Towards a Framework of Design Models

Mathilde Bekker, Stephanie Wilson

HCI Laboratory, Dept. of Computer Science
Queen Mary and Westfield College, University of London
Mile End Road
London E1 4NS, UK
Tel: +44 (0)171 975 5257
E-mail: tilde, steph@dcs.qmw.ac.uk

ABSTRACT

This paper presents a framework for examining design models (e.g. task models, dialogue models) and their usage in the design process (e.g. elicitation, evaluation). The framework provides a basis for describing design methods in terms of the models they incorporate, for exploring the common ground between methods and for analysing studies of design in practice. The ideas are exemplified in a discussion of the use of task models in three design methods. Keywords Design methods, models, design support, task models.

INTRODUCTION

Many types of information, often explicitly represented as 'models', are used in the design of interactive systems, e.g. user models, task models, dialogue models. Each model has various uses. The models in combination with their use form the basis of design methods and techniques. Design methods generally describe a series of phases, where each phase involves one or more activities using models, e.g. eliciting the information in the model, manipulating the model or using the model to create further design models. Overviews (e.g. [5]) and taxonomies (e.g. [4]) have attempted to provide a structured view of the numerous design methods and techniques existing today. These focus on the design methods that are available to support particular phases of the design process, often with the aim of offering guidance to design practitioners. This paper argues that there are advantages to an alternative approach which views design from the perspective of the models involved, and presents a framework which provides such a view of design models and their use. Any given design method can then be described in terms of that subset of the models and their uses that it incorporates. This framework is a tool for discussing and researching design. Specifically, it provides a means of exploring and analysing the design model 'space'; it allows us to examine the common ground between existing design models, and therefore between the various design methods. Further, in our future work we intend to use the framework in an analytical way, to describe studies of design in practice and to assess the use of design models in this context. In the remainder of this paper we discuss the framework and illustrate its use by comparing and contrasting the different uses of one category of models, namely task models, in design. Finally, we discuss how the framework will provide the basis for our future research.

A FRAMEWORK

The significant distinction between the framework presented here and others that have been reported (e.g. [4] and [5]) is that it does not describe design as a linear, if iterative, sequence of organised phases. Clearly, any design project will consist of a sequence of activities, but our claim is that these activities do not fit into the neat pigeonholes described by the promoters of particular design methods. For example, 'requirements analysis' is a phase that design methods generally schedule relatively early in the design life cycle. While many requirements may be identified during this time, others may emerge much later as the design progresses, perhaps during the evaluation of a prototype. In our framework we are concerned firstly with the information that is identified, namely the requirements, and secondly with the actual activity of identifying requirements, which may occur at any time, as opposed to a defined slice of time labelled 'requirements analysis'. Other activities will also inevitably occur during this phase, for example, some initial generation of design solutions. Hence, we are proposing a framework which describes design using a number of dimensions: • The models used in design as explicit representations of design information, e.g. task models, object models, design rationales, dialogue models, designed systems. • The ways in which the models are used: design activities, e.g. eliciting the information, generating design ideas, evaluating design ideas or verifying whether the information expressed in the model is accurate. • The people that are involved in the design activity, e.g. designers, programmers, human factors experts, users. • The representations that are used to express the design models, e.g. textual notations, graphs and prototypes. This is by no means an exhaustive list of dimensions that are of interest in describing design models. Clearly, the relevance of the various dimensions depends on the context of use of the framework. Figure 1 shows a subset of the dimensions of the framework: the categories of models, their use in design and the people involved in their use (space does not allow us to give definitions of the dimensions' categories).

Figure 1: Models, their uses and the people involved in their use (the categories are not intended to be exhaustive).

ILLUSTRATION OF THE FRAMEWORK

One application of this framework can be illustrated by considering the possible uses of one form of model, say task models, in design. Different design methods may incorporate one or more of these uses. For example, consider GOMS [1], ADEPT [2] and CARD [3]. All three make use of some form of task model to describe the users' existing tasks, expressed in terms of goals and subgoals that the user seeks to achieve, and the objects that are used during the task. Note that we are not claiming that the three forms of task model are identical, but that they are used to represent similar information. The other dimensions (activities, representations and people involved) are summarised below: • GOMS [1] represents a task model using a structured textual notation. Designers (with knowledge of human factors) can use the model to document the task and to assess the usability of a system by predicting user behaviour. • The task model in ADEPT [2] is represented using a structured graphical notation. Designers can use it to document the users' tasks and to create design solutions. • The CARD technique [3] uses cards to represent the components of the task model. The model can be used by designers and users for documenting and creating a new work flow and also for critiquing or evaluating an existing system.

Figure 2: Comparing the use of task models

By fitting the three task models into the framework, we can compare and contrast the different uses for which they are intended (Figure 2). This shows how design methods can complement each other in their use of models. For example, by combining the CARD technique for eliciting information with the ADEPT approach for creating design solutions, we could combine elicitation and design generation. Unfortunately, this simple example does not show how a more complete set of models could be compared and contrasted. For example, a more detailed comparison could be achieved by using the framework to compare the complete set of models used in ADEPT [2] against a set of participatory design models, such as CARD and PICTIVE [3].

DISCUSSION

The framework outlined here allows us to describe design methods in terms of the models they employ, irrespective of the design phase for which they are used, providing a basis for comparison of models that use similar information for different purposes. Further, it allows us to identify gaps in the available design methods. In this way, new insights can be gained into how methods may be enhanced or used in combination. A longer term goal is to offer guidance to practitioners in selecting the models and methods that might be appropriate for a given design situation and in assessing the trade-offs involved in selecting one approach rather than another. The framework is continually evolving. While the work to date has been largely based on reports in the literature describing the 'ideal' use of various models and methods, we are continuing to elaborate the framework as we gather further information from ongoing studies of design in practice. In our work, we are particularly interested in how design models can facilitate or inhibit user involvement in design, and the trade-offs associated with that involvement. Thus, an important dimension of the framework to be further elaborated is the level of user involvement in the various design activities.

ACKNOWLEDGEMENTS

This work is funded by the UK Engineering and Physical Sciences Research Council, grant no. GR/K19211.

REFERENCES

Card, S.K., Moran, T.P., and Newell, A. The psychology of Human-Computer Interaction, Lawrence Erlbaum Associates, Hillsdale NJ, 1983.
Johnson, P., Johnson, H. and Wilson, S. Rapid Proto-typing of User Interfaces Driven by Task Models. In Scenario-Based Design, Carroll, J.M. (ed.), John Wiley & Sons, 1995, 209-246.
Muller, M., Tudor, L.G., Wildman, D.M., White, E.A., Root, R.W., Dayton, T., Carr, R., Diekmann, B. and Dykstra-Erickson, E. Bifocal tools for scenarios and representations in participatory activities with users. In Scenario-based design, Carroll, J.M. (ed.). John Wiley and Sons, 1995,135-163.
Muller, M.J., Wildman, D.M., and White, E.A. Taxonomy of PD practices: A brief practitioners guide, Communications of the ACM, Vol 36 (4), 1993, 25-28.
Olson, J.S. and Moran, T. Mapping the method muddle: Guidance in using methods for user interface design. In Human computer interface design: Success cases, emerging methods, and real world contexts. C. Lewis et al (eds). New York: Morgan Kaufman, 1995.