LTL-based declarative process models are very effective when modelling loosely structured processes or working in environments with a lot of variability. A process model is represented by a set of constraints that must be satisfied during the process execution. An important application of such models is compliance checking: a process model defines then the boundaries in which a system/organisation may work, and the actual behaviour of a system, recorded in an event log, can be checked on its compliance to the given model.
A known pitfall in specifying such a set of constraints is allowing for more behaviour than the intended one. Consider, for instance, a model with a constraint specifying that an invoice must be eventually followed by a payment: if invoice never occurs in the event log at all, the behavior is considered as compliant, while it is vacuously-compliant. Is this the behaviour we intended to model? Is this the right idea of compliance we had in mind? Undoubtedly, only the process designer who modeled that process can answer these questions. In this paper, we provide the designer with patterns that, applied to the model, trigger posing the right questions and define options for strengthening the constraints. Our patterns are inspired by vacuity detection techniques working on a single trace. We take the log point of view instead to check whether the constraints of a compliance model constrain enough.
|Status||Gepubliceerd - 2011|