focusing on many relevant aspects: Development of organisations, work activities, in- formation ... user interface design within a specific application area.
Designing for Usability – Domain Specific Human-Computer Interfaces in Working Life. JAN GULLIKSEN Abstract Despite the tremendous development and increasing availability of computers in working life, severe usability problems are frequently observed, leading to bad user acceptance, negative stress, anxiety, and even health problems for the users. We need to create a scientific foundation for the studies of humans interacting with computer systems to achieve a specific work task in a physical and psycho-social work environment. Interdisciplinary studies of systems analysis, software engineering, cognitive and perceptual psychology, sociology, organisation theory, human factors, and design and art, a l l contribute to the science of human-computer interaction (HCI). With an engineering approach to design, scientific principles, technical information and imagination can be used in the definition of a mechanical structure, machine or system to perform prespecified functions with maximum economy and efficiency. This thesis covers methods for the study and modelling of a human-computer system, focusing on many relevant aspects: Development of organisations, work activities, information technology and humans and their competence. The importance of domain knowledge in an iterative user centred development process is emphasised. Analysis o f information utilisation is presented as a method for capturing how information entities are being used in the work process, to support directly design controlling knowledge. A workspace and document-oriented approach to interface design for skilled professionals is described to prevent the problems occurring with design o f graphical user interfaces. Creativity supporting design methodologies are described as means for being able to develop efficient and effective user interfaces, based on the results of the analysis of information utilisation. The concept of domain specific design for the establishment of a higher-level style guide for domain knowledge acquisition is introduced as a design controlling methodological support in the user interface development process. Case handling work activities have been particularly studied, applying the domain specific approach on the health care and tax handling domains. A general framework for case handling work activities has been specified, which supports work related judgement and decision making. The overall goal is to make computer support in working life easy, efficient and pleasant to use. Through this development, computer support can be enhanced by making it more cost efficient and less time consuming. Developers and designers can produce a result that fulfils the users’ objectives and requirements in accordance with the achievements of specific tasks.
1
2
Designing for Usability – Domain Specific Human-Computer Interfaces in Working Life. JAN GULLIKSEN ”Concern about the human being and his/ her destiny must always be the main interest in every technical development effort ... never forget that, amongst your diagrams and equations.” Albert Einstein in ”Science and Values” ”The natural scientist and the humanist need to be Siamese twins.” Tage Danielsson in ”Den akterseglade humanisten”
INTRODUCTION Computerisation and the use of computer based information systems have increased tremendously during the last three decades. Today, almost every existing work situation more or less relies upon the computer support. The efficiency of the work, the quality of the produced outcome, and the satisfaction of the user performing the work is directly related to, amongst other things, the computer support. In administrative, routine work, such as case handling, which is the main focus of these studies, the purpose of the work is never to operate the computer. The computer is merely a tool that has to be used as an aid to help achieve the main purpose of the work, i.e. to perform case handling; to be appreciated. Far too often in real life you come across situations as a customer where you are approached in a non-service manner, resulting in anger and frustration. And almost as often this lack of service is excused by a non-functioning computer support. Regardless of whether this problem is to be related to a lack in competence or knowledge by the clerk performing the case handling, or if it can be related to poor effectiveness or functionality by the computer support, it is a substantial problem that needs to be solved. A careful and dedicated design of the computer support and the work environment should help the clerk performing the work to achieve the goals and, thereby, also increasing the customer satisfaction. The results and methods presented in this thesis w i l l hopefully constitute a step in the complex process of achieving more usable and useful computer support in working life in the future. 3
GOALS AND PURPOSES This thesis aims at identifying factors and providing additional methodological tools for more efficient and effective analysis, design and construction of human-computer interaction in working life. The idea is that by creating better user interfaces to information systems, the efficiency of work will increase, the work environment will be more enjoyable and pleasurable to interact with, and, at the same time, the development times and costs will decrease. This research work has been applied to various work activities, such as tax case handling, health care, and bank and insurance work, to promote its general applicability. Central theoretical aspects that have been stressed are: the need for structured methodological aids for design; the analysis of how information is used in a specific work situation; and the domain studies and the case handling domain specific design guidelines. The concept of domain specific design, using domain specific style guides, is also introduced to create a higher-level platform for user interface design within a specific application area. Case handling work activities in general have been studied as examples of a broad range of administrative routine work activities performed by skilled professionals.
SYSTEMS ANALYSIS This work is a doctoral dissertation in systems analysis. Systems analysis is the study o f complex systems by modelling relevant aspects of the reality and performing experiments on those models in order to predict the effects on the real system [Gustafsson, Lanshammar, & Sandblad, 1982]. A model is an abstraction of relevant aspects of a system fulfilling some special purpose. One of the reasons for studying the reality with the help of systems analysis is the problems, or even impossibilities, of predicting the best, or even a good, behaviour from systems that have not yet been constructed. In this research, the human being, interacting with the computer in a specific physical and social work environment, to achieve a certain task, is the system under study. We study this physical system by analysing it according to several different methodologies and by constructing models to describe some aspects of it. The models help to specify methods and develop tools to enhance design, construction and maintenance of computer systems. Also, aspects relevant for the improvement of other parts of the system than the computer can be traced in this way. Important aspects in this research work, incorporated from the theories of systems analysis, are the importance of restricting the system domain and to be aware of the effects these restrictions can have. Further, the notion o f the subjectivity in modelling is important in order to explain diverging behaviour. A significant method in systems analysis is the method of simulation, i.e. the analysis o f model behaviour by model experiments.
INTERDISCIPLINARY STUDIES One basis for studying the complex system of a human achieving a task in a computerbased work environment is the need for an interdisciplinary approach. The following disciplines have substantial contributions: 1) Software Engineering and Computer Science 2) Human Factors and Work Activity Analysis 4
3) Cognitive Psychology including Perception 4) Sociology, Linguistics and other Human Sciences 5) Design and the Science of Art The assembled knowledge about computers, the architecture, the software and the functionality can be entirely understood by a human being, in that it is a purely technical system created by humans. The other side of the human-computer system, the human being, is not as easy to understand. Still, today the assembled knowledge on the human cognitive and perceptual functionality is not enough for the consequences of a computerisation of a work task to be fully understood. The study of cognitive, perceptual and social psychology is, nevertheless, important and there are some findings that we regard as directly applicable in the process of designing human-computer interfaces. Many different models and theories for explaining and predicting human features exist. Our knowledge is too limited to be able to falsify these models. We are satisfied with an explanation of human behaviour and information processing that can help us fulfil the overall goal of improving the human-computer work situation. Cognitive Psychology is the study of how human beings take in, represent, treat and use information. Central in this process is the organisation and structure of the human memory. According to the Stage Theory [e.g., Gleitman, 1991] of human cognitive capacity, human memory can be divided into short-term memory (STM) and long-term memory (LTM). Together STM and LTM form the nucleus of the information-processing system. LTM has no known capacity limits and some research even indicates that there is no memory loss; LTM represents information on a permanent basis. Memory retrieval can be ”distorted” (e.g., because of interfering retrieval cues to memory units). Another limitation is, however, more important for HCI. The STM, on the other hand, has a limited capacity (7±2 units) but it is possible to group units into larger ”chunks” that occupy less memory. If something new is entered older ”memories” are suppressed to leave space for the new memory units. Information is lost from the STM after about 12-15 seconds, if the information is not further worked upon. To be able to enter information into the LTM, the information needs to be more deeply encoded to be successively transferred from STM to LTM. In human interaction with a computer the load on STM could be decreased by enhancing the possibilities for the computer user to create larger, work-related memory chunks and to relieve the user from needing the STM to deal with controlling and manoeuvring the computer. Lind [1991] showed that utilising the STM as a temporary storage for information sets presented on sequential computer screens demand large cognitive load by the user compared with simultaneous presentation. Cognitive load caused by the information system (e.g., by overloading the STM) can be a severe factor when judging the comfort and efficiency of a work situation [Nygren, Johnson, Lind & Sandblad, 1992]. It can cause negative stress, anxiety, frustration, and even health problems [Johnson & Johansson, 1991]. Perception is a subset of cognitive psychology and incorporates the interpretation o f sensations (that is, immediate and basic experiences generated by isolated simple stimuli), giving them meaning and organisation [Matlin & Foley, 1992]. By using existing knowledge on the human perceptual capabilities, considerations relevant for the visual perception and recognition of information patterns can be derived. For example Gestalt Psychology states that groups of stimuli acquire a pattern quality which is more than the sum of their parts [Gordon, 1989]. In relation to human-computer interaction, this means that the design of user interfaces should aim at creating meaningful information patterns. Gestalt Psychology can be used to help us understand the effects o f grouping information that belong together, using consistent coding of colours, fonts and sizes, and separating figure background of different levels of user interface objects. 5
A central tool to aid the understanding of user interface design and for deriving prerequisites for the design process is the construction and use of models of human-computer systems [Rasmussen, 1986]. These models aim at understanding why people undertake certain performance patterns and how they can be improved in carrying out a specific work task with computer support. One model that has had a great impact on humancomputer interaction is the skill-rule-knowledge framework corresponding to different levels of familiarity with a task or environment [Rasmussen, 1983]. At a skill-based level of human performance, individuals behave according to stored patterns of preprogrammed actions. The rule-based level applies to the tackling of familiar problems by the application of stored conditional rules. In novel situations, the knowledge based level is used for on-line planning, using conscious analytic processes and stored knowledge to acquire expertise. Rasmussen further elaborates this within research on human error. He identifies eight stages of decision making (activation, observation, identification, interpretation, evaluation, goal selection, procedure selection and activation) with possible short-cuts, explaining stereotypical reactions and associative leaps to overcome the laborious and slow knowledge-based processing. The generic error modelling system (GEMS) further elaborates this by introducing the dynamics of human error and tendencies to treat problems on inappropriate levels [Reason, 1990]. A different view on human information processing divides the problem solving capabilities based on (at least two) levels of awareness. On a high conscious level, the human being is single processing; that is, only one such process can be performed at a time. For example, solving mathematical problems, reading, making judgements or decisions are processes that can not occur simultaneously. On a lower level of consciousness, several different things (e.g., driving a car, handling a standard telephone, judging the characteristics of information media (Figure 1)) can be performed in parallel without the involvement of higher-level cognitive processing. Successive repetition of a highlevel task can decrease the conscious level at which these tasks are performed, which we refer to as automatising a task [Schneider & Schiffrin, 1977; Schiffrin & Dumais, 1981]. Level of cognitive awareness
A high level of awareness, single processing
Automatisation is the successive transformation of a task to performance on a lower level of awareness.
A low level of awareness, with high capacity for parallel processing
Capacity Figure 1. There are different levels of cognitive awareness. High cognitive awareness means single-processing, but successive routine repetition can distribute a task to a lower level o f cognitive awareness through automatisation. People’s ability to automatise parts of their behaviour are closely related to cognitive processing and the efficiency in the performance of a specific task. But this knowledge 6
on which parts of an individual’s behaviour is automatic can not easily be observed and communicated to other individuals because of its complex environment and lacking awareness by the individual performing the operation. For example, skilled professionals tend to perform very complex work tasks rapidly, and without any conscious involvement, yet them not being aware of doing anything at all. This is closely related to the concept of tacit knowledge [Polanyi, 1958], referring to aspects on unaware knowledge that can not be communicated. To solve these problems, there is a need for new methodologies that capture design relevant aspects of automatic components. The importance these theories have for the design of human-computer interfaces i s the notion that high level cognitive processing should be funnelled to the work itself; i.e. to demanding problem solving or decision making processes. Cognitive processing should not be consumed on the control of the computer interface or the handling o f other media that should be possible to perform on an automatic level. The user interface should be made ”obvious” to the user [Nygren, Johnson, Lind, & Sandblad, 1992].
HUMAN-COMPUTER INTERACTION Human-Computer Interaction involves a set of processes, dialogues and actions employed by a user to interact with a computer to perform a specific task [Baecker, Buxton, Grudin, & Greenberg, 1994]. We need to extend the scope of this definition to cover a l l activities surrounding the analysis, design, construction, and evaluation of interactive computer systems for human use in a certain physical and psycho-social work environment. The importance of simultaneously facing the effects that information technology development has on the human being and his competence, on the organisation of the work and on the work activity as such have been known for a long time [Leavitt, 1958]. Bearing the historical development of HCI, according to Grudin [1990], in mind, future research implications aims at regarding everything that in any way might influence the computer supported work situation. Initially, a hardware focus described the user interface as something directly coupled to the computer, controlled by programming operators. Higher level programming languages freed users from directly handling the hardware. A keyboard and display focus included some early perceptual and motor issues. Conversational dialogues between the user and the computer initiated a focus upon cognitive issues. Finally, a groupware focus supports organisations in a social work environment. With the emerging development of information technology and the increasing commercial availability of graphical user interfaces (GUI), the theories underlying GUIs become more and more interesting. GUIs are based on the idea of direct manipulation [Schneiderman, 1988; Ziegler & Fähnrich, 1988], where objects in the task domain are represented by interface elements. User actions have a high degree of direct manipulation if it is illustrated with a visual representation engaging the user in a feeling o f immediate control. The degree of direct manipulation could be used as a measure of the learnability of the system but could sometimes come in conflict with the efficiency o f skilled users. The cognitive engineering philosophy that we find useful, as specified in Norman [1986], describes seven stages of goal driven user activity. When performing a task, a user specifies an intention based on some goal and translates this into an action that is to be executed. The user then perceives the system state and subsequently interprets and evaluates this in accordance with the goal. There is a possibility to identify short-cuts in this seven-stage cycle by providing direct manipulated user interfaces.
7
Different approaches to the science of HCI design exist; they all emphasise rather different views on how to analyse, develop, and maintain computer systems. We have found this grouping of the different approaches to user interface design appropriate both to illuminate the background sciences that have influenced the approaches and the methodological aids that have been developed [Wallace & Andersson, 1993]. The craft approach views each design project as unique, where software evolves under the guidance of a skilled human factors expert. Supporters of this approach tend to believe that a structured approach to interface design is an impossibility, as the aesthetics o f interface design can not be achieved through analytic techniques. It focuses on the designers need for talent, not for methodology. The artistic aspects of user interface design are undoubtedly important, but using this craft view as an argument against methodologies or methodological aids for user interface design, is clearly inappropriate. The enhanced software engineering approach attempts to introduce HCI techniques into the repertoire of traditional systems engineering by various methods for task analysis. HCI aspects become an issue for software engineers. Our observations are that the knowledge demands for efficiently being able to utilise HCI techniques, are too big to be put on software engineers. HCI-issues will inevitably receive a lower priority. The role of experience is also central, but software development projects tend to be too time consuming (e.g., years of development are not unusual), why one individual seldom gets the chance of acquiring experience from several projects in practice. To enable ”optimal” design, the cognitive engineering approach aims at applying theories from cognitive psychology to the problems facing the designer. Cognitive metrics models, such as the keystroke-level model [Card, Moran, & Newell, 1983] measures the user’s performance time and, thereby, indirectly estimates the memory load for unit tasks to help predict the efficiency of different design solutions. The grammar models, with formal grammatical notations, describe the mental models and their incorporation into the computer dialogue design. The knowledge methods try to make explicit the mental processes of the user when performing tasks. The user modelling methods describe not only what the user must know to perform a task but also how that knowledge is acquired and manipulated during the execution of a task. The problems of this approach lie in the failure of being applicable in real-life development projects due to its highly complex application of cognitive theories. The technologist approach tries to solve the problems of interface design by providing appropriate tools, especially the User Interface Management Systems (UIMS), a system that is both an interface development tool and an interface artefact, consisting of a special design environment, a linkage module and a management function. It is useful for prototyping and might even be usable for interpretation of formal specifications. It is, however, no real design support, because design begins long before the first prototype is constructed. It is merely a tool that allows bad interfaces to be developed more rapidly.
DOMAIN KNOWLEDGE AND USER CHARACTERISTICS We have defined a domain as a class of work activities that bears similar aspects on the workers situation and performance regarding interaction with customers, case handling, decision making, information processing, and so on. The importance of studying a domain and acquire domain knowledge for the production of information systems to support a work activity is rather obvious. Neisser [1976] argues that instead of spending too much time modelling in laboratories to come up with good models of human-computer interaction ”it is a good idea to go out and study in detail how people actually 8
work and especially the fine-grained details of the information used in the real world”. Recent research within cognitive ergonomics indicates the importance of extending the analysis of the interface beyond the visible user interface and instead regard the entire domain [Fischer, 1993]. The human being is not a passive computer user but an active problem solver in some world [Woods & Roth, 1988]. By incorporating domain knowledge earlier in the development process and in the methods for analysis, design and evaluation, and by manufacturing formal representations of this knowledge, user interface design for specific work activities can be enhanced. The information the users actually use in their work is important for the establishment of the domain model [Neisser, 1987]. By establishing methods for collecting this information and also extend these methodologies with aids to capture informal aspects of the work that the user might not be aware of , user interface design can be enhanced and the resulting user interfaces can be made more efficient and effective. There are also important ways to reuse software engineering components through domain modelling [Arango & Prieto-Diaz, 1991]. On the other hand there are evidence for the notion that knowledge of the application domain does not affect the quality of the design [Sharp, 1991]. In structured software development, software engineers are not encouraged to apply their creative abilities and restructuring the specification of the problem to something appropriate for design. This can not be seen as an argument against domain knowledge acquisition, but as an implication to the need of more structured approaches to the use of domain knowledge in user interface design, for example through the domain specific design methodologies elaborated in this thesis. The difference between novice users, expert users, and skilled users is important i n the domain analysis. Novice users, as opposed to expert users, are unfamiliar with computers; expert users have general computer experience, while skilled users are highly professional actors within their complex domain of work. They may be very skilled at performing their work but unfamiliar with computerised environments. For skilled users, the computer is only a tool that will be used and appreciated as long as it effectively supports the main purpose of the task. Interaction requirements for novices and skilled users vary greatly and there is often a conflict in these demands. Style guides and standards often overemphasise aspects relevant to novices. This is why we stress the dedicated design for skilled professionals when designing in complex administrative domains with high performance rates. Field studies show that properties of the work environment can limit people’s ability to effectively use their skills in the accomplishment of a work task. We usually refer to these hindrances as cognitive work environment problems [Sandblad, Lind, & Nygren, 1991], problems that are often associated with the user interface. Problems knowing the present location in the information space, problems interpreting visual codes with lacking consistency, problems with ”cognitive tunnel view” (i.e., only seeing a smaller subset of a document), loosing the overview, problems with STM overload, etc., are all cases of cognitive work environment problems. Addressing the reasons behind these problems is a necessity in the design of efficient human-computer interfaces.
TASK ANALYSIS An information system will be of no value if it does not contribute to the improvement of the work situation for people in the organisation. Because of this, it is not enough to study the contents of the information system, but also the activities people perform in an organisation and that could somehow be improved [Lundeberg, Goldkuhl, & Nilsson, 9
1981]. Task analysis (TA) is concerned with what people do to get things done [Preece, Rogers, Sharp, Benyon, Holland, & Carey, 1994]. TA is a general concept describing methods and techniques that either (a) elicit descriptions of what people do, represents those descriptions, predicts difficulties and evaluates systems against usability or functional requirements, or (b) predict performance, measure system complexity, measure learnability or the transfer of knowledge between systems. Hierarchical task analysis methods (HTA) decomposes high-level tasks into subtasks and operations or actions based on a graphical structure chart notation [Shepheard, 1989]. It involves an iterative process of identifying tasks, categorising them, breaking them down into subtasks and checking their accuracy. It is performed i n conversation with, and observation of, users working and results in an accurate description of the steps required to complete a task. Cognitive task analysis [Johnson, 1992] is, on the other hand, a group of methods concerned with informing the design process through the application of cognitive theories, capturing some representation of the knowledge users have to complete a task. One of the most well known is the model human processor (MHP). This model focuses upon the capturing of appearances of the users mental model of a computer system by predicting and modelling the user’s representation of the interaction (e.g., through the GOMS-modelling [Card et al., 1983]). The contextual inquiry method [Holtzblatt & Jones, 1993] supports the production o f general purpose systems and provides us with a way to work for short periods of time with users at multiple customer sites that are geographically dispersed. This provides an understanding of the nature of the user’s work through inquiry with users. One problem with existing methods of task analysis is that the definition of the tasks that are traced (e.g. key-pressings, or mouse-downs), views too fine-grained details of work. Bigger, concatenated tasks make the work context more important, where work can be viewed upon as simple tasks to be carried out as efficiently as possible, and where movement between tasks is important. This task switching [Henderson & Card, 1986; Card & Henderson, 1987] is seldom emphasised in task analysis methods. It is, however, important, if not crucial, for efficient human-computer interaction in a specific work setting. Important perspectives on the work activity can be reached by dividing the work into larger units. We refer to these as core work tasks, continuous in time, with a starting point and an end point and terminated by a decision. It has been noted in field studies that a minimal amount of ”key-pressings” and ”mouse-clicks” are performed among a much longer period of professional interaction with a work task that contains information search, judgement and decision making in a typical period of time. Computer supported administrative work of today is no longer about typing as fast as possible, rather to become a professional problem solver using the computer to support this task. The particular tasks that are performed in a work situation are important information, but for the design of human computer interfaces we need methods of analysis describing how tasks are performed. This is why we propose the method for analysis of information utilisation, to create a directly design controlling domain modelling basis for user interface design. Task analysis methods should focus on utility and usability to serve its purpose in practice
DESIGN Design is a creative activity that involves bringing into being something new and useful that has not existed previously [Jones, 1981]. Engineering design is the use of scientific principles, technical information and imagination in the definition of a mechanical structure, machine or system to perform pre-specified functions with maximum econo10
my and efficiency. Design refers to both the process of developing a product, artefact or system and to the various representations (simulations or models) of the product that are produced during the design process [Preece, Rogers, Sharp, Benyon, Holland, & Carey, 1994]. As a complement to these, we argue for a thorough understanding of the users and their tasks. This should result in a maximisation of the satisfaction and entertainment of using the resulting artefact. Despite these various definitions of design, the term i s often misused for the entire software development process, making software engineers into user interface designers. This is neither good nor appropriate in that only a smaller subset of the decisions that are made during the software development process are directly related to the design. We have defined design in human-computer interaction as the creation of a formal description (e.g., formal language, program code) of appearance and functionality, based on partly formal and partly informal (to a large extent) results of an analysis. One advantage to this approach is that it treats the creation of appearance and functionality o f the user interface as separate but related parts, and that different types of products can be designed with this definition, such as data base design, organisational design, design o f the work environment. The literature on design gives guidance on the essential prerequisites for interface design, but seldom supports the actual design process, even though they claim to do so. The usability engineering approach to design [Nielsen, 1993; Nielsen & Mack, 1994] refers to methods for design such as the various methods for usability evaluation and inspection. These methodologies are the only formal approaches to interface design that exists (and that should exist according to this approach). The actual design that is to be evaluated should never be developed based on a methodology. The design rationale approach [Moran & Carroll, 1995] serves as a fruitful step for listing the design questions, giving various options and criteria for choosing among these options for the best possible outcome. It is undoubtedly useful for documentation of the reasons for some design decisions. However, the tremendous workload that occurs due to the collection and retrieval of design background information easily becomes too cumbersome to be efficiently used in working life. User interface design is not just common sense, as shown in an investigation on skilled designers [Tullis, 1993]. Rather, the acquired skill by expertise in designing user interfaces is very important for a productive result of an interface design process. In inhouse development projects the possibilities for the design team to participate in many development projects to acquire skill is limited due to the time-consuming feature o f interface development. From our point of view there definitely is a need for a more structured approach to interface design, identifying different phases and supporting the making of design decisions with methodological aids and tools. Structured design methods and opportunistic processing are competing strategies in the nature of design expertise [Ball & Ormerod, 1995]. A design methodology should not, however, l i m i t the designer’s ability to bring novel approaches to the design process. Due to these restrictions, the best way to introduce and incorporate design knowledge efficaciously i n the development process is through the introduction of a special design role in the development. This designer should be well equipped with knowledge of the user; cognitive psychology and user characteristics; the development tool and software engineering i n general; domain knowledge, analysis methods and human factors issues; and physical and psycho-social work environment problems. The designer could function as a communication link between different participants in the development team, bringing in aspects that otherwise might have been neglected. These aspects include usability aspects, artistic design aspects and novel approaches to the performance of the work task with the computer support system. 11
Are any design methods really used in working life today? The following answers have been found within software development companies1: No, we do not use any special design method, but... • • • • •
we make it as simple as possible for the user. we follow ”Windows”. we have employed a specialist. we discuss the new functionality with the clients and then make a few forms. if you find any good method, please tell us!
As one easily realises there is a widespread disbelief in the work of finding design methodologies. There are two reasons for this. Either because the design methodologies are processes to supply prerequisites to the design and not directly to support the making of the design decisions. Otherwise the lack of methodologies is made purposely, because of the artistic aspects of design argument for talent and experience rather than method. Standards, style guides and guidelines have been manufactured to guide the designer while making of these design decisions. Standards are ”global” documents to supply definitions and rules to follow , defining basic functional requirements and quality measures for a user to be able to achieve effectiveness and efficiency in the performing of a certain work task [ISO, 1995]. A style guide (e.g. Motif) provides a consistent look and feel for products that adopt a specific style guide defining elements and their layout and rules for their use. Guidelines are specific to a certain application area and should support help and guidance in the specific design situation. Design guidelines, such as striving for user interface consistency, have been shown in various situations, to pull the attention away from the proper focus (which includes the user, the work task and the understanding of the work domain) of the user interface design [Grudin, 1989]. We argue that standards today are good and necessary but difficult to grasp and use due to their size and complexity. Style guides deal with physical elements and define more exact rules for their use, but they do not support any direct design support or support the actual development in a specific context. Guidelines are quite useful in specific application domains, but their quality and origin might be questionable. In our work, we extend the concept of style guides to be more domain specific and incorporate domain knowledge. We support our guidelines (heuristics) with empirical evidence [e.g. Nygren, 1996] Metaphors are important clues to the understanding of the user interface functionality and organisation. Today most graphical user interfaces are based on the desktop metaphor [Johnson, Roberts, Verplank, Smith, Irby, Beard, & Mackey, 1989]. The basis for the desktop metaphor is the user’s own ability to arrange the desktop according to his/her intentions and adapt work patterns according to the arrangement of the work space. In real working life, we have seen a number of examples of where the rearrangement of the user interface takes a lot of time away from the actual performance of the task, and where a great amount of mental effort is needed for this procedure. In a specific work situation where the user develops skills to handle work tasks with efficiency, accuracy, and with speed, the desktop metaphor becomes unsuitable. We propose an adaptation of a ”rooms” metaphor [Card & Henderson, 1987; Henderson & Card, 1986] by identifying workspaces to map certain features, known as work situations. The basis for this is the idea that all information needed for the performance of specific tasks is always visible on the interface in a workspace consisting of dynamic documents and indexes. 1
Eva Olsson at CMD is acknowledged for having performed this inquiry.
12
PARTICIPATORY DESIGN AND USER INVOLVEMENT Traditional researchers state that results that have not been obtained in the laboratory are of no scientific value and therefore useless in practice. With the speed that the science of HCI develops it is both unwise and impossible to subject every observation to laboratory research. To develop a good model demands too much time in the modelling laboratories. It would be much wiser to go out and in detail study how people actually work and especially the fine-grained details of the information used in the real world [Neisser, 1976]. The majority of the results outlined in this thesis includes observations from real development projects, so-called action research. This means that the results might not have general applicability. A common opinion is that when constructing computer support for the user, it should be rather obvious that the user him/herself i s the person best suited to design the user interface. Many models and methods for User Centred System Design exist, emphasising the study of the user in the production o f human-computer interfaces [Norman & Draper, 1986]. Unfortunately, too little evidence can be found on the practical use of a user centred design methodology. Today’s methods for system development are old and emanate from the times when interactive end-user applications where popular and important. The development of these seldom focus on the user and his/her task. This knowledge led to the development of (what i s known as the Scandinavian approach) participatory design [Schuler & Namioka, 1993; Greenbaum & Kyng, 1991], with the general assumption that the users should be not only involved, but in control of the analysis and design process. Cooperative design [Bjerknes, Ehn, & Kyng, 1987; Ehn, 1988] emphasises cooperation with the end-user’s union organisations to amplify the legal and democratic right of the users to influence and control the work situation of the future; the user is in direct control of the system development process. The iterative process of analysis, design, testing and redesign in a ”spiral” model has been emphasised, in Boehm [1988], for example. Organisational aspects can, in some cases, work against user participation. Users are of an undisputed importance in the user interface analysis, design and evaluation phases but, according to our observations, users are not always aware of their work behaviour, nor are they experts in design and human-computer interaction. This i s why a user controlled design methodology can have negative effects on the usability o f the resulting information systems. There is apparently a need to bridge the gap between developers and users. One way to do so is to introduce external consultants or independent developers as ”bridges”. One can distinguish among three different types of development [Grudin, 1991a], and they all have different varieties of the involvement of users and developers. •
Contract development focuses upon software development methodologies; for example, the waterfall model is a structured model of information system development involving; establishment of a development plan, requirement specifications, preliminary design, final design, coding, testing, integration, installation, running and maintenance. It is restricted to a ”build it twice design”; a prototype is constructed and tested in the design phase and the final system is then constructed based on these evaluation results. The reliance on written documentation as aids for communication between developers and users tend to build walls between them, aggravating iterative user centred design.
•
Product development focuses upon the user interface with an apparent need to manufacture attractive entertaining systems because it is the market; the client that separat-
13
es the users from the development. The client is in many cases not the typical user, rather a manager or information system specialist. •
In-house development focuses upon user participation. Because the earlier mentioned tendency of documentation to constitute a wall between developers and users, they need to be brought into physical contact. An adaptation of the organisation is often needed. This is the type of development that our research group at CMD mainly deals with.
In a recent Swedish study, a survey was made, trying to rate the usability maturity o f Swedish working life [Katzeff & Svärd, 1995]. Particularly important factors were: Formal and organised usability competence, and; Integrated usability in the design and development methodologies. The survey showed that the development of methods for the development of usable systems received the highest rating. This is in line with our experience that enhancing the methodologies for participatory design in in-house development companies receives the greatest attention and addresses factors relevant for every participant in the development process. Achieving more usable systems is a goal for everyone, there are only different ways, with different results, to reach the goal.
EXPERIMENTAL SYSTEM DEVELOPMENT Computer systems that are to be used by human beings should be (a) Easy to learn, (b) Usable (in relation to the task that is to be accomplished), (c) Easy to use, and (d) Pleasant to use [Gould & Lewis, 1988]. To reach these goals, the designer needs an understanding of who the user is. The presumptive user should be working close to the design team i n the early developmental stages and physically use prototypes for their behaviour to be measurable. Iterative design is needed with a cycle of design, test, measurement and redesign etc., in order to create efficient and effective user interfaces. In working life development projects, these design principles are often highly underestimated for several reasons, such as the diversity in use and user characteristics are often underestimated, too much confidence is put on rationales and reasoning for establishing how tasks are to be performed; there is a common belief that guidelines are sufficient prerequisites for a ”good” design; users do not know what they need; there is a belief that ”good” design i s ”getting it right the first time” and that iterative design is merely ”fine tuning” of the interface, and risk that the development process might be delayed. In working life of today, most development methodologies are still adaptations o f the waterfall model of systems development, that is, successive stages of operational plans, operational specifications, coding specifications, coding, parameter testing, assembly testing, shakedown and system evaluation, with each producing rigorous documentation. In in-house development projects, it is very common that the large amounts of documentation, produced as results of the task analysis, create a wall between developers and users, and especially so if developing according to the waterfall model. It is our experience that the documentation produced is seldom used in the interface design phase, either because they are insufficient or because they are too large and extensive and hence impossible to overview. With an iterative spiral model of continuous analysis, design, evaluation and redesign, etc., development and maintenance of larger information systems can be enhanced [Boehm, 1988]. Such a model has shown to be problematic if it is directly applied to working life, because of the problems of keeping down the iterative life cycle time. Also, the problems related to time and interest by the developers to learn new development methodologies, and the complexity in the knowledge demands on the design 14
phase are some reasons why we propose the development of a certain design competence in the iterative experimental system development process, with essential knowledge prerequisites and methodological tools such as analysis of information utilisation and design methodologies.
EVALUATION METHODS Evaluation methods have become an increasingly important tool for user interface design and need be incorporated into the system development process. We can separate evaluation methods into usability testing methods (i.e. where users are involved) and usability inspection methods (i.e. where users are not involved).
Usability Testing Methods Performance measurement [Nielsen, 1993] collects data on errors and times to determine to what extent a usability goal is reached by having a group of test users perform a pre-defined set of tasks in a laboratory. This method allows a quantitative comparison of the usability of different design solutions. The method is expensive, complex, and time consuming to use. Moreover, the laboratory tests are difficult to perform early i n the design process because it requires a running prototype and a reasonably complete data base. More abstract goals can not be measured using this method. Questionnaires [Nielsen, 1993] are useful for judging users’ subjective satisfaction and anxieties. Questionnaires are inexpensive and may be distributed to many users. However, end users are not able to judge their satisfaction when they have acquired the necessary skills. This is why questionnaires does not identify usability problems concerning efficiency in daily use. Thinking aloud is a method in which a usability expert notes how users view the computer system by verbalising their thoughts while using the system. It is an inexpensive way to identify users’ misconceptions of a system. Unfortunately, it is not natural for users to think out loud while performing work, and there are many automatised information handling routines that the user might not be aware of in daily use. The pluralistic walk through [Nielsen & Mack, 1994] involves users, developers and usability experts, and may be carried out early in the design process. Representatives from the three categories meet and discuss usability problems associated with the dialogue elements in scenario steps. The main focus is on how users react in different situations. The method is effective in evaluating the learnability of a user interface, but is not appropriate for evaluating interfaces in daily use. Users are not able to predict how they will interact with a system when they are skilled.
Usability Inspection Methods Cognitive walk through [Nielsen & Mack, 1994] is a method through which an evaluator examines each action in a solution path to tell a credible story describing why the user would choose certain actions. It is based on assumptions about the users background, knowledge, goals, and an understanding of the problem solving process that enables a user to guess the correct action. The method focuses on evaluating a design for ease o f learning, particularly by exploration and is, therefore, not applicable for inspection o f interfaces for skilled users. 15
Heuristic evaluation [Nielsen & Mack, 1994]is performed by an evaluator using a set of guidelines (i.e. heuristics in form of a checklist) to judge the usability of the user interface. It is easy to learn and inexpensive to use. Many usability problems can be traced and it is usable early in the design process. Evaluators using this method, however, seldom manage to identify domain specific usability problems due to a lack of domain knowledge by the developers. Mostly general usability problems can be identified with heuristic evaluation. Heuristics suggested in Nielsen [1993] emphasise ease of learning and are not "optimised" for efficiency in daily use.
Domain-specific Evaluation Methods Apparently there is a need for evaluation methods that focus upon efficiency in daily use and ease of learning. We also need methods that identify both general and domain specific usability problems. To be usable, the methods also need to be efficient, easy to use and possible to adopt early in the design process. A method to fulfil some of these goals is the method of domain specific evaluation that identifies general usability problems by an adapted heuristic evaluation method, and domain specific problems through an evaluation together with domain experts to be used early in a user-centred design process [Lif & Sandblad, 1996]. The ADA-method [Sandblad, Åborg, & Lif, 1996] is another method for evaluating user interfaces for skilled professionals when they have become skilled at using the new system. It is performed by a usability expert categorising usability problems in accordance with a specific task through observation-interviews.
DESIGNING FOR USABILITY – DOMAIN-SPECIFIC DESIGN OF USER INTERFACES In this extensive introduction to the problem area, the need for domain specific approaches to modelling, analysis and design of human-computer interfaces in working life. Below, I will describe what our contributions are to the problems described above by briefly summarising the papers in the thesis and their interrelationships. This thesis has been divided into three parts: Part I: Methods and tools for HCI Development; Part II: Domain-Specific Design and Domain-Specific Style Guides; and Part III: The Case Handling Domain; Field Studies, Models and Applications. The purpose of this division has been to separate and group what I regard as theoretical reasoning and ideas, the domain specific approach and strictly applied research work.
I.
Methods and Tools for HCI Development
This part of the thesis explains theoretical ideas and foundations, upon which we have based the defined methodologies and methodological aids. It establishes a new definition of the concept of design within human-computer interaction and stresses the need for a more structured design methodology. It specifies the method of analysis of information utilisation and gives design heuristics and metaphors for design dedicated to a specific class of work. The results have been generalised and examples from several domains of work are given.
16
1.
Efficient Development of Organisations and Information Technology – A Design Approach.
In the first paper, a framework for the entire process of organisation and information system development is defined as a structure for the entire process of organisational and information technology development, consisting of separate but sequentially related parts (Figure 2). User User Organisational development
4 Organisational model
Information analysis
Application development 3 Application model User
Information utilisation analysis
2
1. Database 2. Applications
User
Work model
User Interface Design
1 Interface model
Construction
Activities – work processes (analysis, design, construction) Descriptions – resulting formalised models Figure 2. Iterative experimental system development; including consecutive activities and the descriptions they produce. Feedback and possibilities for participatory design are also illustrated. The different development phases (activities) and the resulting models (descriptions) are defined focusing especially on design issues. As mentioned before, the meaning o f the concept design varies depending on the discipline and even on which researcher that answers the question. We have defined the process of design in human-computer interaction as the process of creating a formal description (e.g., program code, prototype, etc.) of the appearance and functionality of an information system. This design is based on both formal and informal descriptions of interaction requirements as a result of a work analysis process. Informal aspects that are encountered in the analysis process (i.e. work routines that the user might perform without him/her being aware of it, or properties observed by the designer without it being possible to formally describe) are very important for the creation of efficient user interfaces. The analysis phase is separated from the design phase; while documenting the results of the analysis, no design decisions should be made. It is a point to try to produce work descriptions in the language of the work activity, without involving design terminology. With such a definition we can speak about organisation design, database design, interface design and program code design (construction). According to the definition of design as a specification in a formal language, it can never completely describe all requirements. In the design process, four different consecutive gaps of communication can be distinguished, in which important information about the actual work situation could be lost. These gaps refer to the formalisation o f the process of organisational development, data modelling, work descriptions and interface design specifications. The gaps can constitute severe obstacles in the process o f developing efficient and usable information systems for specific work situations. By introducing development models that covers the entire process of design, these gaps might bridge, or at least narrow. Some main characteristics of such models are presented in this paper.
17
2.
The Nature of User Interface Design – The Role of Domain Knowledge.
The importance and growing awareness of domain knowledge acquisition in information system development, and especially in the process of design of human-computer interfaces, is becoming more evident. But the introduction of participatory design techniques and user centred system design are, although necessary, very difficult to efficiently adopt in working life. The basis for the discussion is the focus on skilled professional users, using computers to support the performance of a complex work task. In several in-house development projects, methods for efficiently capturing and utilising domain knowledge have been defined and tested (e.g., through the definition of domain-specific style guides and methods surrounding these). Also, methods for modelling case handling work in general terms, which is the main application focus of our studies, have been developed and tested in larger corporations. The second paper aims at relating domain knowledge models to conceptual models in information system development and to user models in human-computer interaction. Based on Norman's [1986] model of user perception of an existing computer system, different limitations are identified and possible extensions are discussed. By introducing the domain context, several additional models of the work task by the user, as well as the designer, occur. If the dynamic iterative system development process is regarded, further models and dynamic changes in the models over time can be traced. The different mental models by the participants in a development project are identified and their changes over time because of the iterative experimental system development process, are discussed. Furthermore, these models are related to the state-of-the-art knowledge on mental models and domain modelling. Implications for design, such as work modelling, iterative system development, analysis of information utilisation, design and evaluation methods, are discussed.
3.
Analysis of Information Utilisation (AIU)
As outlined previously, task analysis methods are necessary but insufficient for deriving specific knowledge that is directly applicable for user interface design. The third paper describes analysis of information utilisation (AIU), a method for the specification of how information entities encountered in the information analysis are being physically manipulated in a work situation. AIU focuses particularly on factors affecting cognitive load. The method supports the human-computer interface designer with relevant and appropriate information structured specifically for a document oriented design methodology, what we call design controlling information. It gives a suitable basis for user centred interface design, without the possible limitations that improper user participation could cause. It focuses on all aspects relevant for a ”good” work environment (e.g., human competence, work organisation, work activity and information technology development). Analysis of information utilisation is performed through ”observation-interviews” in which a human-computer expert interviews a typical user about his work situation i n the domain of work and observes information handling routines that the user might not be aware of. Every analyst adopt individual patterns of how to conduct such an interview, why we do not support special instructions on this. We rather present a framework for the documentation of these analysis results for the incorporation in the design process. A number of development projects has shown that this method can constitute an additional step in existing methodologies for capturing aspects of the work situation that can be 18
directly design controlling. The paper describes the method of analysis of information utilisation, and what information that has to be documented while performing the observation-methods. It suggests how the method can be incorporated into existing inhouse development methodologies, and gives an example of documented analysis results from a development project within the Swedish National Tax Board.
4. Work Spaces and Rooms in User Interface Design. Design of graphical user interfaces are today mostly based on the desktop metaphor and window-based analysis and development methodologies. These methods particularly focus upon the spatial design of windows and navigation by means of menus and buttons. Window-based computer systems have totally come to dominate the commercial market. Originally, this metaphor was not developed for in-house development of dedicated computer support systems for professional users. Even if it means stepping aside from existing standards and style guides, other basic structuring methods should be considered when today’s methods aggravate efficient performance in a work situation. The purpose of the fourth paper is to identify differences between two approaches regarding the basic structuring of user interfaces to computer systems for skilled professionals in working life. The first approach to the design of an information system i s based on a detailed analysis of the total number of work processes of the organisation under study. A basic design problem is to show all available functions of the system to the users, and supply them with efficient facilities for navigation from the overview state to the desired state where a specific task can be performed. The second approach i s based on an analysis where tasks, relevant to each individual in the organisation, are identified. A design goal is to specify an efficient interface for each group of individuals (or work role). The proposed design has been inspired by many concepts from the ”rooms design” metaphor [Card & Henderson, 1986]. By separating different room structures for groups of individuals and base the design of screen documents on domain knowledge, tailored and efficient interface structures can be obtained.
5.
A Structured Approach to Prototype Design in Iterative System Development.
As mentioned before, existing methodologies for design focus upon the capturing of design rationales, methods for domain analysis, or evaluation methods to name a few. There is little or no support for making design decisions; rather, design is regarded as an artistic process that can be improved only through experience. This fifth paper presents a structural approach to the iterative process of designing effective and efficient human-computer interfaces. Different phases of the design process are identified. Four different phases of the design process are described: Identification of the information contents, specification of the layout, dialogue structure, and the behaviour of the user interface. Design decisions made in these phases are based on a set of domain specific heuristics. Prerequisites for the user interface design process is an analysis of the domain, the users, and the technical environment. Based on this, a suitable metaphor can be chosen. A “workspace metaphor“ supporting administrative work in a broad sense performed by skilled professionals is introduced. By creating interface elements and workspaces in an iterative design process, a first prototype can be defined. The following user characteristics are important. (a) Cognitive processing on a conscious level is a limited resource. (b) Transferring information via keyboard and/or mouse i s 19
a slow process as compared to the swiftness of cognitive and perceptual processing. (c) Experienced users decode frequently occurring, meaningful patterns quickly. These three constitute a basis for two general design goals: (1) Design an interface that minimise the users cognitive processing on a conscious level. (2) Create a design that w i l l minimises the time the user has to deal with the keyboard and/or the mouse. These goals can be fulfilled if the designer knows the frequency by which the information entities are needed and the technical limitations of the computer system and its software. Resolving conflicts between the two goals are a major part of the designers work. Some design examples emphasise the resulting design, basing it on those heuristics.
II
Domain-specific Design and Domain-specific Style Guides
The importance of incorporating domain knowledge early in the development process was stressed in the previous section. Of course, there are many ways of doing this, but not so many that efficiently supports the design process. By introducing a domain specific style guide, i.e. a higher level of interface elements with element specific guidelines incorporating domain knowledge, development can be enhanced. The development can be performed quicker and cheaper, with better possibilities to achieve efficient user participation and to produce more usable interfaces. This section introduces the concept o f domain specific design and relates earlier methodological tools to this technique.
6.
The Need for New Application-specific Interface Elements.
The design of user interfaces for skilled workers in professional work settings should be based on style guides that certify efficiency. Most of toady’s style guides and design guidelines over-emphasise general aspects or aspects relevant to novices. To increase efficiency both of the design process and of the resulting interface, more domain specific interface elements should be used. This sixth paper explains the basic ideas of such domain specific style-guides and gives some examples from the health care domain. Interface elements that are described are: The document, consisting of one or more pages in a bundle, improving the possibility of recognition of spatial location; the index, containing information on the contents and structure with graphical features; miniatures (i.e. a proportional diminution of a screen object) that can be recognised through the information patterns it create; and more complex element combinations. The point i s to incorporate domain knowledge already in the element structure and present these in a domain-specific style guide.
7.
Domain-specific Design of User Interfaces.
The use of graphical user interfaces in a computerised work environment is often considered to substantially improve the work situation. The outcome can, however, often be the opposite. Inappropriate use of windowing techniques, scrolling and colours can result in tedious and confusing interaction with the computer. Today's standards and style guides define basic design principles but are insufficient for design of interfaces to end user applications. Here detailed domain knowledge is indeed essential. This seventh paper describes the domain-specific style guide (DSSG) as an extension of today's standard with domain-specific primitives, interface elements, and forms, together with domain-specific guidelines. Careful dedicated analysis of information 20
utilisation in a domain is the development basis for a DSSG. The development is performed with an object-oriented approach to facilitate the reuse of interface components and to support consistency and structure. By using a DSSG, the development of applications can be performed with a simplified information analysis. Therefore, a more effective design process is possible, one in which end users can participate in the design using their own familiar domain-related terminology. Time and costs for the development process can be drastically reduced if domain-specific style guides, design guidelines, and development tools are used.
8.
Domain-specific Design of User Interfaces – Case Handling and Data Entry Problems.
New approaches on human-computer issues are required in designing user interfaces for skilled professionals in specific domains. By including domain knowledge earlier i n the development process, applications development can be simplified. The advantages of a domain specific design are obvious: (1) It provides increased efficiency in the development process and in the final application; (2) It enhances the possibilities for effective user participation; and (3) It reduces development time and cost. Domain specific design requires a careful analysis of information utilisation. The analysis results in a design specified in a special design language interpretable by the construction tool. This paper outlines a framework for a domain specific design and development process. Design was earlier supported mainly by cognitive theories on human computer interaction and user modelling. In the domain specific approach, design issues are concentrated on efficient development of effective human-computer interfaces. Our purpose is to identify the major steps of the process – analysis of information utilisation, design and construction – with special emphasis on their documentation. The different documentation can be seen as interfaces or bridges between preceding and succeeding steps in the development process, and are to be specified in some kind of formal or semiformal language. We are currently working on the specification of these languages. Our development process can be compared with development using traditional UIMS, with emphasis on the possible benefits, such as cost and time savings, increased possibilities for effective user participation and increased modularity. The domain of case handling accentuates the domain specific development as a general approach on a large number of administrative work situations where case administrators interact with customers or clients. Case handling is an important domain with many end users using their skills for complex information retrieval and decision making. This domain is described from the case administrators perspective on customer interaction, information retrieval, decision support and information systems. We stress some general aspects on case handling, which are important for the design of efficient computer support for personnel working in this domain. Some basic work flow entities, to describe and categorise the case handling domain, are presented. These include the initialisation phase, the information retrieval process, the case administrators decision process, and the documentation and information services.
III. The Case Handling Domain; Field Studies, Models and Applications Most of our research has been applied to case handling work (e.g., tax administration work, bank and insurance, personal administration). Case handling is a general concept 21
describing work that bears similar aspects concerning client approaches, information structuring, judgement, decision making and archiving. By attacking these problems with a mutual approach, synergetic effects concerning usability and human-computer interaction can be reached. This section aims at describing current research within the Swedish National Tax Board, applying theories described in the previous sections.
9.
Case Handling Models as a Basis for Information System Development.
Computer use in administrative work is increasing rapidly and today most of the case handling activities depend heavily on computer support. As a consequence, usability problems are increasing due to the complexity of the information systems, which consist o f many different software technologies that need to be combined together in an appropriate way. Software is often developed in in-house development projects, and sometimes even in combination with adaptations of various off-the-shelf products. The load on the software development team becomes enormous, and their possibilities for developing standards and reusing software components are limited. A general approach to case handling work through mutual concepts and process flows is a basis for being able to develop a domain specific style guide (DSSG) for the case handling domain. In this paper, we have modelled the case handling domain based on work activities within the Swedish tax administration. It has been performed in a user centred way and these models have been tested and validated in a number of other work activity areas within the organisation. We suggest how case handling models can be used in an effective way when analysing and designing information systems for case handling. The important aspects that stem from the synchronisation of this technology development with organisational, competence and work activity development being emphasised. Such a framework should be a necessary prerequisite for domain-specific design of case handling work.
10. Domain Specific Design in Practice: Tax Handling Work. This paper describes the state of the art of an ongoing development project within the Swedish National Tax Board. The project aims at introducing the technique of domain specific design, increase the developers competence in the area, and participate i n the design and the development of a domain specific style guide for the Swedish National Tax Board (RSVSG). Important aspects of domain specific design that this project studies include; the introduction of new methodological steps in the existing software development life cycle; the specification of workspaces and the methodological aspects of identifying major work situations; the structure of the domain specific style guide; and, finally, we will give some implications for the future work within the project.
11. Designing Graphical User Interfaces under the Limitations of Old Alphanumeric Mainframe Systems. A common problem in in-house development projects is the tendency to be so dependent upon existing information systems that the only possible development solution is to extend the computer environment with new client-server applications, and thus communicating with the older alphanumeric mainframe systems. Maintaining and refining existing software receives higher priority than building new computer 22
environments. In our work, we have observed and identified these problems in several case handling work activities and studied the impacts these problems have on end users. A better work environment, minimising the cognitive load caused by the handling of the information system, can be achieved by approaching these design problems through the application of design heuristics specific for the case handling domain, and by a w e l l thought out compromise due to these technical limitations.
GENERAL DISCUSSION Because of the increasing psycho-social, physical and cognitive problems in computerised work environments of today, there is a strong need to improve the work situations for the thousands of computer users that today suffer from badly designed computer support systems. Evidence can be found that some of the work environment problems found in real life work situations are related to the efficiency and usability of the computer support. Numerous field studies have shown decreases in work efficiency and satisfaction due to the introduction of computer based information systems into w e l l functioning paper-based work situations. There is definitely a need for efforts aiming at the improvement of the human-computer-work interaction situation. For example, a redesign of displays in an American telephone company decreased the average search time by an amount translatable to 79 person-year savings per year [Tullis, 1981]. S t i l l today, a lot of new development of character-based user interfaces occur in working life, despite nearly 15 years of commercially available graphical user interfaces. Generally, graphical user interfaces are more effective than character-based interfaces, but for tasks where the main part of the interaction is input, old-fashioned interfaces are often better [Rauterberg, 1992]. Although having a nicer look-&-feel, the graphical user interfaces are sometimes less efficient for skilled professionals. This thesis has described a structured approach to organisation and information technology development, focusing especially on analysis of information utilisation and user interface design. The need for structured methods incorporating work activity, organisation, human competence, and information technology development have been stressed. Embryos to the method of analysis of information utilisation and design methodologies have been presented. Advantages have been noted, we have been involved in, using the analysis of information utilisation in our development projects. Design methodologies in the form of design heuristics for skilled professionals have been defined, related to empirical measurement and practically used in development projects. Concerning organisational development no methodological aids have been presented, rather, the need for adapting the work organisation to information technology changes have been discussed through the introduction of ”learning organisations”, i.e. organisations that automatically adapts to required changes. The information analysis has not been given much space due to the acceptable quality of existing methods, but object-oriented design methodologies [Booch, 1990; Rumbaugh, Blaha, Premerlani, Eddy, & Lorensen, 1991] are essential prerequisites for achieving modularity and promoting software reuse in the development process. The present thesis briefly mentions construction and implementation aspects, but without supporting any tools or any general advice. These are aspects that we need to further study in detail. Our research group is currently working on a tool that automatically generates a functioning user interface, by interpreting a design description file, together with an existing database and ”methods” describing the behaviour. Domain-specific design aims at creating a higher-level platform for system developers, designers and for work activity representatives. It is a well known fact that gene23
ral style-guides and standards overemphasise aspects relevant for novices, or aspects relevant for learning to use the interface. With the domain-specific design methodology, development of user interfaces for skilled professionals in a specific work setting can be enhanced, and thus effectively supporting the accomplishment of individual performance of tasks. Ideally, development should be possible to perform in cooperative sessions with designers and users, structuring the interface from a set of pre-specified interface elements, containing domain knowledge. To what extent the domain specific style guide can be used in a specific application project and how much needs to be redesigned and incorporated into the style guide is yet to be tested and evaluated. We are currently adopting the technique in development projects for the Swedish National Tax Board, and the full experience will have to wait a few years to be fully documented. However, so far the ideas and applications, design methodologies and prototypes describing the technique and the resulting user interfaces have been received with overall positive judgements, both by developers and user representatives. The establishment of a tax case handling style guide should be based on a common conceptual framework describing and defining general concepts and processes within the organisation. We have participated in the user centred development of such a framework and evaluated its general applicability on other case handling work activities. Work on a refined establishment and use of the case handling framework have recently been restarted. So far, the results and responses of the establishment of a common conceptual framework and mutual process descriptions for general case handling work have been well received within the organisation. The establishment of methodologies for using such a framework, for basing user interface development on the framework, for identifying needs for organisational changes, competence increase, and business process reengineering are yet to be performed. Also, rules and regulations for updating and dynamically adapting the framework to new work activities and work routines are to be specified. Other official authorities have expressed an interest in establishing similar case handling frameworks. Future research will focus on describing and refining all steps in the organisation and information technology development, according to the structure described in paper one. The development and introduction of domain specific style guides and design methodologies in working life application projects will be evaluated. The case handling framework will be further elaborated and applied to specific work activities. Methods for the use of the case handling framework will be defined, as well as organisation and rules for updating the framework. It is important to incorporate and base the development of the domain specific style guide on this framework, identifying basic elements and information structures. These methodologies strive to enhance development by decreasing the iterative system development time, supporting high level interface elements and guidelines and by the incorporation of domain knowledge. Thereby, development can be performed quicker, better and with reduced development costs. By decreasing development time and restructuring the development process, developers and designers have the possibility o f acquiring design expertise, because of the possibility to participate in several development projects using the same tools. Despite the tradition of user centred and even user controlled design methodologies [Bjerknes et al., 1987; Ehn, 1988; Greenbaum & Kyng, 1991; Schuler & Namioka, 1993], many problems with user participation in the development process exist [Grudin, 1991b]. Cognitive, physical and psycho-social work environment problems must all be considered in order to obtain increased work efficiency, improved effectiveness, and a better work environment. Although user participation is difficult, because of the complexity in the work tasks for which computer support is being 24
developed, it is absolutely necessary for in-house development. We increase the possibility for more effective user participation in the development process and hope to increase the satisfaction and efficiency of computer use in daily, working life for skilled professionals.
REFERENCES BAECKER, R., GRUDIN, J., BUXTON, W., & GREENBERG, S. (1994). Readings in HumanComputer Interaction: Toward the year 2000, 2nd edition, Morgan Kaufman Publishers, Inc. BALL, L.J. & ORMEROD , T.C. (1995). Structured and Opportunistic Processing i n Design: A Critical Discussion. International Journal of Human-Computer Interaction, Vol. 43, pp. 131-151. BJERKNES, G., EHN, P. & KYNG, M. (1987). Computers and Democracy. Gower Publishing Company Ltd., England. BOEHM, B. (1988). The Spiral Model of Software Development and Enhancement. IEEE Computer, Vol. 21, No. 5, pp. 61-72. BOOCH , G. (1990). Object-oriented Design with Applications. Benjamin Cummings. CARD , S.K., MORAN, T.P. & NEWELL, A. (1983). The Psychology of Human-Computer Interaction. Lawrence Erlbaum Associates Ltd., Publishers, Hillsdale, New Jersey. CARD , S.K. & H ENDERSON, A. (1987) A Multiple Virtual-Workspace Interface to Support User Task Switching. In CHI'87 Conference on Human Factors in Computing Systems, (Toronto, Canada, Apr. 6-9), ACM/SIGCHI, New York. EHN, P. (1988). Work-Oriented Design of Computer Artifacts. Arbetslivscentrum, Sweden. FISCHER, G. (1993). Beyond Human-Computer Interaction: Designing Useful and Usable Computational Environments. In J. Alty, D. Diaper & S. Guest (eds.) People and Computers VIII, Proceedings of the BCSHCI ’93 conference, Cambridge University Press, Cambridge, UK G LEITMAN , H. (1991). Psychology. 3rd edition. W.W. Norton & Company, Inc., New York. GOULD, J.D., & LEWIS, C. (1985). Designing for Usability: Key Principles and What Designers Think. Communications of the ACM, Vol. 38, No. 3, pp. 300-311. GORDON, I.E. (1989). Theories of Visual Perception. John Wiley & Sons Ltd., Great Britain. GREENBAUM , J. & KYNG, M. (Eds.) (1991). Design at Work: Cooperative Design o f Computer Systems. Hillsdale, NJ: Lawrence Erlbaum Associates.
25
GRUDIN, J. (1989). The Case Against User Interface Consistency. Communications of the. ACM, Vol. 32, No. 10, pp. 1164-1173. GRUDIN, J. (1990). The Computer Reaches Out: The Historical Continuity o f Interface Design. In J.C. Chew & J. Whiteside (eds.) Proceedings of Human Factors in Computing Systems, CHI ’90, ACM/SIGCHI. GRUDIN, J. (1991a). Interactive Systems: Bridging the Gaps Between Developers and Users. IEEE Computer, Vol. 24, No. 4, pp. 59-69. GRUDIN, J. (1991b). Obstacles to User Involvement in Software Product Development, with Implications for CSCW. International Journal of Man-Machine Studies, Vol.- 34, pp. 435-452. GUSTAFSSON, L., LANSHAMMAR, H., & SANDBLAD, B. (1982). System och Modell – En Introduktion till Systemanalysen (System and Model – An introduction to Systems Analysis), in Swedish, Studentlitteratur, Lund, Sweden. H ENDERSON, A. & CARD , S.K. (1986). Rooms: The Use of Multiple Virtual Workspaces to Reduce Space Contention in a Window-Based Graphical User Interface. ACM Transactions on Graphics, Vol. 5, No. 3, pp. 211-243. HOLTZBLATT, K. & JONES, S. (1993). Contextual Inquiry: A Participatory Technique for System Design. In D. Schuler and A. Namioka (eds.) Participatory Design: Principles and Practices, Lawrence Erlbaum Associates, Publishers, Hillsdale, New Jersey. INTERNATIONAL O RGANISATION FOR STANDARDISATION (1995). ISO/DIS 9241. Ergonomic Requirements for Office Work with Visual Display Terminals (VDTs) Part 10: Dialogue Principles, Part 11: Guidance on Usability, and Part 13: User Guidance. JOHNSON, P. (1992). Human-Computer Interaction, Psychology, Task Analysis and Software Engineering. McGraw-Hill Book Company, London. JOHNSON, J., ROBERTS, T.L., VERPLANK , W., SMITH , D.C., IRBY, C.H., BEARD , M., & MACKEY , K. (1989). The Xerox Star: A Retrospective. IEEE Computer, Vol. 22, No. 9, pp. 11-29. JOHNSON, J.V. & JOHANSSON, G. (Eds.) (1991). The Psychosocial Work Environment: Work Organisation, Democratization, and Health. Baywood Publishing Company, Inc., Amityville, New York. JONES, J.C. (1981). Design Methods: Seeds of Human Futures, 2nd Edition, London: Wiley. KATZEFF, C. & SVÄRD , P.O. (1995). Användbarhet i Praktiken: En Enkätstudie. (In Swedish: Usability in Practice: A survey.) SISU publication. no 95:20. Swedish Institute for Systems Development. LIF, M., & SANDBLAD, B. (1996). Domain-specific Evaluation, during the Design o f Human-Computer Interfaces. In A.G. Sutcliffe, F. Van Assche, & D. Benyon (eds.) Domain Knowledge for Interactive System Design. Proceedings of the IFIP WG 8.1/13.2 26
Joint Working Conference on Domain Knowledge for Interactive System Design, Geneva Switzerland, 8-10 May, Chapman-Hall: London. L IND , M. (1991). Effects of Sequential and Simultaneous Presentation of Information. Report no. 19, CMD, Uppsala University. LUNDEBERG , M., GOLDKUHL , G., & N ILSSON , A. (1981). Information System Development: A Systematic Approach, Englewood Cliffs, New Jersey: Prentice Hall. MATLIN, M.W., & FOLEY , H.J (1992). Sensation and Perception. 3rd edition. Allyn and Bacon, Massachusetts. MORAN, T.P., & CARROLL, J. (1995). Design Rationale: Concepts, Techniques and Use, Lawrence Erlbaum Ass., NJ. NEISSER, U. (1976). Cognition and Reality. San Francisco: W H Freeman N IELSEN , J. (1993). Usability Engineering. Academic Press, Inc., San Diego. N IELSEN , J., & MACK, R.L. (1994). Usability Inspection Methods. John Wiley & Sons Inc. NORMAN, D.A. (1986). Cognitive Engineering. In D.A. Norman & S. Draper (eds.) User Centered System Design, Lawrence Erlbaum Associates, Inc., Hillsdale, New Jersey. N YGREN , E. (1996). From Paper to Computer Screen – Human Information Processing and User Interface Design. PhD Thesis, Systems and Control Group, Uppsala University. N YGREN , E., JOHNSON, M., L IND , M. & SANDBLAD, B. (1992). The Art of the Obvious. Automatically Processed Components of the Task of Reading Frequently Used Documents. Implications for Task Analysis and Interface Design. In J.P. Baursefeld, J. Bennett & G. Lynch (eds.) Proceedings of Human Factors in Computing Systems, CHI '92, Monterey, California, May 1992, ACM, pp. 235-239. PREECE , J., ROGERS, Y, SHARP, H., BENYON, D., HOLLAND, S., & CAREY, T. (1994). Human-Computer Interaction. Addison-Wesley Publishing Company, Wokingham, England. POLANYI, M. (1958). Personal Knowledge. University of Chicago Press, Chicago RASMUSSEN, J. (1983). Skills, Rules, and Knowledge; Signals, Signs, and Symbols, and Other Distinctions in Human Performance Models. IEEE Transactions on Systems, Man, and Cybernetics, Vol. 13, No. 3, pp. 257-266. RASMUSSEN, J. (1986). Information Processing and Human-Machine Interaction. A n Approach to Cognitive Engineering. A.P. Sage (ed.) North-Holland Series in System Science and Engineering. Elsevier Science Publishing Co., Inc., Amsterdam. RAUTERBERG , M. (1992). An Empirical Comparison of Menu-Selection (CUI) and Desktop (GUI) Computer Programs Carried Out by Beginners and Experts. Behaviour and Information Technology, Vol. 11, No. 4, pp. 227-236 27
REASON, J.T. (1990). Human Error. Cambridge University Press, USA RUMBAUGH, J., BLAHA , M., P REMERLANI , W., EDDY , F., & LORENSEN, W. (1991). Object-oriented Modelling and Design. Englewood Cliffs: Prentice-Hall. SANDBLAD, B., L IND , M., & N YGREN , E. (1991). Cognitive Work Environment Problems (In Swedish: Kognitiva Arbetsmiljöproblem och Gränssnittsdesign.) Report no. 20, CMD, Uppsala University. SANDBLAD, B., ÅBORG, C. & LIF, M. (1996). A Practical Method for Evaluation o f Human-Computer Interfaces. In Preparation. SCHNEIDER , W., & S HIFFRIN , R.M. (1977). Controlled and Automatic Human Information Processing I, Psychological. Rev., Vol. 84, pp .1-66. SCHNEIDERMAN , B. (1992). Designing the User Interface: Strategies for Effective Human- Computer Interaction, 2nd edition. Addison-Wesley Publishing Company, Inc. USA SCHULER , D., & NAMIOKA , A. (eds.) (1993). Participatory Design: Principles and Practices. Lawrence Erlbaum Associates, Inc., Hillsdale, New Jersey. SHARP, H. (1991). The Role of Domain Knowledge in Software Design. Behaviour and Information Technology, Vol. 10, No. 5, pp. 383-401. SHEPHEARD A. (1989). Analysis and Training in Information Tasks. In D. Diaper (ed.) Task analysis for Human Computer Interaction. Chichester: Ellis Horwood. S HIFFRIN , R.M. & DUMAIS , S.T. (1981). The Development of Automatism. In J.R. Andersson (Ed.) Cognitive Skills and their Acquisition. Hillsdale, NJ, Erlbaum. TULLIS, T.S. (1981). An Evaluation of Alphanumeric, Graphic, and Colour Information Displays. Human Factors, Vol. 23, pp. 541-550. TULLIS, T.J. (1993). Is User Interface Design just Common Sense. In G. Salvendy & M.J. Smith (Eds.) Human-Computer Interaction. Proceedings of the HCI International ’93, Orlando, Florida. Elsevier, pp. 9-14. WALLACE , M.D., & ANDERSON , T.J. (1993). Approaches to Interface Design. Interacting with Computers, Vol. 5, No. 3, pp. 259-278. WOODS, D.D., & ROTH, E.M. (1988). Cognitive Systems Engineering. In M. Helander, (ed.) Handbook of Human-Computer Interaction. Elsevier: North-Holland.. Z IEGLER , J.E. & F ÄHNRICH , K.P. (1988). Direct Manipulation. In M. Helander (ed.) Handbook of Human-Computer Interaction, Elsevier Science Publishers B.V. (NorthHolland).
28
