So much planning, so much anticipation, and now the 4th International Workshop on Adaptive Case Management and other non-workflow approaches to BPM is over after one marvelous day. We had reserved some time at the end for a round table discussion, with some time in the morning to select topics. The subject of ‘The Purpose and Value of Modeling for Knowledge Worker Support’ quickly emerged as the dominant concern, and ended up being the main discussion point. Before we get to that, let me present a summary of the papers (see the program).
Case Management: An Evaluation of Existing Approaches for Knowledge-Intensive Processes
Matthias Hauder gave a presentation on their study into the field of case management. The problem is that all the different papers have different definitions of case management. They did a survey of the literature, and pulled the common characteristics, as well as common requirements. They propose a good definition for case management. They then looked at the requirements, and identified the ones required for modeling, and then looked to see if CMMN would fit the bill. I found the logic a little bit circular: because they assume that the environment will have some characteristics of a CMMN based environment. Specifically, that design modeling capabilities are a separate feature from regular usage, and that only some of the users would do modeling. Because they assume an environment similar to CMMN, it is not surprising that they find CMMN a suitable fit to the requirements — at least for the modeling portion. What they did not do is to validate that the modeling-based approach actually works in any situation. That migth be the next step for Matthias.
Declarative Process Modelling from the Organizational Perspective
Stefan Schönig gave a presentation on his analysis of five different declarative modeling languages Declare, DCR-Graphs, CMMN, DPIL and EM-BrA2CE. The first three of theses are graphical in nature, while the last two are text-based language. He particularly looked at how well the languages represent a couple of specific organizationally relevant patterns involving roles. The analysis was well done. The somewhat surprising result was that all three of the graphical notations showed significant deficiencies in representing these kinds of task assignment. While the text-based languages might do a better job of representing this requirement, they are widely recognized as being harder to use. DPIL is his preferred choice and he is actively involved in development of it. He spoke about how to put a graphical representation in front that would automatically translate to DPIL. Two comments from me: (1) he needs a better definition of ‘role’ since the one used tended to blur the line between ‘role’ and ‘group’ — a rather different concept. (2) If you have a graphical notation that faithfully represents what DPIL can do, they why bother with DPIL? This could be completely hidden and nobody would need know about it.
A Case Modelling Language for Process Variant Management in Case-based Reasoning.
Andreas Martin compared the expressiveness of various modeling techniques measured in a knowledge-work extensive use case: BPMN, CMMN, Declare, and BPFM. The user case was good: qualifying candidates for admission to their school which involves many different rules as well as international candidates having widely varying supporting evidence. BPMF is a like a decision tree, and seems to be a better fit for case management which is more about identifying goals and less about the process to get there.
Embracing process compliance and flexibility through behavioral consistency checking in ACM, A Repair Service Management Case
Christoph Ruhsam from ISIS Papyrus presented some approaches that might be used to allow for building of process diagrams at run time. One problem with letting knowledge workers change process models, or with composing process models from a set of pieces, is that the resulting model might contain internal consistency problem. This approach would allow for automatic checking of those consistency rules, to warn users immediately about problems.
Modeling crisis management process from goals to scenarios.
Elena Kushnareva presented the idea of using State Charts to model the process for an emergency response organization. For example, as a flood rises to different levels, there are associated necessary responses, such as closing bridges, roads, or evacuating certain regions It is a good use case because emergency response is very unpredictable, and it is important. State charts are particularly strong when you have nested state, but I remain unconvinced that the scenario requires a lot of nested state. It seems that they particularly use it in such a way that you have a lot of independent states that just happen to be in a containing box “emergency occurs.” The following box, recover from emergency, would need to go to some great length to retain the internal state of the previous box, but that following box was not elaborated in the presentation. Once again, my main question is “why do modeling”. In this case I see a clear need for an elaborate model done to run simulations and understand what needs to be prepared — that model done by people who specialize in modeling such things — but the actual emergency response worker don’t show any need to do modeling, or even to change the aforementioned model while an emergency is unfolding.
Supporting Adaptive Case Management Through Semantic Web Technologies
Wilhelm Koop presented an idea to use semantic representations (OWL, RDF) to help guide knowledge workers in choosing what is and is not an acceptable enhancement to a process model. This is also important as a way to determine whether the extensions made by two different knowledge workers are the same or not. Some form of semantic analysis of case history seem obviously critical in order to eliminate the arbitrary differences cause simply by the choice of words the workers use.
Supporting Knowledge Work by Speech-Act Based Templates for Micro Processes.
Johannes Tenschert proposed that instead of making models on deconstructing the human activities involved, we should instead base the models on how people communicate, particularly how they communicate to get things done. This is the realm of speech acts. They have a couple of basic patterns: a promise, a commitment, a question, a declaration of completion, etc. I personally think there is a really a lot of promise to this approach, because organizations are ultimately social entities, and what matters is what you say was done, not necessary how you did it. Their research is just starting on this, and what they need is to elaborate a practical example, and to show that this approach would result in an improved ability to support the work.
Towards Structural Consistency Checking in Adaptive Case Management.
Christoph Czepa gave the second paper on how to automatically detect consistency problems in a language like CMMN. For example, the exit criteria of one node can directly contradict the entry criteria of the following node, and produce a graph that can never be traversed. Model checking is used to extend this detection ability to logical contradictions that would not be immediately obvious. This type of logical check would seem an excellent feature on any process modeling tool.
Towards Process Improvement for Case Management. An Outline Based on Viable System Model and an Example of Organizing Scientific Events.
Ilia BIder put forth that if you want a system that modifies itself, or at least an organization that modifies the system it uses, then the Viable System Model from Cybernetics is worth investigating. Gave a VSM overview. Then present the use case of running a scientific meeting, much like the workshop we all are attending. He gave a list of tasks, and pointed out that each is likely to be needed in any event, so the goal is never to ‘optimize the process’. He mapped the various parties that are involved in a workshop to the 5 different ‘systems’ defined by the VSM. He stopped short of demonstrating that this was an effective way to structure an ACM system, but leave that up to a future research to take to the next step.
This all led up to a large discussion on the merits of modeling: what good is it? what should it be used for? what are the goals? what should it not be used for? how do we measure the effectiveness? and do we need to model at all? The first day of the BPM conference is starting here in Innsbruck in a few minutes, so you will have to wait until tomorrow for my summary of this discussion on the merits of modeling for knowledge workers.
Pingback: No Model is a Good Model | Thinking Matters