A Goal-Based Approach for Learning in Business Processes - Intentional Perspectives on Information Systems Engineering - page 236

limiting ourselves to specific ones. This raises two main challenges. First, different
soft goals may exist and may even conflict with each other (e.g., quality and cost).
Process changes may positively affect one soft goal while negatively affecting the
other. Second, soft goals may be affected by more than one process. For example, the
quality of a product may be affected by the production process and by the purchasing
of raw materials. It follows that considerable attention should be devoted to the
precise specification of soft goals with respect to a specific process for learning to
be effective.
As explained, learning should assess the level of success of each process execu-
tion (process instance) with respect to the defined goals of the process (both hard
and soft goals). We refer to the combination of hard and soft goal achievement by a
process instance as the outcomes of the instance.
2.2 The Role of Context
The success of a process instance can be affected not only by the actual path per-
formed, but also by environmental conditions, not controlled by the process, which
we term the process context. Specifically, the context of a process includes the initial
state at which the process is triggered (which may hold specific case characteristics,
such as customer properties in a sales process) and events in the environment that
may occur during its execution. The initial state is specified by values of state vari-
ables known when the process is initiated; events in the environment are external to
the process domain but affect its state.
Process instances of different contexts may need to be addressed differently (i.e.,
take different paths) in order to achieve desired outcomes. Alternatively, we may
say that if exactly the same path is applied to process instances of different contexts,
it might lead to different outcomes. Considering a sales process, a regular cus-
tomer may place an order and pay once the goods are supplied, while an unknown
customer would be required to pay in advance to reduce the risk.
Furthermore, threshold levels of soft goals for determining whether an outcome
is “desired” or not may also depend on context. For example, a desired outcome of
a broken leg treatment process for an old person would be to be able to walk freely
again, while for a young person it would only be considered successful if he never
suffers pain again.
It follows that to learn effectively, process instances should be classified accord-
ing to their context, so that learning could take context into account. However, initial
information and external events may relate to a variety of factors, and it is not nec-
essary that all factors indeed affect the outcome of the process. Hence, the challenge
faced is to identify the relevant factors that should be taken into account by learning.
While some factors may be well known to domain experts and even incorporated
into the process model as decision criteria (e.g., regular vs. first-time customer),
others may be guessed intuitively by some workers, and some even unknown in
advance.