Conclusions

In this paper we put forth the argument that logic is an appropriate language to represent process models for work activity coordination. First, by using logic one is able to uniformly define process models, policies, constrains and so on. Second with logic it is possible to define a much more flexible relation between case and model. For example in a Petri-net based modeling (of which the standard WfMC workflow language can be seen as an example), the case is seen as a marking of the Petri-net. On one hand, this allows for a simple computation of what is the next activity, but it limits how much the case can diverge from the model, and how to represent less well defined models. In our logic framework, the relation between the case and the process model is one of consistency: the case will be said to follow the model if it is consistent with it. This allows for a much more flexible relation between model and case.

On the other hand, the flexible relation between case and model makes it impossible to compute what activities should follow another activity, if all is normal. So we defined a stronger logic that, using the same process models, allow for the determination of the next activities that should be performed. This extended logic allows one to model different modes of use of WACS.

Some previous research, mainly in the software process domain have similarities with the one proposed here. [1] also proposes a logic based formalism, enhanced with non-monotonic logics to represent processes. Rule based representations are common in both the workflow and the software process research. But in most of these research rules are interpreted procedurally, that is as condition/action pairs, instead as logic implication as it is used in this paper. Thus rules can only be used in a ``figuring out what to do next'' mode.

Although we used a modal temporal logic in this paper, nothing in particular hinges on that choice. We used that logic because we were interested on the (relative) temporal aspects of the representation. The logic representation would have to change to first order logic if one is interested in representing some of the other aspects of work activity coordination, such as absolute time (dates and deadlines), objects, data, resources, and actors involved in the execution of the activities, coordination across many workcases (in scheduling applications), and so on. We are currently expanding the logical framework presented here to include some of these aspects.

If logic is an appropriate representation framework for work activities coordination, it is not user-friendly. We make no argument that logic should be the representation visible to the user; a more user friendly representation must be devised. That user friendly representation would then be converted to logic expressions that are used by the WACS engine.

Finally, this paper claimed that logic is an appropriate representation for the issues regarding work activity coordination. It is not clear yet if it is an efficient representation: the complexity of the computations proposed must still be evaluated.