11. BIBLIOGRAPHY

The participants at the Workshop were asked to identify papers particularly relevant to their own position paper and give the reasons. These are given below:

Carl Engelman

1) Indicated:

Interactive Design of Computer Consultants by Carl Engelman, Proceedings of International Conf. on Interactive Techniques in Computer Aided Design, Bologna, Italy 1978.

Reason:

This paper discusses computer programs that act as interactive consultants focusing on the support such systems present for the interactive acquisition and modification of their own domain specific knowledge. It results that such systems incorporate the primary mechanisms for their own evolutionary computer aided design.

2) Indicated:

Acquisition of New Inference Rules by Randall Davis, Fifth International Joint Conf. on Artificial Intelligence, Cambridge, Mass.

Reason:

The program discussed in the paper is designated to function as an assistant in the task of building large, knowledge based systems.

Alan Shaw

1) Indicated:

Software descriptions with flow expressions by A. Shaw, IEEE Trans. Software Engineering, vol. SE-4, No. 3, (May 1978), 242-254.

Reason:

1) extension to regular expressions to describe concurrences, interleaving, cyclic activities, and synchronization.

2) applications in a wide variety of software areas, including command languages.

Relevance: a model for command language specification and processing, that includes finite state model but overcomes some of the limitations of a finite state model.

2) Indicated:

Definition and use of higher level graphics input tools by J. van den Bos, Proc. SIGGRAPH '78, Computer Graphics 12, 3 (Aug. 1978) , 38-42.

Reason:

1) Uses an expression language to specify and define input sequences from logical and physical devices.

2) Proposes a programming language that uses input expressions to drive the system.

Relevance: A new approach to programming graphics systems, with some of the virtues of the state driven techniques but more powerful potentially.

F.R.A. Hopgood & D.A. Duce

1) Indicated:

Rand Intelligent Terminal Agent Design Philosophy by Anderson & Gillogly, Rand Report No. R-1809-ARPA.

Reason:

There are a number of position papers which imply that a single syntactic translation of input is all that is necessary in a graphics system. We think that such an approach may be alright in the short-term but will seriously degrade the progress of interactive systems in the future. Graphics input will be a rich multi-channel mode of working in the future.

Gergely Krammer

1) Indicated:

Definition and Use of Higher-level graphics Input-Tools by J. van den Bos, Proc. ACM-SIGGREPH, 1978, Atlanta, GA. USA, pp. 38-42.

Reason:

This paper considers Algol-68 as an implementation language for interactive systems, A method for the formal specification of input is discussed.

2) Indicated:

Techniques for Processing Interactions in Fortran by G.A. Butlin, Notes for Seillac I, 1976.

Reason:

This paper considers FORTRAN as an implementation language for interactive systems. Various aspects of interaction and their solution in FORTRAN are discussed.

3) Indicated:

A Complete Calculus for Interactive Programming and for Communicating Sequential Processes by T. Gergely and L. Ury, SZAMKI Report, 1979.

Reason:

Seillac-II is expected to produce both

• a consistent, state-of-the-art overview of the theoretical aspects of interaction; and
• a consistent foundation for state-of-the-technology interactive systems. While I have dealt only with the second in a pragmatic manner, this report is a theoretical treatment related to our theme.

4) Indicated:

Conversation Theory by G. Pask, Elsevier, 1976, Amsterdam, New York book.

Reason:

Seillac-II is expected to produce both

• a consistent, state-of-the-art overview of the theoretical aspects of interaction; and
• a consistent foundation for state-of-the-technology interactive systems. This is an excellent book for the first.

Peter Bono

1) Indicated:

Special Issue on Graphics Standards by ACM Computing Surveys, vol. 10, no. 4 (Dec. 1978) especially pp. 402-428 and pp. 458-459.

Reason:

Explains more fully than the GSPC Core Report (Computer Graphics, vol. 11, no. 3 (Fall 1977) how the CORE Graphics System Input facilities were intended to be used.

2) Indicated:

Considerations for Future Programming Language Standards Activities, by J.A.N. Lee, Comm. of the ACM, vol. 20, no. 11 (Nov. 1977), pp. 788-794.

Reason:

Should the Seillac-II discussions on input get around to issues of standardization, then this paper explores the philosophy, politics, and methodology of formal standardization efforts in a most thought-provoking manner.

Ketil Bo

1) Indicated:

An Interactive Geometrical Design System with Handwriting Input, by M. Hosaka and F. Kimura, Proceedings IFIP '77, North-Holland.

Reason:

This publication is an important step towards a natural way of inputting bulk data directly from the hand of the user.

2) Indicated:

Logic Diagram Recognition by Divide and Synthesize Method by S. Kakumoto, Y. Fugimoto and J. Kawaski, AI and PR in CAD, North-Holland, 1978.

Reason:

This publication is an important step towards an automatic method for input of bulk data.

Richard A. Guedj

1) Indicated:

Towards a Theory of Telling by B.N. Lewis and J.A. Cook, International Journal of Man-Machine Studies, 1, (1969), pp. 129-176.

Reason:

• In relation with what I called the "asymmetry of the dialogue" the form of communication by the computer is more likely to be related to the concepts of telling than to anything else. Lewis and Cook are making in this paper an extensive study of preliminary steps towards a theory of telling.
• My paper also draws the attention on the necessity to take into consideration "modes of communication". Lewis and Cook (1969) although restricting their study to a Theory of Telling suggest two criteria that should help us evaluate a classification of modes of communication:
  1. classification should be associated with a plausible and rich model of man.
  2. classification to be shown, should be demonstrably applicable in the real world.

2) Indicated:

Behavioural issues in the use of interactive systems, by L.A. Miller and J.C. Thomas Jr., International Journal of Man-Machine Studies 9, (1977), pp. 509-536.

Reason:

This paper presents an extensive list of so-called "behavioural issues". It is a recent review of studies such as: the influence of the system's response time on user's behaviour, the list of properties desirable for an interactive system. However, it is difficult to weigh and organize those properties.

Seillac-II should have a different approach looking, for a framework and models; should go much beyond a listing of issues.

3) Indicated:

User-perceived quality of interactive systems by W. Dzida, S. Herda and W.D. Itzfeld, IEEE Trans. on Software Engineering, vol. SE-4, no. 4 (July 1978), pp. 270-276.

Reason:

This reference does not appear in the references listed in my paper. However, my paper states as a pragmatic and short term goal the set up of guidelines to improve the quality of interactive systems. Just a little bit later I read Dzida's paper on user-perceived quality of interactive systems (Dzida & all, 1978). In retrospect, this reference is important to my paper. Any designer for interactive systems must be fully aware of those qualities and of major user requirements "problem adequate usability" and "ease of learning".

Relevance: the user-perceived qualities of interactive systems to guidelines and principles may be revealed at Seillac-II Workshop.

4) Indicated:

Interactive Command Language Design Based on Required Mental Work by S. Treu, International Journal of Man-Machine Studies, 7, (1975), pp. 135-149.

Reason:

It is a reference to the concept of types of computer-aided tasks (CAT). Necessity to study CAT. Treu proposes 5 types of CAT. Treu proposes an interesting scheme for user's commands.

5) Indicated:

The user interface in interactive systems by J.L. Bennett, Annual Review of Information Science and Technology, Caudra (Ed.), vol. 7, (1972), pp. 159-196.

Reason:

In relationship with the importance of a conceptual framework for interaction I listed Bennett's reference for one important idea. (There are many others in his paper.) The importance of an effective transfer to the user's mind of the conceptual framework that guided the design of the system: "... the impact on system performance of the user's concept of the tool is too important to be left to chance".

Jan van den Bos

1) Indicated:

The Art of Natural Graphic Man-Machine Conversation by J.D. Foley and V.L. Wallace, Proc. IEEE, vol. 62, 4, (1974), pp. 462-471.

Reason:

Although not worked out, this paper, especially in Section III: "Action language" spells out that man-machine discourse should be structured in sentences the tokens of which are input actions. It should be considered, in hindsight, as a precursor to the ideas presented in my position paper. It is only unfortunate that the above authors did not work out these ; ; ideas but instead chose to illustrate input actions by means of finite-state diagrams because the latter do usually more to hide structure rather than reveal it.

2) Indicated:

Status Report of the Graphics Standards Planning Committee by ACM-Siggraph, vol. 11, no. 3, (1977), pp. 1156-1160.

Reason:

The description of input primitives in the CORE especially the Association mechanism, more than anything else has triggered our thinking about structuring primitives for input, because they are so sorely lacking from the CORE.

Jim Foley

1) Indicated:

The Art of Natural Graphic Man-Machine Conversation by Foley and Wallace, Proc. IEEE, vol. 62, 4, (1974), pp. 462-471.

Reason:

One can look at interaction from a language (semantics, syntax) view or from a psychological view. Basic user actions/inputs can be logically categorized.

Jean Weydert

1) Indicated:

Behavioural issues in the use of interactive systems by L.A. Miller and J.C. Thomas Jr., Int. J. Man-Machine Studies 9 (1977), pp. 509-536.

Reason:

This paper tries above all to present a structure for the discussion of issues in the use of interactive systems. The basic idea consists of identifying the general behavioural problems when nothing is known about the particular user or its particular task.

A classification of the characteristics of interactive systems shows at the same time the connection between certain problems, above all the structuring of information on different levels.

A well structured description of the state-of-the-art of designing interactive systems, emphasizing this last aspect, could be part of the results of Seillac-II.

2) Indicated:

Controlling User Interaction by D.J. Kasik, ACM SIGGRAPH/Computer Graphics vol. 10, no. 2, (Summer 1976), pp. 109-115.

Reason:

This paper presents a structured approach to controlling user interaction. The interaction is centralized for an entire application, which allows it to be designed along uniformly for all system components. This helps both, the user in choosing a path through the application and the programmer in structuring interaction before coding new modules. As a concept for the realization of this approach, the command tree has been chosen, which above all allows easy and consistent definition of features for controlling interaction. A classification of general interaction mechanisms could emphasize the use of such models.

3) Indicated:

Spatial Concepts as an Organizing Principle for Interactive Bibliographic Search, by J.L. Bennett, Interactive bibliographic search/the user-computer interface, pp. 67-83, AFIPS-Press, Montvale, New Jersey.

Reason:

A characteristic of the communication with an interactive system is the constant reformulation of the user's task on the basis of information gained during its execution. This process must be integrated in an organizing structure which tells the user his actual status and shows him how to go on. The paper presents a spatial framework as an overall organizing structure, which allocates the different types of information the user acts upon, to different functional 'areas'. The command language is conceived as a uniform tool for controlling the notion between the inside of these areas. A characterization of overall organizing concepts, such as spatial model, would be of great interest.

W. Dzida, S. Herda and W.D. Izfeldt

1) Indicated:

User Perceived Quality of Interactive Systems, by W. Dzida, S. Herda and W.D. Izfeldt, IEEE Trans. Software Ent., vol. SE-4, no. 4, July 1978.

Reason:

The quality of interactive systems is empirically investigated. Factors of system quality such as ease of learning and problem adequacy are described in terms of categories as perceived by users. These user-oriented design goals are theoretically substantiated by a model of human performance. This leads to the concept of task-oriented man-computer interaction. Seillac-II Workshop should emphasize the user's point of view.

2) Indicated:

Allgemeine- und Ingenieurpsychologie by W. Hacker, Huber, Bern, 1978. 2) Auflage, Lizenzausgabe des VEB Deutscher Verlag der Wissenschaften, Berlin.

Reason:

A starting point of our investigation in the design of man-computer interaction is a model of human performance. From the TOTE performance model it is derived that:

• a decomposition of human tasks should be hierarchically organized;
• the theoretical separation of knowing, from doing, should be considered in the design of various tours in dialogue.

Seillac-II Workshop should encourage interdisciplinary cooperation.

Setsuo Ohsuga

1) Indicated:

A Machine Oriented Logic Based on the Resolution Principle by J.A. Robinson, JACM vol. 12, no. 1, pp. 23-41.

Reason:

This paper presents the theory of the resolution process in the form of a system of first order logic. The completeness of the system is proved. In my opinion the interactive system is going to be endowed with more intelligence, and the deductive inference is, the most basic function of intelligence. Therefore this paper has deep concerns to the purpose of the Seillac-II Workshop.

2) Indicated:

Predicate Logic as Programming Language by R. Kowalski, Proc. IFIP Congress 74, Stockholm, Sweden, pp. 569-574.

Reason:

The predicate logic is interpreted as a programming language. It is entirely user oriented. This paper deals in a preliminary way with some of the issues raised by the consideration of predicate logic as a programming language. In my opinion, an interactive system must be able to deal with knowledge and its representation must be such that it can be transformed into language expression and program. This paper concerns the latter requirement.

3) Indicated:

Recent Investigations in Relational Data Base Systems by E.F. Codd, an ANSIA3/SPARC Interim Report 1975.

Reason:

The objectives and characteristics of the relational approach to the management of large formatted and integrated data bases are reviewed. Since the data base technology is expected to have direct concern to interactive systems, we need to pay attention to the trend of the technology.

4) Indicated:

A Panel on Knowledge Representation by D.G. Bobrow, 5th IJCAI 1977, pp. 983-992.

Reason:

Different forms of knowledge representation are compared. In my opinion the interactive system will become more and more intelligent so that it can support man's intellectual works. The intelligence of systems is a function of the knowledge they contain and the utility of this knowledge is in part dependent on the form of its representation. In this paper, eight different forms of representation of knowledge are listed with their features. It is a good survey of the current state of the art.

Rolf Eckert

1) Indicated:

Use of the Concept of Transparency in the Design of Hierarchically Structured Systems, by D. Parnas and D.P. Siewiorek, Comm. ACM 18, 7 (July 1975), pp. 401-408.

Reason:

Specification techniques force a software designer to think more precisely about a problem to be solved. From the Seillac-II Workshop I suggest that we define 'WHY1 and 'WHAT' is to be done in the near future.

J.P. Crestin

1) Indicated:

Contribution a L'etude des Systemes Interactifs by C. Queinnec, These ENSTA-Paris VI, Octobre 1978.

Reason:

An interactive program may be described by the set of its traces, i.e. pairs (sequence of inputs, sequence of outputs). These traces are formally generated by an automaton, called a Formal Dialogue System (FDS), with usually an infinite number of states. Decomposition techniques are given which allow finite decompositions of the FDS. A link with the finite state diagram's method (see Newman for instance) is derived from this decomposition. Thus finite state diagrams show up as either a very simple case of FDS or as a framework for more complex FDS.

The definition of a program obtained by a FDS is totally independent of any implementation and of any communication device. It is thus a way to gain device independence for graphical interactions. Moreover, equivalence between graphical and non-graphical interactions may be thus defined, permitting to have the same mainframe program used with interactive graphics or with pure alphanumerical dialogue.

2) Indicated:

Using Assertions About Traces to Write Abstract Specifications for Software Modules, by W. Bartussek and D.L. Parnas, Information systems methology proceedings, Bracchi and al (eds.), Venezia 1978, Lecture Notes in Computer Science, Springer-Verlag.

Reason:

Comparison of different methods for abstract specifications of programs. Proposal to define a program by its traces, illustrated by several convincing examples. Leads independently to similar conclusions as those of Queinnec. The presentation by Queinnec is more algebraic, this one based on more pragmatic considerations. They should lead to an efficient approach of interactive programming. It is a very good bibliography on program specification.

3) Indicated:

Functions Describing Interactive Programming by J. Kupka and N. Wilsing, International Computing Symposium (1973), North-Holland, 1974.

Reason:

A definition of Formal Dialog Systems (FDS) which was at the origin of the work by Queinnec. It introduces the mathematical concepts of abstract input and output commands and interrupts.

Thomas P. Moran

1) Indicated:

Introduction to the Command Language Grammar by T.P. Moran, 1978, Xerox PARC, Technical Report SSL-78-3.

Reason:

It proposes a notation for describing human-computer interfaces that:

• defines the space of command language systems;
• captures the human's conceptual model of the system;
• is a representation for designers to specify user interfaces.

2) Indicated:

The Keystroke-Level Model for Predicting User Performance Time with Interactive Systems, by S.K. Card, T.P. Moran and A. Newell, 1979, Xerox PARC Technical Report SSL-79-1 (submitted to Communications of the ACM).

Reason:

Illustrates that models of user performance can be built, that are:

• accurate enough;
• easy to use.

3) Indicated:

Applied Information-Processing Psychology: the Human-Computer Interface, by S.K. Card, T.P. Moran and A. Newell, Erlbaum, in preparation (draft should be ready by April, 1979).

Reason:

Proposes the form of an applied psychology of user interfaces:

• it is theory-based;
• it is based on calculations rather than on experiments;
• it will apply to all stages of the design process;
• it will by applied by systems designers themselves, rather than by "human factors specialists".