Information Technology Reference
In-Depth Information
background knowledge. However, observations of working practice did not capture
much concrete detail and often turned into a dialogue between the requirements ana-
lyst and the users, since epidemiology is cognitive-intensive work which involves
abstract concepts rather than specific details.
When we first introduced the use of scenarios the epidemiologists found the
approach somewhat abstract and felt it was hard to contrive situations that were
not grounded in current work. However, as they became more familiar with the
method, they were able to think of specific examples of working practices, cur-
rent problems and how they would like the system to support their investigations.
Domain knowledge workshops were intended to capture abstract knowledge about
the concepts, terminology and semantics within the users' domain. The workshops
placed the epidemiologists in an unusual situation, asking them to discuss aspects
of their world that they take for granted. In order to make this task easier, elicita-
tion started with more concrete concepts, e.g. the different types of epidemiological
study, and then moved to more abstract questions. This approach worked well,
and the epidemiologists commented that they found the workshops interesting and
engaging.
Two representations are notable omissions from the case study analysis: infor-
mal and formal models. The absence of formal models is not surprising in an
iterative user-centred RE process that was closely related to agile development
approaches. However, the role of informal models is diverse and needs more expla-
nation. Use cases, data flow diagrams and class diagrams were used but only among
the developers. The reasons why these representations were not shared with the
users, in contravention of the common ground framework, were twofold. First, the
requirements analyst was a bioinformatician and domain expert rather than a com-
puter science-trained requirements engineer; hence her common ground with the
stakeholders was not directly supported by models. Secondly, even though the intro-
duction of use cases was encouraged, scenarios and other concrete representations
provided sufficient common ground to support development of mutual understand-
ing even for abstract viewpoints on process and data structures. This did incur
some penalties in misunderstandings that might have been discovered earlier in the
process; for example, the expert researcher workflows were not articulated clearly
and use of process dependency diagrams could have focused attention on resolving
ambiguities. In data specification, the division of continuous distributions into dis-
crete categories was important for all statistical analyses, yet this remained unshared
tacit knowledge for some time. More explicit data modelling may have identified
this issue.
In spite of these limitations the use of value-based RE to focus attention on topics
which required mutual understanding, coupled with use of multiple representations,
served the project well. Storyboards were useful in developing a common ground
of design ideas, since alternatives and variations could be deployed quickly within
requirements sessions, either by drawing, swapping illustrations or adding post-
it notes. Prototypes supported validation by eliciting detailed feedback on design
features; however, this was combined with discussion of more abstract concepts,
such as workflows for both stakeholder groups and how the system design might be
Search WWH ::
Custom Search