own work. Two within-subjects menu experiments were performed. The sorting-based categories with investigator titles were superior to the expert categories in selection times, selection errors, "goodness of fit" ratings, and memory recall errors. A detailed analysis showed that the expert categories contained more "miscategorization" errors and vague category titles than the sorting-based categories, while both conditions contained overlapping categories.
Novices' Debugging when Programming in Pascal BIBA 707-724
  Carl Martin Allwood; Carl-Gustav Bjorhag
In this study an analysis was made of novices debugging their own Pascal programs. Eight novices verbalized their thoughts aloud while attempting to solve a programming task. Novices' debugging is seen as taking place in negative evaluation episodes (henceforth: evaluation episodes). During the three hour programming session, the novices spent 51% of the time in evaluation episodes. This percentage would presumably have been higher if the subjects had been given more time for the session. Evaluation episodes were found to be triggered in four different contexts: Reaction to an error message (67% of the total time spent in any evaluation episode), Reaction to the resulting value of a test run (23%), Hint from the experimenter (4%) and Other (6%). When related to results presented by Gray and Anderson (1987), our results indicate that novices perform the substantial part of their debugging after they have compiled the program, or part of it, for the first time. Despite the information given in the computer's error messages, the percentage of errors eliminated in episodes triggered by such message was not higher than could be expected from the time spent in these episodes. Our results indicate that the importance of activity oriented towards understanding one's program during debugging depends on: (1) whether the error elicits an error message from the computer or not; and (2) the general programming strategy used by the subject.