Research Challenges on Person-centric Flows

Abstract

Research in the domain of Work ow Management focuses increasingly on service orchestrations. Often the fact is neglected that a huge part of the activities of business processes are performed by people. Especially, in the domain of pervasive computing processes are describing sequences of real world activities which are invariably performed by people. Therefore we consider the role of people participating in work ows from a new perspective. The basic idea of this work is to transfer the work ow metaphor to people processing their tasks. Therefore, we introduce the concept of a person-centric ow, which denotes such an implicit ow scheduled and executed by a single person. Secondly, we provide a list on research challenges on person-centric ows.

Introduction

Most research in business process management focuses on the aspects of service orchestration. The fact, that in certain types of businesses a business process consists of many manual activities, has not been recognized as a major research item, yet. Manual activities are carried out by people. We call those activities in the following tasks to di erentiate them from the other types of activities, such as an invoke activity that requests the execution of a Web Service. For example, pervasive computing processes are typically describing sequences of activities which are performed by people. People are involved in many business processes at the same time, which requires appropriate support from the underlying IT infrastructure to help organize the work they need to carry out. Among many other functions such a infrastructure requires capabilities for work prioritization, context-sensitive assistance, or the capability of structuring a task by dividing the task into smaller tasks and ordering them as required by the user. Several speci cations such as [ 1 ] and WS-Humantask [ 2 ] have been developed to address the integration of people into the world of services. We claim that those speci cations only partially provide the required support and that additional capabilities must be provided by work ow management system to better integrate people in the execution of business processes.

This work is partially funded by the ALLOW project. ALLOW (http://www.allow-project.eu/) is part of the EU 7th Framework Programme (contract no. FP7-213339).

Since the cognitive capabilities of people do not allow to carry out many tasks in parallel, people order the di erent tasks in a way that is convenient for them to process the di erent tasks. Current state of the art in work ow management typically provide the notion of work lists (cf. [ 3 ]), which the user can sort in any desired way using the properties that are associated with each task. Incidentally [ 4 ] shows that the e ciency of users highly depends on the capability for individual ordering of the work list. A disadvantage of larger work lists is the increased probability of missing the schedule for a task.

Studies have shown that people not only use the basic ordering of the work lists to select the next set of tasks, but also use some correlations or similar context information between the di erent tasks to select them [ 4 ]. This is similar to traditional work ow paradigm where the execution of the di erent activities is driven through the context that is associated with a process instance. We propose in this paper to order the individual tasks via process templates that are dynamically generated from the available tasks via the user preferences and automatically instantiated.

We introduce the concept of a person-centric ow, which denotes such an implicit ow scheduled and executed by a single person. Basically, a personcentric ow consists of the tasks on a person's worklist extended by additional ordering information. However, the knowledge about the existence of such a implicit person-centric ow cannot be utilized. Thus, making the person-centric ow explicitly visible and observable opens great opportunities to support people in doing their work. For example recommendations on the task ordering can be made which helps people e.g. to save resources or time. The goal of this work is to list research challenges on person-centric ows, their modeling, and execution. The aim of person-centric ows is not to prescribe people an ordering for performing their tasks. In fact, we want to support people e.g. by calling their attention to possible errors or to the fact, that the chosen ordering is prohibited. Another advantage of person-centric ows is the possibility to adapt the environment to needs of a future task. This is particularly useful in the case of actions that need time consuming preparation (cf. [ 5 ]) For example the lling of the bathtub can be started in time, as soon as it is known when a nurse will start washing a patient. For simpli cation reasons this work assumes that all tasks are generated by a Work ow Management System (WfMS).

The outline of this paper is as follows: Section 2 presents a scenario which is used to explain some issues of person-centric ows. A brief de nition of personcentric ows is presented in Section 3. Section 4 lists the research challenges concerning person-centric ows and related approaches. Finally, Section 5 concludes the paper. 2

Scenario

Our scenario is situated in the domain of Healthcare (cf. [ 6 ]). Figure 1 shows simpli ed versions of the work ows executed each morning for each patient. All steps of the two work ows have to be done by a nurse and all steps of the medication ow has to be done by the same nurse. Imagine there are two patients Frank and Tobias situated in two di erent rooms and one nurse Hanna. Imagine the situation where Hanna has executed the prepare medicine and fetch medicine task for both patients. As a consequence, Hanna has four tasks to perform which all appear on her worklist. Her person-centric ow could be rst to medicate and wash Frank and afterwards to medicate and wash Tobias. In another situation she may decide to medicate both patients before washing them. This can be considered as an adaptation of her ow.

Patient  Medication 

Flow

Prepare

Fetch

Administer

Patient  Washing 

Flow

Wash

The scenario shows that a person-centric ow does not just follow a static ow model. It is highly dynamic due to context and changes of the worklist. Hanna implicitly forms her ow according to her actual situation. Her ow is not visible to the outside completely. Only the history of the ow is visible. Secondly, while executing a prepare task of an instance of the medication ow, by experience she schedules a fetch and a medication task even if these task are actually not on her worklist yet. By experience she knows that the a fetch medicine task is followed by a medication task. Hence, she considers the medication task right at the moment of scheduling her tasks, although only the fetch medication task is active. 3

Person-centric Flows

The aim of this section is to provide a rst informal de nition of the term and concept of person-centric ows. Later on the de nition serves as foundation in order to identify research challenges. As mentioned before all tasks arise from ow instances executed by a WfMS which also provides a person's worklist. In the second part of this section we present the types of person-centric ows we identi ed during our research. 3.1

De nition De nition 1 (Person-centric ow). A person-centric ow de nes a partial ordering over a set of tasks which have to be performed by one single person. Tasks of a person-centric ow can be classi ed in three tenses: past tasks, present tasks, and future tasks. Past tasks are completed either correct or incorrect and their ordering is known. Present tasks are tasks currently present at a person's worklist. Their ordering is planned by the person but can change dynamically. Future tasks are tasks which are assumed to be executed in the future. Both the set of future task as well as their ordering may change dynamically.

Present

Future

During our research we recognized the following classes of person-centric ows:

Predicted ow: This is a predicted ow which corresponds to a person's ow with a high probability. Such ows are used for example to adapt the environment of a person. If washing a patient is supposed to be a nurse's next task the light in the shower could be switched on automatically. Regarding our scenario the knowledge of the nurse's ow helps to decide which patient will be washed rst and consequently in which shower the light has to be switched on.

Guidance ow: A guidance ow is based on a predicted ow. However, a guidance ow may be extended by external ordering constraints. The aim of a guidance ow is to guide people actively through their ow. For example a constraint demands that a nurse must disinfect her hands between washing two patients. If she decides to perform both washing tasks without disinfecting her hands between, the constraint will be violated and the system will inform her. 1 The gure uses an illustration by Patrick J. Lynch, medical illustrator and C. Carl Ja e, MD, cardiologist.

Enforcement ow: An enforcement ow orders the tasks of a patient only based on external constraints. In this case the person-centric ow of a person is given by the environment. For example the fact that the medication must take place before washing can be enforced by an enforcement ow and the system can drive the person accordingly. A major disadvantage of this kind of person-centric ow is that it deprives people of the freedom to decide in which order they want to perform their tasks.

However, there is no clear distinction between the three classes of personcentric ows. A ow can comprise enforced as well as predicted as well as guided parts. Mostly, only a meaningful scenario-based combination of the three classes facilitates to reduce the cognitive load of a person necessary for scheduling task while retaining a high degree of freedom in scheduling tasks. Sometimes the characteristic of a person-centric ow may change over time. While a unexperienced nurse should be guided, with increasing experience the guidance should be reduced. As of that time the person-centric ow is only used for adapting the environment. 4

Research Challenges

In this section we present the identi ed research challenges on person-centric ows derived from the scenario analysis. The list is not exhaustive. Analyzing other scenarios may identify further issues.

Relation between implicit and explicit ow: As mentioned earlier a person-centric ow is implicit created by a person. In order to utilize the personcentric ow a explicit representation must be created. Like a hidden markov chain the observable behavior of the explicit ow may di er from the hidden implicit ow. Anyhow, the relation between these two ows must be described. The relation may be de ned by describing the possible deviations of the predicted person-centric ow compared to the person-centric ow a person has in mind. Such an approach for modeling process variants is presented in [ 7 ]. Additionally, the predicted ow can be annotated with a maximum likelihood of its occurrence.

Rethinking the notion of ow model and instance: A person-centric ow has no prede ned ow model. Since tasks emerge and disappear continuously, the person-centric ow has to be adapted very often. Thus, a language has to be de ned that can deal with such a high degree of exibility. Furthermore, the notion of an instance must be rethought. An instance of a common ow model has a de ned beginning and a de ned end. In contrast to traditional work ows a person-centric ow has no de ned beginning and end. People are continuously performing tasks. Although there might be gaps in the ow where no tasks are performed.

Imperative vs. declarative process modeling languages: Basically, two types of work ow languages exist: imperative languages (cf. e.g. [ 3 ]) and declarative languages (cf. e.g. [ 8 ]). While imperative languages are assumed to be easier to understand, declarative languages promise a higher degree of exibility. Furthermore, di erent orderings of the same set of tasks are hard to model using imperative languages. In this work we use [ 9, 10 ] as starting points for our discussion on declarative and imperative work ow modeling languages, since the authors aim to contribute to a rigorous, theoretical discussion of this topic. Since high exibility is a requirements for person-centric ows we decided to use a declarative language for the rst attemp to describe person-centric ows (cf. [ 11 ]). The language is based upon the interval relations of Allen's interval algebra [ 12 ]. In contrast to existing declarative languages (cf. [ 8, 13, 14 ]) we extended the constraints of our language by time parameters to add support for temporal restrictions. To validate the consistency of the relation constraints we follow an approach similar to the one suggested in [ 15 ]. Also case handling [ 16 ] provides more exibility in executing work ows. In case handling the control- ow related information does not drive the process but the state of the process, i.e. the existence of data objects. Since our approach focuses on the tasks themselves and not on the state of a data object, case-handling is inappropriate for modeling person-centric ows.

Algorithms for prediction: Since we are not able to capture the personcentric ow a person has in mind, we need to nd algorithms to determine the task orderings. Especially history based algorithms are promising since people often have behavior patterns which can be detected by history-based algorithms. But also algorithms operating e.g. on task deadlines without considering the history (e.g. scheduling algorithms) seem to be useful in certain scenarios. There are many approaches which can be used as a starting point. In [ 17 ] the authors present an approach, which can be used for predicting a person's next steps based on the ow history. The recommendation in this approach is made for a single process instance and doesn't consider the interleaving of di erent process instances. In [ 18 ] scheduling algorithms are used to order worklists. Such an approach is suitable, if the person-centric ow should be enforced. [ 19 ] provides an approach which can be used for mining behavior patterns of people. Algorithms for predicting a person's next location are presented in [ 20 ]. The latter two approaches can be used, if there are recurring behavior patterns. However, they have to be adapted in order to operate on work ow histories.

Granularity of tasks: Tasks in business process are often modeled on a higher level than they will be actually executed by the people. Hence, people divide tasks in a set of subtasks. For example the washing of a patient is divided in several sub-tasks by a nurse. Algorithms for predicting person-centric ows also have to consider the fact that people re-structure their tasks. In [ 21 ] a case study is presented showing how people re-structure tasks.

Parallelism of human actions: It has to be discussed whether it is reasonable to allow parallelism in person-centric ows. Hence, granularity implies that if task are divided in sub-tasks which might be executed interleaving with subtasks of other tasks urge for parallelism capabilities.

Visualization: If person-centric ows are made explicit they should also be visualized in order to present them to a person, e.g. in the case that the personal ow is a guidance ow. However, meaningful representations must be found. For example, the person-centric ow of a nurse may be integrated within a map of the hospital. Such a scenario may be realized based on an approach for workitem visualization as presented in [ 22 ]. In [ 23 ] a work ow visualization framework is presented, which enables a exible business-oriented process visualization (e.g. using a gant chart). This approach can also serve as foundation for a visualization discussion on person-centric ows.

Privacy: Person-centric ows contain a lot information about the speci c behavior of a person, which should be protected. Hence, discussing privacy is an important issue. [ 24 ] gives an overview on the e ects of combining work ow systems and context-aware systems and to discuss its implications for workplace privacy and human-oriented design. 5

Conclusion

In this paper we introduce the concept of person-centric ows. Since personcentric ows are orthogonal to the existing work ow concept the two concepts complement each other. The main goal of person-centric ows is to support a person performing her tasks for example by reducing the cognitive load necessary for scheduling their tasks. However, person-centric ows can also be used to enforce external constraints. Initially, we will use the declarative work ow modeling language proposed in [ 11 ] to representing person-centric ows. In future work, we investigate and implement a prediction algorithm to predict person-centric ows prototypically. The aim of the prototype is to guide unskilled nurses. The prototype is also used to improve the accuracy of activity sensing and context recognition. In that case the person-centric ow reduces the search space of the activity recognition system since two activities having the same characteristics of sensor data may be distinguished according to their position in the person-centric ow.

References

1. OASIS: WS-BPEL Extension for People Speci cation Version 1 .1, Committee Draft 06. ( 2009 )

2. OASIS: Web Services Human Task Speci cation Version 1 .1, Committee Draft 06. ( 2009 )

3. Leymann , F. , Roller , D. : Production Work ow { Concepts and Techniques. Prentice Hall PTR ( 2000 )

4. Card , S.K. , Newell , A. , Moran , T.P.: The Psychology of Human-Computer Interaction . L. Erlbaum Associates Inc., Hillsdale, NJ, USA ( 1983 )

5. Nurmi , P. , Martin , M. , Flanagan , J.A. : Enabling Proactiveness through Context Prediction . In: In: CAPS 2005, Workshop on Context Awareness for Proactive Systems . ( 2005 )

6. ALLOW: D3 . 1: Results of Scenario Analysis . ( 2009 ) Available online (http://www. allow-project.eu/).

7. Hallerbach , A. , Bauer , T. , Reichert , M.: Issues in Modeling Process Variants with Provop . In Ardagna, D., Mecella , M. , Yang , J., eds.: Business Process Management Workshops. Volume 17 of Lecture Notes in Business Information Processing. , Springer ( 2008 ) 56 { 67

8. van der Aalst , W.M.P. , Pesic , M. , Schonenberg , H.: Declarative work ows: Balancing between Flexibility and Support . Computer Science - R&D 23 ( 2 ) ( 2009 ) 99 { 113

9. Fahland , D. , Mendling , J. , Reijers , H. , Weber , B. , Weidlich , M. , Zugal , S. : Declarative vs . Imperative Process Modeling Languages: The Issue of Maintainability. In: 1st International Workshop on Empirical Research in Business Process Management (ER-BPM'09) , Ulm, Germany ( September 2009 ) 65 { 76

10. Fahland , D. , Lubke, D. , Mendling , J. , Reijers , H. , Weber , B. , Weidlich , M. , Zugal , S. : Declarative versus Imperative Process Modeling Languages: The Issue of Understandability . In: Proceedings of the 14th International Conference on Exploring Modeling Methods in Systems Analysis and Design (EMMSAD'09) , Amsterdam, The Netherlands, Springer-Verlag ( June 2009 ) 353 { 366

11. Wagner , S.: A Concept of Human-oriented Work ows . Diploma thesis , University of Stuttgart, Germany ( January 2010 )

12. Allen , J.F. : Maintaining Knowledge about Temporal Intervals . Commun. ACM 26 ( 11 ) ( 1983 ) 832 { 843

13. Pesic , M.: Constraint-Based Work ow Management Systems: Shifting Control to Users . PhD thesis , Eindhoven University of Technology ( 2008 )

14. Pesic , M. , Schonenberg , M.H. , Sidorova , N., van der Aalst , W.M.P.: ConstraintBased Work ow Models: Change Made Easy . In: OTM Conferences (1) . ( 2007 )

15. Lu , R. , Sadiq , S.W. , Padmanabhan , V. , Governatori , G. : Using a temporal constraint network for business process execution . In: ADC . ( 2006 ) 157 { 166

16. van der Aalst , W.M.P. , Weske , M. , Grunbauer, D.: Case handling: a new paradigm for business process support . Data Knowl. Eng . 53 ( 2 ) ( 2005 ) 129 { 162

17. Schonenberg , H. , Weber , B. , van Dongen , B.F. , van der Aalst , W.M.P. : Supporting Flexible Processes through Recommendations Based on History. [25] 51{66

18. Han, R ., Liu, Y. , Wen , L. , Wang , J.: A Two-Stage Probabilistic Approach to Manage Personal Worklist in Work ow Management Systems . In Meersman, R., Dillon , T.S. , Herrero , P., eds. : OTM Conferences (1) . Volume 5870 of Lecture Notes in Computer Science., Springer ( 2009 ) 24 { 41

19. Vrotsou , K. , Ellegard , K. , Cooper , M. : Everyday Life Discoveries: Mining and Visualizing Activity Patterns in Social Science Diary Data . In: IV, IEEE Computer Society ( 2007 ) 130 { 138

20. Petzold , J. , Bagci , F. , Trumler , W. , Ungerer , T. : Comparison of Di erent Methods for Next Location Prediction . In Nagel, W.E., Walter , W.V. , Lehner , W., eds.: Euro-Par . Volume 4128 of Lecture Notes in Computer Science., Springer ( 2006 )

21. Schroeder , W.: New Tools for Task Work ow Analysis . In: CHI EA '09: Proceedings of the 27th international conference extended abstracts on Human factors in computing systems , New York, NY, USA, ACM ( 2009 ) 3877 { 3882

22. de Leoni, M., van der Aalst , W.M.P. , ter Hofstede , A.H.M. : Visual Support for Work Assignment in Process-Aware Information Systems . [ 25 ] 67 { 83

23. Bobrik , R. , Reichert , M. , Bauer , T. : View-Based Process Visualization . In Alonso, G., Dadam , P. , Rosemann , M., eds. : BPM . Volume 4714 of Lecture Notes in Computer Science., Springer ( 2007 ) 88 { 95

24. Wieland , M. , L angerer, C. , Leymann , F. , Siemoneit , O. , Hubig , C. : Methods for Conserving Privacy in Work ow Controlled Smart Environments - A Technical and Philosophical Enquiry into Human-Oriented System Design of Ubiquitous Work Environments . In: UBICOMM 2009 , Sliema, Malta, IEEE ( October 2009 )

25. Dumas , M. , Reichert , M. , Shan , M.C., eds.: Business Process Management, 6th International Conference, BPM 2008, Milan, Italy, September 2- 4 , 2008 . Proceedings. In Dumas, M. , Reichert , M. , Shan , M.C., eds. : BPM . Volume 5240 of Lecture Notes in Computer Science., Springer ( 2008 )