Collaborative Requirements Engineering: Bridging the Gulfs Between Worlds - Intentional Perspectives on Information Systems Engineering - page 354

allocation, are applied to RE tasks and processes to suggest how RE might be
improved. Activities in the RE road map [ 16] are reviewed from a common ground
perspective while investigating the role for appropriate representations.
4.1 Elicit and Summarise
Elicitation commences with little common ground between the users, domain
experts and requirements engineer. This task involves not only capturing informa-
tion but also making sure that there is a shared understanding about domain facts
and user goals. Representations play an important part in summarising information
so it can be discussed and checked by all parties. Without representations transient
speech in interviews can allow ambiguities and mismatches in understanding to
go unchecked. Information representations: sketches, drawings, photographs and
video, all help to supplement text and enable elicited facts to be inspected and their
meanings evaluated. Scenarios provide records of specific episodes; furthermore,
their realism encourages the development of common ground since specific stories
and descriptions of the real world are anchored in the users' experience. Similarly
sketches, drawings and photographs record reality. However, common ground in
RE presents a paradox. The conversation has to progress towards a mutually agreed
project: realisation of the users' goals. But this dialogue, as has been increasingly
acknowledged in RE, involves exploration of the solution space that might satisfy
the users' goals. Specifications inevitably have to be generalised and abstract. While
an abstract view of specifications is closer to the personal common ground of soft-
ware engineers, abstractions tend to be alien to most users. Hence common ground
has to be established between abstract and concrete views, which causes a tension
that runs through all RE activities. At the elicitation stage, representations will be
biased towards the users' common ground and hence tend to focus on specific details
in scenarios, sketches, etc., although goal trees may start development over a more
abstract view of intentions.
4.2 Analyse and Reflect
The tension between abstractions and specific representations is central to mod-
elling. Generalisations are derived from specific examples, so models arise from
scenarios. Unfortunately scenarios impose a constraint on developing abstract com-
mon ground, since each scenario focuses on a specific slice of reality. This raises
the coverage question: how many scenarios are necessary to capture not only the
generalities but also the exceptions in a domain? Increasing the number of scenar-
ios suffers from a law of diminishing returns since the additional detail imposes an
information overload on stakeholders. Furthermore, adding more scenarios runs the
risk of focusing attention on superfluous detail. There is no ideal solution to the cov-
erage problem, although systems sampling and generative tools can help to increase
confidence that models have captured the more important variations [ 35] .