is in the position to formulate pertinent requirements3.

Overcoming the Barriers the organisational, tool, dialogue and i/o for interfaces. The "communications barrier" between applier, user Sufficient empirical evidence has accumulated by AN SOFTWARE PROCESS MODEL and on end-user theITERATIVE-CYCLIC one hand and the software developer now that task and user oriented procedures in to show on the other is essentially due to the fact that technical software development not only bring noticeable savings in costs, Matthias Rauterberg intelligence is only inadequately imbedded in the social, but also significantly improve the software produced historical and political contexts of technological development. [6] [9] [17]. How then, can the both barriers mentioned above Communication Work between and those Organisational involved in Psychology the Unit, Swiss Federal Institute of concept Technology be overcome? The answer is: the of Nelkenstrasse 11, CH-8092 Zuerich, develop(ETH) ment process can allow non-technical facts to slip "optimisation". rauterberg@czheth5a.bitnet through the conceptual net of specialised technical language,3.1 The Optimisation Cycle which therefore restricts the social character of the Abstract 2. Barriers in the Framework of technology to and instrumental. the functional Systems theory distinguishes between control (i. The current state of traditional software development is Traditional Software Development The application-oriented jargon of the user e. forward or open loop control systems) and feed surveyed and essential problems are investigated on flounders on the technical jargon of the developer. This gap can regulation (i. e. feedback or closed loop control systems). the basis of system theoretical considerations. The concept The specification barrier is an immediate problem only be bridged to a limited extent by purely linguistic The following are minimum conditions for a feed of optimisation cycle is presented. The relationship even at a cursory glance. How can the software means, because the fact that their respective semantics are forward or open loop structure, that functions: between developer several different kinds of local optimisation cycles with ascertain that the client is able to specifiy the conceptually bound make the ideas applied insufficiently (1) precise knowledge of the response of the system known techniques for user-participation is discussed requirements for the subsystem to be developed in a the complete and controlled, i. e. of the relation between controller precise. Overcoming this fuzziness requires creating jointly being and in- into a general concept of participatory accurategrated te way which will notand be modified while project is put on the one hand the output andthe interference experienced, perceptually shared contexts. Beyond verbaloutsoftware development. The pros and cons of essential problems being carried out? The more formal and the as changes in the specifications on detailed the communication, visual means are the ones best suited to such known medium to obstruct optimal software development and possible used by the client to formulate requirements, theaffect easier (2) precise knowledge of those quantities whose this purpose. The stronger the perceptual experience oneother; it isthe ways them are for the system software is detrimental developerto tothe incorporate intended these influence into an has ofof solving the semantic context of the other, the easier it is to on ference or perturbance, such as technical feasibility, considered. ap(interpropriate software system. But this presumes that the overcome the communications barrier. the system has a response delay, then a prognosis etc.); if 1. Introduction client has command of a certain measure of expertise. is At its best, software development is a procedure for for at to least for the period the needed However, the these client interferences is not prepared acquire this or of perhaps op- designing a product with interactive properties timally (3) knowledge of procedures controller delay; is in Analysis of current software development processes part not in a position to do sofor deriving before the beginning for supporting the performance of work tasks. Because output from such information. brings to light a series of weaknesses and problems, of the software development process. It is therefore computer science has accumulated quite a treasure trove of These conditions are hardly ever met in practice. That the necessary sources of which lie in the theoretical concepts applied, to find and implement other ways and means, using very broadly applicable algorithms, software developmentis why is it is constantly necessary to supplement or the from traditional procedures followed (especially project informal through semi-formal control by regulation [18]. to formal specification inreplace creasingly focussing attention on those facets of management) as well as in the use of inadequate cost analysis methods. The application of the highly effective regulation application-oriented software which are unamenable to models [5]. The literature contains an ample store of It would be a grave error with dire consequences principle actually only requires a knowledge of thoseto algorithmic treatment. While the purely technical aspects of a proposed solutions based on current practice in software assume that clients usually from middle controller outputs which steer the people ouput in the the desired software product are best dealt with by optimisation procedures development. These point to the significance of participation and upper by We echelons of the management are able to provide direction. designate Test-Action-Cycle based on the attuned to a technical context, the non-technical context all pergroups affected. Analysis of these cases shows that tinent and adequate information on all requirements gulation" principle as optimisation cycle. An important of the application environment aimed at requires the "rethere essential are for an three barriers to optimisation: the interactive system. cycle As a result, the following mension of software the optimisation is its length, i. e. the implementation of optimisation procedures of a different dispecification difbarrier, the communication barrier and the ferent perspectives must be taken into consideration time required to complete the cycle once. Depending in nature. It would be false indeed to expect that at the outset optimisation the barrier. analysis phases. on the ofand nature the specification activity and the testing, the length can be of areorganisation of a work system any single larger The Appliers Perspective: Every person in a position to Speaking quite generally, one of the most important anything from a matter of a few seconds to up to possibly group of could have a complete, exact and persons make a contribution to defining the requirements for problems lies in coming to a shared understanding by several years. The longer this period, the more costly the comprehensive view of the ideal for the work system to be set up. Only tal thework to- system is considered to be an applier. This perall the affected goups of the component of the timisation cycle. It is the with aim of user-oriented software opduring the analysis, evaluation and planning processes spective often coincides that of the client, and worksystem to[14] that is, to find the answers to the be automated development to incorporate an as efficient optimisation are the people involvable to develop an increasingly clear into consideration general requirements concerning the takes questions of if, where and how for the planned cycle as possible into software development procedures [9] picture of what it is that they are really striving for. This efis fectivity, organisational structures, project costs, global implementation of technology, to which a shared [15] [16]. basically why the requirements of the applier seem to plication and implementation goals for the technical apcommitment can be reached. This involves, in particular, system, The optimisation as well as the criteria desired are effect all relevant on the technical total work subchange they do not really change but simply determining all the characteristics of the work system that are to be system. The appliers perspective thus embraces the to and social factors [19] [21]. Testing ascertains the extent become concrete within the anticipated boundary constraints. planned anew. Every work system comprises a social and a mentsthe for optimisation the organisational requirecriteria are met, subject to the This process of crystallisation should be allowed to unfoldwhich technical subsystem. An optimal total system must The Users Perspective: Users aretaken thosecould persons who interface. boundary constraints. The action come as completely, as pertinently and from a global integrate both simultaneously. the results obtained from using the software system need from aof extremely range different procedures, methods or for perspective as inexpensively as possible. Completeness can be forming their tasks. The dominant factor influencing perIn order to arrive at the optimal design for the total techniques. All of this depends on the nature of the work reached by ensuring that each affected group is perspective is human-to-human communication with their working system, it is of paramount importance to Interference could come from the three barriers as well involved at representatives. least through Iterative, interactive output. end-users (e. g. heads, secretaries, etc.) and their the regard the social subsystem as a system in its own right, as from technical and social problems in the realising progress makes the ideal concept increasingly concrete. There ion usually includes requirements for tool the contributendowed with its own specific characteristics and conditions, and project. Of course current software development also avails are methods available for supporting the process of The End-Users Perspective: End-users are all those who interface. a s ystem to be optimised when coupled with the itself of the principle of control in various places. What communication which ensure efficient progress [15] rectly use the software system as a work tool. This ditechnical subsytem. we [16]. group 0-8186-2830-8/92 $03.00 1992 IEEE 600 601

have in mind here are decisions made and directives action and decision-making; 2. adequate time available; physical 3. sufficient activity; 4. concrete contact with issued by the client, the project management or other bodies rial and social conditions at the workplace activities; 5. as a consequence of experience, ignorance, exercise of mateacpower or purely and simply the pressure of time. It is frequently tual use of a variety of the senses; 6. opportunities for the case that feed forward control systems operate morevariety when executing tasks; 7. task related econocommunication mically than is ever possible with regulated systems and immediate interpersonal but contact. only if the named conditions prevail! This is one Once those concerned are sufficiently clear about important which are amenable to automation, the next step reason why the attempt is made to come as close as functions possiwhich ble to a particular control system in software should be taken is to test the screen layout on the enddevelopment, users namely the Waterfall Model". If, however, the barriers with hand-drawn sketches (the extremely inexpensive dispen cussed above, are taken seriously, then we must and paper method). If the range of templates is very determine those places in software development procedures at large, then a graphics data bank can be used to manage the which temcycles are indispensible in a software process plates produced on a graphics editor. The use of model. prototyping tools is frequently inadvisable, because tool-specific 3.2 The Analysis Phase presentation offers a too restrictive range of possibilities. The effect of the structuring measures taken can be The analysis phase is frequently the one most assessed with the help of discussion with the end-users, or by neglected. This is essentially due to the fact that methods and means of checklists. techniques need to be used primarily the way occupational The use of prototypes to illustrate the dynamical and organisational sciences have developed and applied teractive and inaspects of the tools being developed is them [13]. Inordinately high costs incur from the indispensable for specifying the dialogue interface. However, troubleshooting required because the analysis was less than optimal [6] prototypes should only be used very purposefully and [20]. The time has come to engage occupational and selectively to clarify special aspects of the specification not organisational consultants at the analysis stage who have been indiscriminately. Otherwise there looms the inescapable especially trained for software danger of investing too much in the production and development! While traditional software development partly maintenance of A very efficient and inexpensive display goods. includes a of the tasks in the work system, global analysis variation is provided by simulation studies, for example, with the analysis of work activities and their effect is largely excluded from use of hand prepared transparencies, cards, etc. which consideration. The results of the detailed analysis of the appear before the user in response to the action taken objective conditions of a work system need to be [9]. 3.4 The Implementation Phase supplemented with the subjective conditions experienced by the employees if the organisational measures to be drawn up jointly byaffected are to have a chance of finding all those Having invested the necessary effort in optimising consensus. Yaverbaum and Culpan [24] determined important theboth analysis and specification phases, it is time to enter criteria the for further qualificational and organisational measures implementation phase. This phase is made up of the by folmeans of Job Diagnostic Survey (JDS). lowing three steps [2]: 1. design of the programme architecture; 2. design of the individual programme modules 3.3 The Specification Phase (object classes, etc.); 3. coding and debugging. It is important to check the extent to which already available software Once the analysis of the work system to be can be before commencing coding. The use of re-used optimised has been completed, the next stage is to mould the velopment software deenvironments can result in increases in results obtained into implementable form. Methods of productivity of up to 40% [4]. specification with high communicative value are recommended here. The first thing is to determine if and where it 3.5 The Trial and Assessment Phase makes sense to employ technology. Although the view is still widely held that it is possible to use technology to Once a working version is available, it can be put to eliminate the deficiencies of an organisation without test in usability studies (user-oriented benchmark questioning tests: the structures of the organisation as a whole, the [17] [19]) in concrete working situations. This is the conclufirst sion is nevertheless usually a false one place where it is possible to clarify the problems with [10]. the The intended division of functions between man and actual organisational and technical environment. Whiteside machine is decided during the specification of the tool et al. [23] point to the necessity of empirical evaluation interface. The tasks which remain in human hands must techniques as opposed to laboratory tests in have concrete the following characteristics [22]: 1. sufficient freedom working situtations. By contrast to laboratory studies, of such 602

field studies take into account the aspect of ecological (e.g. scenarios, Wizard of Oz studies) are available vafor de- completely new systems without requiring lidity [9] [23]. An operational version of the system veloping must any be available for such methods to be used. This is only special hardware or software. Spencer [19] presents a possumsible with the framework of a version mary of techniques for the analysis and evaluation of concept. interactive computer systems. Comparative studies, e. g. 4. Local and Global Optimisation Cycles useroriented benchmark tests [17], can be undertaken after the second time through, when working with a version conThe literature contains a whole series of suggestions cept, for then there are at least two versions for available. incorporating optimisation cycles into software The global optimisation cycle together with its incordevelopment procedures [3] [7] [13] [17]. The various authors porated local cycles, can be subdivided into four have different emphases in their concepts, depending on regions (Quadrants I IV of Figure 1). Quadrant I includes the their anaown background and experience. lysis and approximate specification. Communicative, inforWe now turn to a discussion of those aspects which mal methods are mainly applied here. Detailed specificaneed to be borne in mind when commencing as well as tions are optimised in Quadrant II using prototypes. throughout the course of an individual optimisation The specified hardware and software are implemented in cycle. QuaThe type of software to be developed has proved to be drant III and the test data assessed. Quadrant IV one comprises of the essential factors governing software assessment, maintenance and optimisation of the development. system in operating environment. The following four types can be distinguished the real-life Type A: Specific application for an internal division; [20]: The effort spent on optimisation in each individual the division placing the order and the one developing both quadrant varies according to the type of the project and of software belong to the same the the type. However all software development project Type B: Specific application for external users; the divicompany. analyses completed to date indicate that increasing the effort sion placing the order and the one developing the expended on optimisation in Quadrant I reduces maintenance belong to different companies. software in Type C: Standard solutions for external users; this Quadrant IV and saves costs [13] [20]. arises from projects of Type A or Type B, when often

5. Participatory Methods and Techniques software solutions (Type A, Type B) are specially individual for further users. adapted In the introduction we used systems theoretical Type D: Standard software for an anonymous user considerations to explain how each individual optimisation group. The global optimisation cycle begins at Start A consists of a test and an action component which are (Figure 1) when developing completely new software and at cycle ably coupled. Each component can be of a widely Start in theB case of further development and refinement ofsuitvarying nature. Table 1 provides an overview of the main focus existing technology. Different concept-specific local effort, of the nature of activity, and of the tests, the optimisation cycles are used to optimise specific work tasks, of outcome and the expected range for the length of the depending on the particular type of the project. It is up to the cycle. The shorter an optimisation cycle is, the more rapidly project management to settle on the actual procedure and therefore the more often it can be used to reach a and this decision is reflected in the development form truly optimal result. chosen. In order to reach the goals of a work-oriented design concept [21] [22] the first project phases (requirements A major problem in assuring adequate meshing of diffeoptimisation cycles is synchronisation. If several analysis and definition) should be replete with a range ofrent optimisation cycles are simultaneously active at different different optimisation cycles. The system design is settled in on after carrying out a complete as possible clarificationplaces ofthe iterative-cyclic software development concept 1), then these must be suitably synchronised. This is the requirements of the client (work and task analysis, (Figure particular important, being the only way to minimise division of functions between humans and computer, etc.). The sistencies with to the overall development process. If, question remains as to which design specifications inconexample, there are additional consultations with the for need clarification by means of additional optimisation user and if specification analyses are undertaken parallel to cycles. Simple and fast techniques for involving users implementation phase, then it could easily happen that include groups with various communication aids the discussion the programmers end up writing programmes for the waste (metaplan, layout sketches, screen-dumps, scenarios, etc.), paper basket because they are working to superseded questionnaires for determining the attitudes, opinions specifications. This problem is caused by the differing lengths and requirements of the users, the walk-through of the optimisation cycles involved and it becomes technique for clarifying all possible work steps, as well systematically blatant whenever the separate optimisation cycles have not as targeted interviews aimed at a concrete analysis of the been ad- synchronised. equately work environment [7] [13] [23]. Very sound simulation methods
603

I : Analysis Phase
START A

Trial and Assessment Phase : IV

START B

Protocols

Discussions, Work-Shops, Division of Functions between Humans, Global Task Analysis

Statistics, Interview Results, Assessment Results

Usability Test, useroriented Benchmark Test

"Final Version"

[End] User Requirements Human-Machine Division of Functions, Feasibility Studies Provisional Definition of Requirements Production of Simulations Optimised Definition of Requirements

Assessment Results Operation and Maintenance

Evaluation of Simulation

Deliverable Version

Simulation

Benchmark Test Beta Test

Running Version Alpha-Test of Correctness, Performance, etc.

Detailed Task Analyis Formal Specification

Provisional Design

Provisional "Release"

Test Results

Preparation of Formal Specifications Prototypes for - "Walk-Through", - Explorative Studies

Detailed Design Specification

Module or Object Oriented Programming

"Bug"

II : Specification PhaseImplementation Phase : III

Figure 1 Flow chart for an iterative-cyclic software process model showing the local optimisation cycles tween individual quadrants (I - IV). The systematic use of application and maintenance cycles within and bethe withrequirement feedback to phase thus subsumes the version concept as the global optimisation cycle (see also [7]). 604

MethodActionTestOutcomeCycle-Length
Discussion-Iverbal communicationverbal interpretationglobal design decisionsseconds - minutes Discussion-IImeta-plan, flip-charts, etc.visual & verbal inter- specific design-decisionsminutes - hours pretation Simulation-Isketches scenarios, Wizard of visual & verbal interspecification of the in-minutes - days Oz, etc. pretation put/ output interface Simulation-IIdraughting of stuctural blue- visual & verbal inter(semi)-formal descriptive hours - weeks prints, etc. with (semi)-formal pretation with suitable documents methods qualification Prototyping-Ihorizontal prototypingthinking aloud; specification of dialogue days - weeks walk-through component Prototyping-IIpartial vertical prototypingheuristic evaluationpartial specification of days - weeks application component Prototyping-IIIcomplete vertical prototypingtask-oriented benchspecification of appli- weeks - months mark tests cation component Versions-Irun through entire developmentuseroriented bench- first largely complete months- years cycle mark tests version Versions-IIrun through entire development useroriented bench- several largely complete months- years cycle mark tests versions

Table 1 Overview of Different Methods of User-Participation in the Framework of Optimisation Cycles

By the functional synchronisation principle we mean thods, going beyond their influence on the length of the fixing the sequence of the individual optimisation quacycle. drants in the sense of a functional phase distribution. Discussion Methods I & II (see Table 1): Discussion is This principle is used primarily in the Waterfall Model (the the method most frequently used, because it is fast1, milestone concept) and leads necessarily to a and to a certain extent informative (see the familiar dramatic crease in inthe total length of the cycle when applied communications barrier). But because it rests essentially on exclusively to the version concept. verbal communication, a series of misunderstandings purely One way to at least partly overcome this drawback arise which often never come to light or only do so can is to use the informational synchronisation principle is already when it too late. Discussion must therefore be mented with methods using visual communications approsupplepriate information links are established between the niques. techvarious optimisation cycles, so that every person in each cycle is kept informed about the current state of the cycles Simulation Methods I & II (see Table 1): Simulation which are active in parallel. This can be achieved using such methods comprise all techniques which illustrate the simple aids as document folders at a fixed location and system to be optimised in as realistic, visually work regular conference times. But technical support can also be way as possible. This ranges from simple quickly comperceivable pleted sketches, through template layouts to formal used (mailboxes, version data banks, information tion techniques. descriprepositories). Another important synchronisation principle is to enThe unequivocal advantage of formal analytical and sure that participation in the different optimisation de- techniques is that they force one to perform a scriptive cycles includes the same circle of people. However, this rough and detailed investigation of the domain to be principle often flounders on organisational forms based on thethodescribed. The analysis focusses on different aspects, division of labour, which is frequently encountered in on the particular procedure involved. However the software companies. These software development divisions depending concrete work environment of the end-user is almost completely require a reorganisation according to the occupational psychological criterion of completeness of work tasks [21]. Since neglected by most descriptive techniques. Another caveat: The more detailed this specification is, the more nobody can be in two places simultaneously, we call this hensible it becomes. Additionally, the more formal the principle the personal synchronisation principle. By incompremethod of representation, the more time consumed by their own admission software developers significantly its underrate the benefits of the informational and the personal 1 In the sense of the cycle length; an optimisation cycle synchronisation principle [11]. We list some further aspects to in the discussion method is limited to communicative be considered when applying the various participatory mesuch as statement-counterstatement, questionunits answer. 605

preparation. This is partly due to the fact that when less, there is no guarantee that the user is also a good users also participate, they first need adequate training in designer. One possible consequence of this is that only opeless rating and interpreting formal than optimal solutions end up being produced specifications. iteratively [8]. Industrial norms and standards, and design criteria Prototyping Methods I, II & III (see Table 1): As already interactive software [21] can be of assistance mentioned, prototyping methods make it possible to for here. quaint end-users with the procedural character of the acVersions Methods I & II (see Table 1): As vertical protobeing developed. Prototyping is an attempt to system typing is extended by increasingly enriching the image a part of or the entire application system in a prototype adequately with programmed functionality, it undergoes a smooth ing model for the future user to be able to grasp the metamorphosis into the version method. This workthe planned system works. It is in this sense that protoway gained in significance in the 80s because it clearly procedure typing provides a particularly effective means of matches the Waterfall Model in a software lifecycle. best cation between the user and the developer [1] communiversion cycle as a global optimisation cycle is seen here The [8].Since the use of prototypes is always within the test the proposed participatory development concept in the in component of some optimisation cycle, they must befeedback between the application phase and the phase (Quadrant IV & Quadrant I). Some authors speak readily modifiable. Thus the period elapsed between the prescription evolutionary software development suggestion by the user for modification and his assessment of of [12]. The basic advantage of the global optimisation cycle the modified prototype must be as short as possible, for lies unequivocally in the fact that it provides the first otherwise motivational problems arise. portunity to determine and test all interactions Two kinds of prototypes can be distinguished: the opthe practicality of each version within the usability and cal vertiand the horizontal. Horizontal prototypes contain between of the concrete work environment. If the system being only a very small number of application-oriented functions context veloped is sufficiently complex, then certain flaws in its from the end-product, the emphasis being mainly on the depresentation of the sequence of templates incorporated in adesign can only be detected in the real-life situation of the implementation and maintenance phase. In order to dialogue structure. Vertical prototypes, on the other hand, keep to a minimum the modifications required, the system go deeper. In a partial vertical prototype only a few must be applications functions are implemented and only in a rather developed from its very inception according to modern programming concepts (documentation, modular rudimentary fashion, whereas a complete vertical ob-programming, etc). ject-oriented prototype nearly implements every application function. This last structure, procedure comes closest to the prevailing notion of what a 6. Conclusion prototype in the traditional industrial context is. The disadvantages of prototyping lie in the fact that One of the principal problems of traditional software the prerequisites the developer must produce development lies in the fact that those who have been incomplete software (rapid prototyping) and then deal with primarily involved in software development to date have not critique from the user are difficult if not impossible to meet. been willing to recognise that software development is, in Another aspect is that the prevailing notion of a protocases, mainly a question of occupational and/or type refers to a fully functional product. But in the most tional planning. Were software development to be apcontext of software development, this is more properly calledorganisaproached from such a perspective, it would be planned an end-product and not a preliminary variant. The sad from the beginning to engage experts in occupational and truth is that as an industry, data processing routinely sational planning in the process of software design. delivers a under the guise of a finished product [1]. organiprototype would require interdisciplinary cooperation between Both ofaspects support the observation that when This theses occupational and organisational experts on the one hand prototyping is adopted, the best prototype is often a failed software development experts on the other. The project [5]. The fundamental idea of prototypes is to and extensive qualification required in each of these fields makes it iterate the design, not to FREEZE it [8]. Several authors place virtually impossible to dispense with such interdisciplinary great value on simpler and quicker participatory cooperation. techniques [7] [9] [15] [16] to banish this danger. We have presented here an iterative-cyclic software The method of prototyping in the context of an development concept which integrates solution proposals optimisation cycle faces the peril of leading to an inadequate developed to date for overcoming the specification, optimum. One reason for this is the fact that being communication and optimisation barriers based on the notion of occupied with a concrete prototype can blind one to an optimisation cycle. This consists of a test and an action fundamentally different alternatives. One remedy is to precede prototyping with and to superimpose suitable optimisation cyclescomponent, coupled to each other by feedback. The feedback loops recommended at various places in the aimed primarily at application contexts (see Figure 1). literature Neverthe606

7.GRUDIN J, EHRLICH S F, SHRINER R, 1987: Positionhave been incorporated into a global cycle as local optiing Human Factors in the User Interface Development misation ones. This global optimisation cycle can be Chain. In: Proceedings of CHI + GI (Toronto, 5th - 9th subdivided into four regions: the region where April 1987). New York: ACM. 125-131 requirements are determined (Quadrant I), the region of specification 8.JRGENSEN A H, 1984: On the Psychology of Prototyp(Quadrant II), the region of implementation (Quadrant III) ing. In: BUDDE R, KUHLENKAMP K, MATHIASSEN L, and the region of application (Quadrant IV). ZLLIGHOVEN H (eds.) Approaches to Prototyping. Different aspects of the work system to be designed Berlin: Springer. 278-289 9.KARAT C-M, 1990: Cost-Benefit Analysis of Iterative be can progressively optimised as one moves from quadrant to quadrant. The various perspectives of the ideal sought Usability Testing. In: DIAPER D et al. (ed.) HumanComputer Interaction - INTERACT '90. Amsterdam: take on progressively more concrete form. An appropriate Elsevier Science. 351-356 investment in optimisation in Quadrants I and II not only 10.KLOTZ U, 1991: Die zweite ra der Informationstechnik. helps to reduce the total cost (development costs and Harvard Manager 13(2):101-112 application costs), but also lead to optimally adapted 11.KRAUT R E & STREETER L A, 1992: Coordination in hardware and software solutions. This is due to the fact that all Large Scale Software Development. Communications of subthe ACM:(in press) sequent users are involved at least through representatives, 12.LEHMAN M M & BELADY L A, 1985: Program Evolution and can therefore inject their relevant knowledge into - Processes of Software Change. London: Academic. the design of the work system. 13.MACAULAY L, FOWLER C, KIRBY M & HUTT A, 1990: The more effort that is expended on optimisation in USTM: a new approach to requirements specification. the first quadrants, the less is needed in Quadrant IV. The Interacting with Computers 2(1):92-118 amount of effort required for optimisation in the second 14.NAUR P, 1985: Programming as Theory Building. Microand third quadrants depends in essence upon the processing and Mircoprogramming 15: 253-261 complexity of the work system to be designed. The investment in 15.NIELSON J, 1989: Usability Engineering at a Discount. Quadrant II can be minimised for example with the help of In: SALVENDY G, SMITH M J (eds.) Designing and Using modern prototyping tools and specification methods which Human-Computer Interfaces and Knowledge Based Systems. Amsterdam: Elsevier Science. 394-401 the user finds easy to understand. Employing powerful development environments and suitably qualified software 16.NIELSON J, 1990: Big paybacks from 'discount' usability engineering. IEEE Software 7(3):107-108 developers minimises the investment in Quadrant III. But 17.RAUTERBERG M, 1991: Benutzungsorientierte Benchfirst, we must start learning to jointly plan technology, mark-Tests: eine Methode zur Benutzerbeteiligung bei organisation and the application of human Standardsoftwareentwicklungen. In: ACKERMANN D qualification. &ULICH E (Ed.) Software-Ergonomie '91. (Reports of the

Acknowledgements

German Chapter of the ACM, Vol. 33). Stuttgart: Teubner. 96-107 The preparation of this paper was supported by the German Minister of Research and Technology (BMFT. AuT programme) 18.SCHIEMENZ B, 1979: Kybernetik. In: KERN W (ed.) Handwrterbuch der Produktionswissenschaft. Stuttgart: grant number 01 HK 706-0 as part of the BOSS "User oriented Poeschel. 1022-1028 Software Development and Interface Design" research project. 19.SPENCER R H, 1985: Computer usability testing and evaluation. Englewood Cliffs: Prentice Hall. References 20.STROHM O, 1991: Projektmanagement bei der SoftwareEntwicklung. In: ACKERMANN D &ULICH E (Ed.) Software-Ergonomie '91. (Reports of the German Chapter of 1.BOAR B H, 1984: Application Prototyping: A Requirethe ACM, Vol. 33). Stuttgart: Teubner. 46-58 ments Definition Strategy for the 80s. New York: Wiley. 21.ULICH E, RAUTERBERG M, MOLL T, GREUTMANN T, 2.BOEHM B W, 1981: Software Engineering Economics. STROHM O, 1991: Task Orientation and User-Oriented Englewood Cliffs: Prentice Hall. Dialogue Design. International Journal of Human Compu3.BOEHM B W, 1988: A spiral model of software developter Interaction 3(2):117-144 ment and enhancement. Computer (May 1988):61-72 22.VOLPERT W, 1987: Kontrastive Analyse des Verhltnis4.CRELLIN J, HORN T, PREECE J, 1990: Evaluating Evalses von Mensch und Rechner als Grundlage des Systemuation: A Case Study of the Use of Novel and ConventDesigns. Zeitschrift fr Arbeitswissenschaft 41:147-152 ional Evaluation Techniques in a Small Company. In: 23.WHITESIDE J, BENNETT J, HOLTZBLATT K, 1988: UsaDIAPER D et al. (eds.) Human-Computer Interaction bility Engineering: Our Experience and Evolution. In: INTERACT '90. Amsterdam: Elsevier Science. 329-335 HELANDER M (ed.) Handbook of Human-Computer Inter5.CURTIS B, KRASNER H, SHEN V, ISCOE N, 1988: A action. Amsterdam: Elsevier Science. 791-817 Field Study of the Software Design Process for Large Sys24.YAVERBAUM G J & CULPAN O, 1990: Exploring the tems. Communications of the ACM 31(11):1268-1287 Dynamics of the End-User Environment: The Impact of 6.FOIDL H, HILLEBRAND K & TAVOLATO P, 1986: ProtoEducation and Task Differences on Change. Human Relattyping: die Methode - das Werkzeug - die Erfahrungen. Anions 43(5): 439-454 gewandte Informatik 3:95-100

607

Proceedings Fourth International Conference on Software Engineering and Knowledge Engineering

SEKE'92
June, 15-20, 1992 Europa Palace Hotel Capri, Italy
Copyright and Reprint Permission: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy beyound the limit of U.S. copyright law for private patrons use of those articles in this volume that carry a code at the bottom of the first page, copy fee indicated provided the pre- in the code is paid through Copyright Clearance Center, 29 Congress MA 01970. For other copying, reprint or publication permission, write to IEEE Copyrights Street, Salem, IEEE Service Center, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855-1331. All rights Manager, reserved. Copyright 1992 by the Institute of Electrical and Electronics Engineers, Inc.

IEEE Catalog Number: 92TH0438-2 ISBN (paper): 0-8186-2830-8

IEEE Computer Society Press Los Alamitos, California Washington Brussels Tokyo