HCI Bibliography Home | HCI Conferences | BCSHCI Archive | Detailed Records | RefWorks | EndNote | Hide Abstracts
BCSHCI Tables of Contents: 8586878889919293949596

Proceedings of the HCI'85 Conference on People and Computers: Designing the Interface

Fullname:Proceedings of the Conference of the British Computer Society Human Computer Interaction Specialist Group
Note:People and Computers: Designing the Interface
Editors:Peter Johnson; Stephen Cook
Location:University of East Anglia
Dates:1985-Sep-17 to 1985-Sep-20
Publisher:Cambridge University Press
Standard No:ISBN 0-521-32066-6; QA 76.9 S88B755 1985; hcibib: BCSHCI85
  1. The Design Process: Models and Notation for Interaction
  2. The Design Process: Applications of Task Models
  3. The Design Process: Interaction Techniques
  4. The Design Process: Dialogue Management and Design
  5. The Design Process: Evaluation
  6. Applications: Simplified User Interfaces
  7. Applications: Expert Systems
  8. Applications: Computer Based Training
  9. Applications: Office Automation
  10. Applications: Electronic Mail
  11. Design Issues: Speech Interaction
  12. Design Issues: Models of Programming
  13. Design Issues: Graphical Interfaces

The Design Process: Models and Notation for Interaction

Modelling User Behaviour with Formal Grammar BIBA 3-12
  A. J. Fountain; M. A. Norman
Formal descriptive tools have been used to specify user behaviour at the human-computer interface. Two prominent examples of this approach are the GOMS Theory (Card et al, 1983) and Reisner's Formal Grammar (Reisner 1981, 1984). The GOMS Theory and Formal Grammar are shown to be equivalent in their power of describing use of an interface. Both GOMS and Formal Grammar describe human behaviour in Backus Normal Form. GOMS has a stack discipline for organising the relationships between a user's Goals and Methods; Formal Grammar of Reisner (1981, 1984) has the more general production rule control structure for organising the hierarchy of grammatical symbols. Differences are shown in their use as analytic tools. The GOMS Theory directs attention towards the selection of methods for predefined tasks and the points at which this selection occurs. The Formal Grammar does not consider method selection but determines the complexity of both physical and cognitive operations in pre-defined tasks. It is suggested that the approaches need to be combined in order to provide a more comprehensive formal method for modelling user behaviour.
Abstract Models of Interactive Systems BIBA 13-22
  Alan Dix; Colin Runciman
We propose an abstract model for a large class of interactive systems. In these systems the user provides a sequence of commands that determines both a corresponding sequence of displays and a net effect or result. Editors, for example, usually fit this model. We show how our model can be used to address issues such as display laws, error correction, exception handling and command types. We give some formal statements of design principles, and also discover ways in which these interact or even conflict. Such results are of value whether or not a formal development method is used.
Definitive Notations for Interaction BIBA 23-34
  Meurig Beynon
This paper explores some methodological and pragmatic aspects of the design of the human-computer interface. In particular, it argues that many interactive dialogues can be formulated conveniently and clearly using notations based upon sequences of definitions ("definitive notations"). Such a notation is an implicit ingredient in the "spread-sheet" packages which have recently become so popular in business applications. To apply similar principles to more complex tasks, such as CAD applications, requires abstraction and generalisation, and poses challenging technical problems.
   The three sections of the paper respectively consider: background and motivation, elementary definitive notations (illustrated by an unconventional desk calculator), and complex definitive notations (with particular reference to the design of ARCA, a definitive notation for the interactive description and manipulation of combinatorial diagrams).
User Modelling Techniques for Interactive Systems BIBAK 35-45
  I. Clowes; I. Cole; F. Arshad; C. Hopkins; A. Hockley
In order for an interactive system to adapt to the needs and preferences of individual users, the system requires techniques whereby it can build up and maintain a representation, or model, of the user. Research into developing effective user modelling techniques forms part of research being carried out by the Alvey Adaptive Intelligent Dialogues (AID) project. The primary objective of this project is to research and develop techniques for building user interfaces that adapt to the needs of individual users.
   This paper reports the findings of a survey of computer-based user modelling techniques, mainly from the areas of Intelligent Tutoring Systems and Intelligent Front Ends, that have already been developed. The paper goes on to suggest research directions for user modelling.
Keywords: User modelling, Adaption, Intelligent front ends

The Design Process: Applications of Task Models

Towards a Task Model of Messaging: An Example of the Application of TAKD to User Interface Design BIBA 46-62
  P. Johnson
This paper describes how Task Analysis for Knowledge Descriptions (TAKD) was used in the identification and description of a series of messaging tasks. The purpose of this analysis was to identify user requirements and to form a basis from which a new interface for an extended electronic mail system could be designed. The techniques used to identify and describe the messaging tasks are described and exemplified. A form of notation for task descriptions (KRG), originally developed to describe training requirements, was applied. The application of this notation is described. The analysis and notation were in part influenced by the requirements of the system designer. A top-down, bottom-up design method can be supported by the use of task analysis. The appropriateness of this form of task description for this style of design is discussed. Particular emphasis is placed upon the requirements of system implementation using an object-oriented programming approach (Smalltalk80). It is argued that this method of task analysis can identify the objects and operations (actions) which are to be used in object-oriented programming.
Analysing the Learning of Command Sequences in a Menu System BIBA 63-75
  M. D. Wilson; P. J. Barnard; A. MacLean
Although there is a substantial literature on both novice and expert performance, there is little data on the transition from one to the other. This paper presents data from 8 subjects performing a core set of tasks in each of word processing, graph drawing and calculation environments during this transition. A descriptive model of the command structure used in these tasks is presented which permits the analysis of both the successful attempts to complete the tasks and those involving deviations from optimal performance. The pattern of deviations changes over learning in that the proportion of those involving major re-attempts at tasks decreases while that involving local corrections increases. Two classes of mental representation are used to explain the changing performance: those involving general system principles, and those using specific procedures. The changes that take place during learning are characterised as an increase in the proportion of specific procedures in the repertoire of representation sampled during task performance.
Effects of System and Knowledge Variables on a Task Component of "Teleshopping" BIBA 76-91
  P. K. Buckley; J. B. Long
This paper reports the background to and the results of an experimental evaluation of the usability of videotex for teleshopping. The experimentation represents part of phase two of a project intended to provide guidelines for designers of videotex transaction dialogues and an understanding of the sources of user difficulty. The experiment was intended to assess the usability of videotex options for the transaction task element of 'evaluation' in which shoppers were assumed to assess the adequacy of potential purchases for their needs. The system variable 'Extent of Description of Goods', and related knowledge variable 'Knowledge of Transaction Domain', were manipulated experimentally and found to affect usability in terms of errors made and time taken. The study stresses the need to provide appropriate descriptions of goods relative to what users need to know. Within the range of display options available to videotex transaction designers, the appropriateness and consequent usability will depend strongly on the users' knowledge. The experimental results are expressed as implications for the design of teleshopping systems.

The Design Process: Interaction Techniques

Helping Both the Novice and Advanced User in Menu-Driven Information Retrieval Systems BIBA 92-101
  D. L. Heppe; W. H. Edmondson; R. Spence
Of the variety of ways in which information can be retrieved from a database, the conventional menu-based approach is said to be especially suited to the novice. It does, however, suffer from many drawbacks such as ambiguous category labels and artificial hierarchies. Additionally, users can become lost in the retrieval structure and be compelled to ask, 'Where am I?'. To overcome these difficulties various navigational aids can be deployed. This paper defines and critiques these techniques, with references to examples in the literature and illustrations taken from a newly implemented working system. The more general cognitive aspects of these procedures are also discussed with suggestions for a more cognitively sensitive approach to interface design.
Requirements for an Intelligent Form-Filling Interface BIBA 102-116
  D. M. Frohlich; L. P. Crossfield; G. N. Gilbert
Although forms have been used as an alternative to menu selection and command language interfaces, their full potential for aiding human-computer interaction has rarely been realised. Most form interfaces at present provide little support to the user other than straightforward data type checking and some static help facilities that are displayed upon request. In this paper, the scope for improving the support provided to users of form-filling interfaces is identified by analogy to the support required by form-fillers in general. Some findings from the work on paper form-filling and design are reviewed, in relation to the comprehension of questions, the representation of answers and form navigation. A preliminary design for a Forms Helper system is then described. This illustrates the kind of knowledge which must be represented in any system which is to provide intelligent assistance to users of form-filling interfaces.
Use of Conceptual Maps as Human-Computer Interfaces BIBA 117-127
  A. G. Sutcliffe
A human computer interface is described which represents a hierarchically organised system as map in the form of a tree diagram. It is argued that this interface could be more efficient than standard techniques because it is more space efficient in representing entities within a system than menus and imposes less memory load on users than command languages. Operation of the interface is compared with menus and a command language using a method of predicting interface operation based on the work of Card et al (1980). For novice users the map interface is predicted to be much slower than other techniques but for expert users the reverse is true. It is proposed that map interfaces which exploit human abilities of pattern recognition should combine characteristics of being easy to learn with operational efficiency, and provide a more adaptable interface than other techniques.

The Design Process: Dialogue Management and Design

A Path Algebra Support Facility for Interactive Dialogue Designers BIBA 128-137
  J. L. Alty; R. A. Ritchie
Path algebras have been shown to be useful in interactive dialogue design. They can be used to carry out consistency checks, path analyses and other analytical functions on interactive dialogues described as recursive transition networks. Such analyses are important for large networks. A path algebra support facility has been built which runs on the VAX 11/750 and which allows different algebras to be defined by the designer in an easy to use manner. The algebraic operators are described to the system using LISP and they act upon a uniform arc-labelling scheme based upon lists. The resultant adjacency matrix for a particular dialogue network together with its powers are displayed and a number of iterative techniques are used to obtain the weak and strong closure matrices. The designer can not only analyse these networks using existing algebras but can also define new algebras and investigate their properties. The system and its application to dialogue design are explained. Possible future uses of this algebraic approach are outlined.
Three Transition Network Dialogue Management Systems BIBA 138-147
  Gilbert Cockton
A Dialogue Management System has to be programmed and must therefore be able to construct and sequence actions. The control model adopted must be based on a specified formalism. A minimal Transition Network formalism for dialogue is presented and it is shown that, when examined with reference to this formalism, the three Transition Network Dialogue Management Systems are weakened by unjustified restrictions.
GUIDE: A UNIX-Based Dialogue Design System BIBA 148-160
  P. Gray; A. Kilgour
This paper describes GUIDE, a graphical user interface design environment currently being implemented at the Computing Science Department of Glasgow University. Dialogues are described by dialogue scripts which are sections of the file system maintained by the UNIX operating system on which GUIDE is built. The directories correspond to dialogue units, consisting of individual files (and subdirectories) which define the components (such as prompt, echo, response etc.) of an individual unit of interaction. Most of the files are text files so they can be simply modified by an editor. A graphics editor is planned which will permit direct graphical control over the dialogue component.
   The other elements of GUIDE are a dialogue interpreter, which executes a dialogue by traversing the script structure in response to input events, and a dialogue design script, which when executed by the interpreter permits other dialogue scripts to be constructed or modified. The dialogue designer, thus, uses the same interactive tools as are made available to the dialogue user, as in the MENULAY system of Buxton et al., and the whole dialogue design system is available as a subdialogue during execution of any script by the interpreter. GUIDE is intended to provide fast prototyping, because no compilation stage is involved before trying out a newly modified script.
   The paper describes the structure of dialogue scripts and the implementation of the interpreter, and an example is included showing the application of GUIDE to the simulation of simple data structure algorithms.
Formalising Guidelines for the Design of Interactive Systems BIBA 161-171
  M. D. Harrison; H. W. Thimbleby
User engineering principles may be formalised as theorems over specifications of interactive systems. In this paper we discuss some different categories of user engineering principle and expose issues that must be resolved to produce effective formalisation.

The Design Process: Evaluation

Evaluating the Human Interface of a Data Entry System: User Choice and Performance Measures Yield Different Tradeoff Functions BIBA 172-185
  A. MacLean; P. J. Barnard; M. D. Wilson
When people use computer systems, they are often faced with alternative methods for carrying out a given task. They have to be able to judge which method is likely to be most appropriate for the particular task with which they are faced. A study is presented which compares the most common means of evaluation used in computer design, the time to carry out a given task, with the method the user actually chooses to carry out the task. The results suggest that users are not good at optimising their behaviour on the criterion used by designers.
The Relationship Between Cognitive Style and Dialogue Style: An Explorative Study BIBA 186-198
  C. J. H. Fowler; L. A. Macaulay; J. F. Fowler
This paper presents and discusses findings from an explorative study which examined the relationship between cognitive style and dialogue style as reflected in various measures of performance.
   The cognitive style dimension used in this investigation was field-dependence/independence (Witkin and Goodenough, 1981), as measured by the GEFT, the main dialogue style variable was the kind of command structure (linear and substructure), and the main performance variables were time taken to complete a task ('thinking' and 'doing'), the number of 'Help' requests, the number of corrected/uncorrected errors, and the kind of error made (usage or typographical). These variables were examined over two different learning blocks using 48 subjects.
   The results were analysed by the use of correlational techniques. The results suggested that field-dependent persons tend to prefer a substructured command language and that field-independent individuals show a general preference for linear command structure.
   These results were interpreted in terms of differences in mode of information processing arising from a greater or lesser reliance on external referents. The authors conclude that cognitive style is an important concept in relation to human-computer interaction, and may be of considerable use in the area of interface design.
Use of Man-Modelling CAD Systems by the Ergonomist BIBA 199-208
  P. L. Rothwell
This paper discusses man-modelling CAD systems, focusing on their applications to ergonomics and the quality of their interface with ergonomists as users. Three specific systems are examined to demonstrate the applications and limitations of computer man-models that are presently available, and requirements for an ideal system are identified. Recommendations are made for future development of man-modelling CAD systems, and their presentation to the user.

Applications: Simplified User Interfaces

SUSI -- A Smart User-System Interface BIBA 211-220
  J. Jerrams Smith
The necessary qualities of a human-computer interface include the discovery and prompt correction of the user's misconceptions about the system, thus involving an instructional component. The interface should provide an intelligent response suited to the user and it should be modular for ease of modification.
   The SUSI interface has been designed with such requirements in mind. It is intended to be generally applicable to systems which are currently available (for example, the BLEND electronic journal and Operating Systems such as UNIX and MULTICS) so that they can be used easily and efficiently by novices as well as experts.
   The first implementation is an interface between the user and the UNIX shell. The SUSI interface possesses knowledge about UNIX and knowledge about the behaviour and types of error made by UNIX users. This knowledge is held in the Knowledge Base of an Expert System. The interface maintains a user model and interprets user input in order to discover the user's misconceptions. The interface differentiates between misconceptions and spelling/typing errors. This is particularly important so that users are not given unwanted help: most expert-user errors are in typing and spelling. The interface provides suitable instruction based upon the user model and the interpretation of user input.
Interfaces for Database Systems BIBA 221-229
  M. A. Newton; S. M. Sussmann; J. Watkins
The structured databases controlled by database management systems need to provide access for many users for various purposes. Query languages are intended for users to specify their own database retrieval operations though this can be difficult when it is necessary to manipulate database structures. Perceived records have been proposed as a way of providing each user with a single data structure to contain the data relevant to their queries. The paper describes a prototype implementation of a system using perceived records and the interfaces necessary to support their definition and use. The purpose of using perceived records is contrasted with a relational interface. The functionality required for user queries is explained and the dialogue to control the specification of perceived records is justified.
The Interactive Process Scheduler BIBA 230-238
  D. J. Browning; G. M. Cain; F. G. Gouldstone; J. McEntegart
The use of conventional software development procedures for constructing systems suitable for interactive scientific applications is not always fully compatible with optimising the user-computer interface in terms of the user's perception of the system as a tool for solving problems by logical progression through a sequence of processes.
   Consideration of methods for optimising this user-computer interface has resulted in the development of a requirement specification language, which via a hierarchical menu system definition, can be transformed into an interactive software framework to which the application specific software processes are added. This transformation is accomplished using a system known as the Interactive Process Scheduler (IPS); software systems developed using this technique exhibit both a conventional menu structure for individual process selection and sequence structures reflecting the original requirement specification.

Applications: Expert Systems

What Kind of System Does an Expert Need? BIBA 239-247
  A. L. Rector; P. D. Newton; P. H. Marsden
Many potential users of 'expert systems' are themselves experts. Yet most models for expert systems treat users as novices, placing them in a passive role and ignoring their expertise. Experts are fundamentally different from novices; not only do they know more, they know differently. They perceive their field in terms of rich interrelationships rather than isolated facts and can make use of far more information than can novices. The transition from novice to expert involves fundamental cognitive shifts, analogous to those described by Piaget in children's cognitive development. Decision support systems for experts such as doctors must take these characteristics into account and provide a much richer environment than current expert systems. What 'experts' and what current 'expert systems' do are fundamentally different. Systems for experts should be designed to allow cooperation between the system and the user and should provide users with real examples from which to draw their own conclusions.
The Consultative Role of an Expert System BIBA 248-254
  A. L. Kidd
The development of expert systems has concentrated on simulating the problem solving role of human experts and has largely neglected their role as consultants. This shortcoming may account for the failure of many expert systems to provide effective decision support in field applications. In an attempt to redress this balance, this paper outlines two important shortcomings in the consultative function of current expert systems: rigid dialogues and inadequate explanations. The main part of the paper then describes features of naturally occurring human consultations and suggests that to be successful, expert systems need to support a range of user decisions, allow the user to take an active role in the problem solving process and provide explanations which increase the user's understanding of the domain.
The Use of Rule Induction, A Knowledge Acquisition Technique for Expert Systems, to Interpret HCI Experiments BIBA 255-263
  A. Brooks; J. L. Alty
Traditional statistical analyses can be carried out only on data gathered from carefully designed experiments. However, even when HCI experiments have been carefully designed, it is all too often the case that some other unconsidered problem attribute is a major contributory factor to the results. An alternative approach suggested here is based on the technique of automated rule induction from examples. This technique has been successfully used to develop rule bases for expert systems. Using this technique all possible problem attributes that can be thought of, before and after an experiment, may be included in the analysis. Hidden patterns or rules are automatically induced and less relevant problem attributes discarded. Results are reported of using rule induction on experimental data collected with Alty's CONNECT system which show rule induction to be a worthwhile method. It is further suggested that rule induction should be used to form experimental hypotheses from naturalistic or unplanned experiments as a first step toward more meaningful traditional experimentation.

Applications: Computer Based Training

Intelligent Computer Based Training BIBA 264-272
  P. S. Roberts
This paper discusses the merging of interactive video with expert systems to produce intelligent CBT focussing on the man machine interface. Three aspects of the interface are dealt with in detail. Firstly the output interface, comprising the very powerful combination of video images plus computer generated text and graphics; secondly, the use of expert system techniques in developing student models for more effective control of the training session; lastly, the use of expert system techniques to provide a comprehensive explanation facility to allow students to question both the knowledge base (visual and textual) and the training process itself. This paper draws on both commercial experience and current research projects.
The Presentation of Learning Material via Microcomputers and a Method for Assessing its Effectiveness BIBA 273-281
  G. M. Mills
The categorization of screen material into four main groups is proposed, ie reference, exposition, user response and feedback. One main aim of the program developer should be to clearly differentiate each category to avoid confusing the learner. The assessment of the presentation of screen display material for educational microcomputer programs has been studied through ratings exercises in courses on computers in education for the past four years. The original rating scale of eight factors indicated wide variations in assessment between raters on individual factors but there was more consistency in ranking programs overall. Since some of the original factors covered more than one presentation category, a revised version with sixteen factors was prepared in the hope that this would improve consistency of individual factor ratings. In the trials, no significant improvement was found but the use of the revised version is shown to be justified in its provision of greater diagnostic power, particularly when a number of raters are used.
Mathematical Students and Computers: An Interface for Experimentation BIBA 282-289
  J. A. Glen
Students in the first level course of the Science Degree Scheme at Paisley College of Technology have used microcomputers as part of their laboratory work in the Mathematical Sciences course since September 1982. We attempted to create an environment where, with limited keyboard skills allied to an understanding of mathematical terms, techniques and concepts, the student could use the micro network as a versatile experimental aid for mathematics. The major effort in producing the software has gone into the display of mathematical information in a format familiar to the student, in providing graphical display and in creating reliable modes of interaction. This effort has been greatly facilitated by the production of software utilities for these aspects. This layer of interface, between the student user and the individual software has been tried and tested. The paper reports on progress. A recent development has been the tentative provision of a Simple Knowledge and Information Processing System (SKIPS). SKIPS allows students to browse through the available packages and to enquire about some of the mathematical background to the packages. The paper discusses the evident shortcomings of SKIPS in using standard interfaces, in particular the need for some measure of expertise in the interface will be discussed.

Applications: Office Automation

What Do Clerical Workers Think About Computers? BIBA 290-298
  L. A. Macaulay; C. J. H. Fowler; R. Porteus
The attitudes towards computers of a sample (N=48) of clerical workers were measured by means of a questionnaire. The individual items used were taken from a questionnaire originally designed and used by Zoltan and Chapanis (1982). The clerical workers' responses were then compared with those of the four professional groups sampled in Zoltan and Chapanis' study -- i.e. accountants, lawyers, pharmacists and physicians.
   The results suggest that with respect to computers the attitudes of clerical workers closely resemble those of pharmacists, and tend to be more positive than those of lawyers and physicians. The findings are explained in terms of the similarities and differences between the various occupations and the degree to which computers are able to assist in the sort of tasks which characterise each occupation.
A Committee Secretarial Support System BIBA 299-309
  B. M. Brown; P. R. Davies; W. A. Gray
Many organisations use committee meetings to control their decision making processes. These meetings have a well defined structure with an agenda identifying the business to be discussed at a meeting. After each meeting the committee's Secretary prepares the Minutes of the meeting -- a report which summarises the discussion about each item of business, records any decisions taken on the matter and identifies actions to be taken. Prior to a meeting the Secretary prepares an agenda, ensuring the necessary supporting papers and reports are available and that any items of business due for discussion as a result of decisions at previous meetings are brought forward
   This paper will analyse the agenda minute cycle determining the associated Secretarial functions, the underlying information flows with their associated control functions and the essential links between minutes and agenda items.
   Two possible models for a computerised Secretarial Support System will be discussed -- a linear and cyclic model. Reasons will be given as to why the cyclic model was chosen for the implemented prototype system. This prototype system was created using a script driven editor. It is designed to be a portable system which is supportive of the secretarial functions, while minimising any adverse effects it may have on the committee's functioning. The design, construction and functionality of this prototype system will be discussed in the paper.
Looking Back on the Office of the Future BIB 310-316
  Barry Sheffield

Applications: Electronic Mail

Mailbox Advances & MMI Needs BIBA 317-330
  Paul A. Wilson
Mailbox systems represent a new medium which allows people to communicate at times and places of their own choosing. This paper attempts to alert the Human Factors Community to the way in which mailbox systems will be developed and used in coming years in order to ensure that the Human Interface requirements of such systems can be anticipated and researched in a timely fashion. Some hypothetical examples of mailbox useage are provided and the concept of mailbox structures outlined and illustrated. The potential for structuring to embrace and unify all different communication systems and media is discussed. Finally, the urgent need for Human Factor input to the design of such facilities is emphasised and the way in which such work might be organised is described.
An Intelligent Mail Filter BIBA 331-341
  C. D. B. Boyle; M. R. B. Clarke
Electronic Mail systems handling a large volume of traffic need some form of filtering to avoid user overload. Bulletin boards are one widely employed solution. This paper investigates the use of user profiles as an additional method of finding out what users might find interesting. In a production mail system such a filter might fit below bulletin boards in the mail structuring hierarchy.
   A prototype intelligent mail filter has been implemented and some experiments carried out to evaluate the seven alternative algorithms presented. There are two subsystems, a profile creation program and a message evaluation system. Both active and passive systems are considered. Active systems allow querying but involve the user in more work. Passive methods build up a profile from observations of the users interest level.
   Context based methods, which attempt a semantic analysis of the message were found to be relatively better with good user profiles. Keyword methods were more reliable with poor profiles. The problem of planted keywords is considered.

Design Issues: Speech Interaction

Speech Technology: Is It Working? BIBA 345-358
  Mike Talbot
This paper examines the current state of automatic speech recognition devices and of speech synthesisers. Some examples of current applications of this technology are given, but it is noted how the technology has not become a widely used replacement for more conventional methods of human-computer communication. It is thought that much of the reason for this is not only that performance levels are much lower than is generally expected, but also that the lack of reliable and valid performance measures has led to: a) a lack of comparability amongst the great number of devices that are currently available, and b) problems arising from people having unrealistic expectations regarding the capabilities of the technology. Alternative measures are proposed.
A Voice Input Programming System BIBA 359-368
  D. Crookes; E. Murray; F. J. Smith; I. T. A. Spence
A system is being constructed which allows a user to input, edit, compile and execute Pascal programs by voice only through an inexpensive isolated word speech recogniser on an Apple IIe microcomputer. The aim is primarily to provide a programming tool for a handicapped person exploiting the well defined syntax and semantics of a programming language to improve the performance of the word recogniser. However, our long term aim is the much more difficult problem of applying syntactic and semantic techniques to improve the recognition of spoken natural language.

Design Issues: Models of Programming

A Model of Program Designer Behaviour BIBA 369-379
  Jawed I. A. Siddiqi
Findings of a research investigation into problem decomposition strategies used in program design are introduced. A semantic model of designer behaviour is advanced, that views program design as a problem solving task involving decomposition and elaboration. The former activity is viewed as goal generation, whilst the latter activity is considered to consist of 'clustering' of components and their allocation to the existing decomposition structure. The paper details those aspects of the model which relate to the latter activity. It argues that clustering is performed in terms generic categories and provides a rationale for errors made in this process. Results on subjects error frequencies from an observational experiment provide empirical support for the model. Moreover, it is noted that the solution based on a high quality decomposition corresponds to the solution that preserves generic grouping.
Towards a Framework for Human-Computer Discourse BIBA 380-393
  Mark Elsom-Cook
Research communication between humans and computers requires an interaction which is a structure related to the goals of the participants in that interaction. Psycholinguistic research has attempted to analyse this structure in terms of a syntax of discourse, a division based upon topic change and a division based upon functional sub-components. This paper describes the means by which a system for tutoring programming languages (IMPART) blends the more specific knowledge with general knowledge about interaction to produce a structured teaching interaction.

Design Issues: Graphical Interfaces

The Effect of Advanced Workstations on CAD BIB 394-403
  B. Gay; A. P. H. Jordan
Interactive Graphical Tools in the Social Sciences BIBA 404-414
  Nick Ryan
The interactive use of computers by social scientists has been seen primarily in the area of statistical analysis with such packages as GLIM and MINITAB. Alternative ways of viewing data may be explored whilst sitting at a terminal, and this has led to an increasing use of computers for exploratory analysis. The quality of output has been improved in some cases, although few packages are capable of producing results which may be considered fit, from a typographic viewpoint, for publication. The degree of interaction of much of the available software remains limited and the style of interaction, using a conventional display screen and keyboard, remains primitive.
   Modern microcomputers and, in particular, the more powerful personal workstations, provide viable alternative methods. In addition to their basic processing capabilities, many of these machines provide extensive graphical and word-processing facilities which may be combined in an integrated approach to analysis and report preparation. The use of these facilities is leading to an increased awareness of the potential applications of computers, particularly amongst those who, in the past, have had little use for statistical packages.
   This paper describes how the interactive graphical capabilities of a powerful personal workstation may be used to manipulate information which is often conceived in graphical form. It will be illustrated by reference to the analysis of genealogical data, one of many applications of networks in social research. The methods employed to shield relatively inexperienced users from the complexities of a multiple window environment are fundamental considerations in the design of this software.
A Window Manager with a Modular User Interface BIBA 415-426
  William Newman; Nigel Stephens; Dominic Sweetman
Modern graphics systems need to support multiple independent areas for graphic interaction on a single screen. Software packages (Window Managers) to support this are becoming more common on personal computers and workstations. It remains a challenge to construct a window manager which provides an effective front-end to UNIX
   The paper describes how we have approached this problem by dividing the system into a flexible but low-bandwidth manager and a less flexible but high-performance screen driver. The division is based upon our understanding of the general user interface requirements of window systems.