CONTEXTUAL DESIGN | Going Virtual: Distributed Communities in Practice

Contextual Design is broadly ethnographic in its approach and falls ideally between participant and non-participant observation so that the researcher is not completely immersed in the work yet is more than a mere observer. The method was primarily developed for work analysis and redesign and provides a structure for data collection but also offers models for working with the data and clear steps for how to move from the rich data to design issues. It provides a structure, models, and steps; however, these are not a rigid, restrictive framework but are intended as a support (Beyer & Holtzblatt, 1998):

How to get data about the structure of work practice…how to make unarticulated knowledge about work explicit…and how to get at the low-level details of work that have become habitual and invisible…These problems suggest an open-ended, qualitative approach that brings us in contact with the customer's real work^[1] (p. 37).

Contextual Design moves from qualitative data to themes and models to work redesign. It is intended to be a multi-disciplinary team effort and goes through seven clearly defined stages:

Inquiring and collecting data;
Interpretation session;
Work models;
Affinities and model consolidation;
Work redesign;
User environment design; and
Prototype evaluation.

The first four of these stages were used for exploring the interactions of WWITMan.

The Contextual Design method is based on Beyer and Holtzblatt's (1998) experiences in the field, but it has its roots in ethnography and the development of Grounded Theory (Glaser & Strauss, 1967), which has recently been used in Information Systems research (Howcroft & Hughes, 1999). In the development of Grounded Theory, data is collected. The researcher then develops conceptual categories from the data, followed by the collection of more data to expand on and inform the categories already created. The theory develops from the data itself (Frankfort-Nachmias & Nachmias, 1996). This procedure uses an ethnographic approach and develops the categories as research is proceeding, starting with an in-depth study and being followed by subsequent studies, not necessarily to the same depth, being used to inform the findings that have already been developed. These subsequent studies tend to be searches for confirmation or elaboration. Contextual Design adds to the development of Grounded Theory as Glaser and Strauss (1967) emphasised the need for a codified procedure for analyzing the data in order to convey credibility. Although a purely ethnographic approach would not use frameworks, the models in Contextual Design provide additional support to the researcher and go some way to addressing some of the difficulties outlined above and provide the researcher with a method that can be followed.

The advantages of using Contextual Design are several. The data gathering is ethnographic in form, allowing the researcher to see the work, practice, and interactions of a CoP in context. The method provides a range of models to handle the data, using and extending standard techniques for handling qualitative data, but also employing five different types of models designed to gain a full understanding of the practice being observed. I intended using the method as a framework and adapting it as necessary.

For the initial stages of the study I used a slight adaptation of the method. This involved the gathering of data using Contextual Design methods. The models were used but were adapted as necessary. Contextual Design is essentially a team-based activity, but I tackled this by involving other colleagues and interested parties to provide different perspectives. The fact that I was working primarily alone on the study did have benefits, as it meant I had a detailed view of the whole data set. Contextual Design is intended to involve the interviewees (customers) in the process as much as possible. It was not possible to do this according to Contextual Design principles, so I made use of the "Challenging Assumptions" stage of Dearden and Wright^[2] (1997), which takes propositions back to the respondents. Time and cost constraints meant that the case study could not be longitudinal and therefore took the form of two shorter periods (snapshots) with the CoP. Stage One formed the first of these two periods.

The study of the CoP was also supplemented by smaller studies of other CoPs. This follows the principles of the Discovery of Grounded Theory (Glaser & Strauss, 1967) on which Contextual Design is partly based, and provided expansion and confirmation (or otherwise) of the issues that arose from the study of the CoP. In summary, I spent Stage One with part of the CoP. After examining the issues that arose from my time with the CoP, I went back and spent more time with them to focus on any issues that arose. Finally, I undertook shorter studies with other CoPs where I could focus purely on key issues arising from the study of the main CoP. Due to access and time limitations, and the fact that I now had a much tighter focus, open semistructured interviews were used with these other CoPs.

^[1]Customer: as the overall aim of Contextual Design is the creation of a computer system, respondents/subjects are referred to as customers.

^[2]One approach to tackling the problem of handling a wealth of ethnographic data was used by Dearden and Wright (1997) who used a mix of situated and non-situated approaches. The study reported by Dearden and Wright (1997) was subject to time constraints (22 person days) and was intended to analyse the ‘quality of fit’ between the work undertaken by a specific group of office workers and the IT system which had been put in place to support the work. The time constraint of the study meant that time could not be lost handling large amounts of ethnographic data. Dearden and Wright (1997) therefore used an ethnographic technique (contextual inquiry) for observing the work in context after having had a training session to quickly learn the procedures and terminology in use. As the focus of the work was welldefined it was possible for them to use non-structured techniques such as semi-structured interviews and rich pictures as used in soft systems analysis (Patching, 1990). The understanding of the work was validated by the use of three techniques:

Model building: this forces the researcher to work through his or her understanding of the situation.
Challenging assumptions: the researchers returned to the organisation and presented groups of interviewees with a number of propositions or assumptions, some of which were deliberately contentious, in order to encourage debate.
Wish Lists: The researchers made suggestions and created a ‘wish list’ from open-ended questions in the interviews. The wish list was then presented to the groups of interviewees (who did not know which were suggestions made by the researchers and which had come from the interviews) in order to obtain more ‘wishes.’ The final list was put into groups and then displayed for discussion.

The mix of situated and non-situated techniques, using models to handle some of the data, appeared to work very well for Dearden and Wright (1997) allowing them to undertake the research, handle the data and report within a short time scale.