A Goal-Based Approach for Learning in Business Processes - Intentional Perspectives on Information Systems Engineering - page 235

In this chapter we adopt the notions of business process goals as defined in the
Generic Process Model (GPM) framework [19] . GPM is a state-based and goal-
oriented view of a process, which relates to two types of process goals: hard goals
(or simply goals) and soft goals. Below we discuss these two types of goals and their
possible use for process learning.
The hard goal of a process is defined by GPM as a set of stable states at which the
process terminates. The goal set is specified by a predicate over values of the state
variables of the domain in which the process operates. Since a process is executed in
order to bring about some state of affairs in the domain, the predicate expresses the
conditions under which this state of affairs is achieved so the process can terminate.
The goal is a set of states (rather than a single state), since there might be different
specific states that meet the termination condition. For example, a sales process
reaches its goal once the order is fulfilled and paid for. This may include different
states (e.g., the goods were shipped to the customer, the customer has taken the
goods himself, payment was made in advance or upon delivery, etc.).
Considering learning, a process instance (namely, a specific execution of the pro-
cess) may end up in a state which is in the goal set or in a stable state which is not
in the goal set (an exception). In the sales process example, it might be that the cus-
tomer received the goods, paid with an invalid credit card, and lost contact, so he
cannot be located any more. In this case, the state of the process domain is stable,
namely, it cannot be changed by actions of the organization, but it is not in the goal
set since payment was not received. Learning seeks to avoid exceptional situations
or to minimize their occurrence over time.
As explained, the (hard) goal of a process is a set of states, all satisfying the con-
dition under which the process can terminate having achieved what it was intended
to achieve. These states might be different from each other in business terms. For
example, in the above mentioned sales process it might be considered more desirable
to supply goods in two days than in two weeks (although both lead to goal states).
Soft goals represent business objectives which differentiate states in the goal set
according to how desirable they are. In other words, soft goals define a desirability
order relation among states in the goal set [ 19] .
The term “soft goals” is borrowed from requirements engineering, where it
relates to desired properties whose satisfaction is not on a binary scale. Similarly,
considering business processes, soft goals correspond to performance indicators
whose increased values are sought, but they can only be considered successful in
comparison to others rather than absolutely. It is possible to define thresholds to
performance indicator values, so values above the threshold are considered “good”
(e.g., delivery time shorter than one week) as opposed to values below the thresh-
old. Yet, different values of soft goal related performance indicators denote different
levels of success even if all values are above the threshold.
Learning in a business process should seek to achieve higher levels of soft goal
related indicators over time.
It should be noted that specific soft goals (e.g., minimizing execution time) and
their relationships to actual paths have been addressed to some extent by process
mining approaches [1, 6] . Here we address soft goals at a generic level, without