Social Workflows - Uni Trier

Social Workflows - Vision and Potential Study Sebastian G¨org, Ralph Bergmann∗ Department of Business Information Systems II University of Trier 54286 Trier, Germany [email protected], [email protected]

Abstract Social workflows pervade peoples’ everyday life. Whenever a group of persons works together on a challenging or multifaceted task, a social workflow begins. Unlike traditional business workflows, such social workflows aim at supporting processes that contain personal tasks and data. In this work, we envision a social workflow service as part of a social network that enables private individuals to construct social workflows according to their specific needs and to keep track of the workflow execution. The proposed features for a social workflow service could help individuals to accomplish their private goals. The presented idea is contrasted with established research areas and applications to show the degree of novelty of this work. It is shown how novel ideas for knowledge management, facilitated by a process-oriented case-based reasoning approach, support private individuals and how they can obtain an appropriate social workflow through sharing and reuse of respective experience. Two empirical studies confirm the potential benefits of a social workflow service in general and the core features of the developed concept. Keywords: Social Workflows, Potential Study, Process-oriented Case-Based Reasoning, Knowledge Management

1. Introduction During the past years, social networks have been a subject of increasing interest and have demonstrated significant benefits for their users. They allow users to make new friends and to stay in contact with old ones, but also to make connections with people with similar interests and goals. Today, social networks provide and integrate an increasing number of communication and collaboration services as well as specific apps. For example, the social network Facebook enables to form groups for information sharing about specific topics, it includes mailing and chat services for communication, instruments to enact polls, and apps such as Doodle.com for event scheduling or Spotify for sharing music playlists. While such online services already support the execution of particular activities of private individuals, they still cannot be integrated to form a more complex flow of activities. However, quite often private individuals have goals that are much more complex to reach and hence require a more detailed planning of several tasks to be done, involving different people such as friends or professionals. Examples include moving to a different city, changing jobs, changing cars, planning group vacations, etc. These more complex goals require the planning and execution of a particular type of workflow. For instance, moving to a new city requires searching available apartments first, then selecting potentially approved ones, making appointments with the landlord, visiting the apartment, and taking a decision. In this process, information about the quality of the city district are collected, appointments must be scheduled, and information about the local infrastructure are regarded. In this paper, we introduce the term social workflow to describe workflows which are of such a personal nature. We show that social workflows differ from traditional business workflows, which leads to different requirements on workflow management systems. Today, there is no support at all for constructing and executing social workflows, ∗ Corresponding author: Ralph Bergmann, Department of Business Information Systems II, University of Trier, 54286 Trier, Germany, Phone: +49 651 201-3875, Website: http://www.wi2.uni-trier.de, E-Mail: [email protected]

Preprint of paper accepted by Information Systems on 18 December 2014

Final paper available under http://dx.doi.org/10.1016/j.is.2014.12.007

although there are many useful services available in social networks or on cloud service platforms. We argue that this lack of support could be filled by a social workflow service that establishes linkages between these services and supports users to construct and execute social workflows according to their specific needs. In textual form, such social workflows are already described and shared in various Internet communities which provide how-to knowledge for achieving different goals, in areas like home repair or computer troubleshooting. Today, this is an important body of the existing experiential content on the web and a clear indication of private users’ needs and willingness for exchanging experiences on workflows to solve complex problems. The next section introduces social workflows and social workflow services in more detail. Section 3 describes a concrete example scenario of a social workflow and how a social workflow service could facilitate its execution for a group of private individuals. In Section 4, the concept of a social workflow service is described, including the technical components it involves. Section 5 presents an overview of related work, while Section 6 presents a formative evaluation of the overall idea of social workflows and the concept for a social workflow service. Finally, Section 7 draws conclusions, presents the current state of implementation, and outlines our future steps. 2. General Approach and Research Goals The research topics of this paper have been investigated as part of the WEDA project1 . WEDA aims at transferring the benefits of traditional workflow management to the everyday ‘business’ of private individuals. We call such workflows for private individuals social workflows. Of course, the requirements of private individuals differ from the requirements of a company. For this reason, we developed concepts which are expected to close the gap between traditional workflow management and the requirements of private individuals on workflow management. In contrast to a business workflow, a social workflow will most likely only be created if its automated execution significantly simplifies the life of the users in the coordination of a complex, collaborative activity. While business workflows shall ensure compliance to business processes by a repetitive and standardized execution, this is not an issue for private individuals. However, social workflows will rather be executed repetitively within a larger community of people with common interests or goals. In addition, different people have at least slightly different requirements, contexts, or environments, thus social workflows demand process flexibility [1, 2] rather than standardization. An essential step to enable private individuals to use a workflow management system is to exploit the features of social networks, because they are already used as platform for the organization of private individuals’ life. Based on this assumption, we developed a generic concept which integrates a social network with a workflow management environment [3]. Furthermore, this concept puts process-based knowledge management in a private context because it enables private individuals to share and reuse workflows as well as every kind of resource associated with them. The idea is to let private individuals model their knowledge and experience directly in the structured form of a workflow instead of using arbitrary free text and to give private individuals means to share their experience within a community. This approach, of course, relies on the willingness of users to collaboratively create repositories of workflows, which may continuously evolve and expand to new areas. However, similar ideas have already proven successfully for scientific workflows (cf. myExperiment [4], Wings [5]), i.e., addressing a community of professional researchers (see also Section 5), which collaboratively model and share scientific workflows. In summary, the management of social workflows will benefit from a social network by • using the network of people with similar interest to describe and share experience on social workflows, • exploiting existing online services for task enactment in a social workflow during its execution by a workflow engine, • using the network of people (and possibly professionals) during task enactment. Based on the above considerations, we now concisely define the terms social workflow and social workflow service as follows: 1 WEDA is an abbreviation for “Web-basierte Erstellung, Wiederverwendung, Dokumentation und Ablaufunterst¨ utzung f¨ur agile Workflows” which means “web-based creation, reuse, documentation, and execution support for agile workflows”. For more information refer to http: //www.uni-trier.de/index.php?id=40538&L=2.

2

A social workflow is an executable process representation, serving private individuals and groups of people to fulfil their objectives by providing means to describe and link personal activities and data objects according to procedural rules. A social workflow service provides a modelling and flexible execution service for social workflows addressing private individuals as users. This service includes means to organize, share, and reuse social workflows and the related workflow data within a virtual community of private users. Our research on social workflows is driven by concrete use cases which are subject of several empirical studies. In the next section, a detailed social workflow scenario is presented which shows how people could use a social workflow service to work together to pursue a complex goal. Using this scenario, we illustrate a technical concept of a social workflow service in Section 4. Further, the usefulness of this scenario (and others) is estimated in two empirical studies. These studies serve two purposes: first, as an exploratory investigation in order to identify promising application areas for social workflows and second, to evaluate if the proposed features of the social workflow service are perceived as useful. 3. A Social Workflow Scenario We now outline a scenario in which private individuals employ a social workflow service. It is designed according to the personal experiences of students from the University of Trier, who recently attended the famous German rock festival “Rock am Ring”. The fully elaborated social workflow is illustrated in Figure 1 to Figure 3. These graphical representations have been created using the current prototypical implementation of our social workflow service and its graphical workflow representation language (see Section 4). Let’s assume a group of friends wants to attend the “Rock am Ring” festival. When thinking about what needs to be done to make this an enjoyable experience for everyone, they recognize that there are several task to be completed, some of which require action by everyone, while others involve only some people or can be even automated. The group decides to use a new social workflow service, which promises comprehensive support for these kinds of group activities. Peter, one member of the group who has lots of experience in attending large rock festivals, volunteers to set-up a social workflow for attending this event. He thinks about what needs to be done and in which order. He uses the social workflow service to search for social workflows from other community members for similar purposes and thereby he is quite certain not to forget something important. Finally, he decides to model the festival workflow for their ‘Rock am Ring” event shown in Figure 1 to Figure 3. By doing so, Peter becomes the owner of this workflow. He sends a Facebook message to the group of friends with a link to the social workflow service in which the festival workflow is started and shared with them. In Figure 1 the beginning of this festival workflow is illustrated. The first task is to “Buy tickets for festival” which is assigned to all friends who want to go to the festival. The task assignment specified by Peter contains all relevant information for buying tickets (e.g. a link to the online shop) as descriptive information, making it easy for everyone to perform the task. Once, everyone has ordered a ticket, the first task is completed. Then a parallel execution of the next two tasks is automatically initiated by the social workflow service. The task “Create Spotify playlist” is a service task and is automatically executed. Spotify is a music streaming platform that allows to create personal and shared playlists. The automated task reads the Facebook profile information of all workflow participants and includes their music preferences (“likes”) in order to create a shared Spotify playlist. Later during the festival, the participants can access this automatically generated playlist and enjoy the music. At the same time the purchases of all items needed during the festival are planned in the “Plan purchases” task. This task is assigned to all workflow participants. They shall collect a list of needed items using a cloud editor such as ‘Google Docs’. If everyone agrees, the task is completed. Three people, Mary, Paul, and Sue are selected to perform the subsequent three tasks, i.e., to purchase the items from the list. These tasks are executed in parallel by the three and the previously created cloud document is passed to them (the dashed arrows indicate the data flow). In Figure 2 the next cut-out of the festival workflow is shown. First, the automatic task “Weather on destination” is executed by the social workflow service, which triggers a weather service to get information about the weather at the destination of the festival at the specified date. After that, a human task is activated that reminds all friends to “Pack everything” for the festival. At this step, not only Mary, Paul, and Sue need to pack their purchases, but everyone. To 3

Figure 1: A Social Workflow: Visiting “Rock am Ring” with friends. Part I.

support this activity, this task also passes a Word document which contains a standard equipment list. This document was discovered by Peter as part of a different social workflow he found quite useful when creating the current one. In addition, the collaboratively created purchase list is passed to everyone, so that no item is missing. In Figure 2, the box with the dashed border on top shows the user interface for the friends who are assigned to the task. For them, not the process view is of interest but only the needed information to fulfil the task. In traditional WfMS this is the worklist component of the workflow system. The box shows all information to the assigned users that are linked with the task as input documents. The weather conditions and the temperature at the destination are shown, as well as a link to the purchase list and a download link to the Word document. It is combined with a description of the task containing the advice to watch the weather conditions. Figure 3 shows the part of the festival workflow after packing is completed. Then, the car is fuelled by Bob, who is the dedicated driver of the car. During fuelling, Bob records the current price for fuel per litre. The group drives to the festival location, attends the event, and finally they drive back home. When they arrive, Bob also enters the full distance in kilometres and then he completes the “Fuel and drive to destination” task. Subsequently, the automated task for the cost accounting (“Calculate individual costs”) is activated as well as the last human task for all workflow participants (“Upload photos”). The cost accounting divides the cost for fuel equally among the friends and allocates these costs with the bought items according to the persons who bought them, so that everyone has to pay the same amount. The “Upload photos” tasks offers a Dropbox link to the friends so that they can upload and share their eventful pictures. After the completion of the workflow, other features of the social workflow service show their advantages. Especially the features for sharing and reuse offer some additional subsequent opportunities for the community members: 4

Figure 2: A Social Workflow: Visiting “Rock am Ring” with friends. Part II.

• Paul, who attended a festival for the first time, finds the collaboratively created online purchase list very helpful. He copies this cloud document and stores the online link in his private library to reuses it for the next camping trip. • Sue reused the “Create Spotify playlist” task in a new social workflow for planning her next birthday party. • Peter was glad that the social workflow he designed simplified the whole trip. He decided to share the workflow among all other users of the social workflow service by granting public read access. • Tamara, a different member of the social workflow community, finds the public festival workflow, while she tries to model a workflow to attend a Spanish festival. She reuses this workflow and adapts it to her needs, for example, by adding a task for the online purchase of train tickets and by removing the tasks related to fuelling. This scenario demonstrates various benefits of social workflows compared to the use of unconnected services and plain Email communication: 5

Figure 3: A Social Workflow: Visiting “Rock am Ring” with friends. Part III.

• The execution of the various activities is automatically coordinated, thus everybody is always aware of what s/he needs to do and what the others have done already. • Relevant information is shared automatically with everyone who needs it (e.g. by providing a link to the ticket shop or the shared equipment list), thus many Emails are avoided. • Online services are automatically enacted, thus collected information can be automatically combined during whole course of the workflow (e.g. during playlist generation and cost accounting). • Reuse of workflows avoids forgetting important steps (e.g. when Tamara organizes her festival visit). Before we evaluate the concept of social workflows systematically by empirical studies, we now describe a new concept of a social workflow service that supports the just described scenario. 4. Concept for a Social Workflow Service We now briefly sketch the social workflow service CAKEFLOW, which we are currently developing. It is a specific instance of the Collaborative Agile Knowledge Engine (CAKE), which is a prototypical generic software system for integrated process and knowledge management, developed within various previous research projects addressing the needs of diverse application areas2 . CAKE integrates recent research results on agile workflows, case-based reasoning (CBR), and Web technologies into a common platform that can be configured to different application domains and needs. 2 See

http://cake.wi2.uni-trier.de

6

4.1. System Architecture Figure 4 shows the overall architecture of CAKE (and thus of CAKEFLOW) including the various components it consists of. The CAKE Agile Workflow Engine (see Section 4.2) is used for the enactment of social workflows and supports their collaborative modelling and adaptation in a consistent manner. CAKE supports process flexibility as already running workflow instances can be changed and placeholder tasks can be employed to support late modelling. During workflow enactment, certain tasks may have to be executed by services while other tasks may require an activity of human workflow participants. For workflow modelling, we developed a block-oriented graphical workflow modelling language further described Section 4.3. The CAKE Knowledge Engine supports process-oriented CBR and provides intelligent methods for reusing experiential knowledge. It supports similarity-based retrieval of workflows based on semantic descriptions as well as the automatic adaptation of workflows [6, 7]. The aim is to support inexperienced workflow modellers with gathered and structured knowledge from a repository. This experiential knowledge can then be discovered and reused in order to build or to improve a workflow, which is just designed. Section 4.4 explains this component in more detail. In CAKE, all resources can be managed and controlled by their users. The resource model implemented in the CAKE Storage Layer (see Section 4.5 ) ensures that any stored resource (e.g., a workflow, a task, a document, meta data) is accessible and possesses a clear ownership. The CAKE Interface Layer (see Section 4.6) provides communication means which encapsulate the functions of the two engines. It enables the development of complex web applications such as for workflow modelling and for controlling workflow execution and it supports their integration into a social network.

CAKE Client

User Administration

CAKE Server

CAKE Interface Layer

Workflow Modelling and Execution

Worklist

Knowledge Support for Modelling and Adaptation

Client Support (GWT-RPC, Servlets) CAKE Server API

CAKE Agile Workflow Engine Services in the Cloud

Service Connector

Agile Workflow Execution

Worklist Manager

CAKE Knowledge Engine Retrieval Engine

Adaptation Engine

Case Bases (Semantic Workflows)

Domain Ontologies & Similarity Models

CAKE Acquisition Layer

CAKE Storage Layer Access Control Document/ Data Store

Data Base

Worklist Access

Content Repository Documents

User Profile Access

Workflow Access

CAKE Store Workflows & Data Items User Profile & Worklists Figure 4: CAKE System Architecture.

7

Workflow extraction

Web

Ontology & Similarity Import

Ontology

Similarity

Web Content and Models

4.2. CAKE Agile Workflow Engine The agile workflow engine is used for the enactment of social workflows and supports their collaborative modelling and adaptation in a consistent manner. Agile means that workflows can be modelled and changed on demand at any time by any user who is granted the respective access right [8, 9, 10]. Workflow instances already running can be paused, not already executed parts of the workflow instance can be adapted, and the workflow execution can then resume while consistently considering the changes just made. Thus, the CAKE workflow engine implements flexibility by definition, flexibility by change, and flexibility by underspecification [1, 2]. The internal structure of this workflow engine is strongly tied to the reference architecture of the Workflow Management Coalition (WfMC) [11]. Thus, the workflow engine provides interfaces for modelling and execution of workflows, for invoking applications (service connector), and an interface for the delegation of tasks to humans (worklist manager). During workflow enactment, certain tasks may have to be executed by services (e.g. to check the weather conditions) while other tasks may require an activity of human workflow participants (e.g. friends in the social network). The agile workflow engine ensures that a workflow including human and service tasks can always be enacted, because the internal, executable workflow representation follows the correctness-by-construction principle [9]. Without this principle, inconsistent workflows could occur which would require more effort for testing and correction. For the synchronization of collaborative modelling and execution of workflows, CAKE has a real-time push-event system and an internal synchronization that ensures workflow consistency [12]. In the festival scenario previously described, the workflow engine controls the initial modelling of the workflow by Peter as well as its execution. In particular it pushes active tasks on the respective worklists (see dashed box in Figure 2) of the friends, detects the termination of each task and starts each new task. It also controls the execution of the automatic tasks, such as the weather service. For this purpose, the service connector implements specific application-programming interfaces for each external service available. The agility of the workflow engine is an important feature, if some unexpected event occurs during workflow execution. For example, assume that Sue is having an accident happening just before she has started to buy the required items for the festival. It turns out that she needs a wheelchair for some time, but she still wants to attend the festival. Thus, the friends need a larger car, which none of them owns, and they need to make a reservation for space on the campground for handicapped people. Peter uses the social workflow service to temporarily stop the running workflow instance, which allows him to edit it. First, he reassigns Sue’s purchasing task to himself. Then, he inserts tasks for “Making a rental car reservation” and for “Making a reservation for the campground” as two additional parallel branches in the workflow fragment shown in Figure 1. Finally, he inserts the task “Pick-up the rental car” as additional task prior to the fueling task in Figure 3. Then the workflow resumes and is now suited to appropriately address the new circumstances. 4.3. CAKE Workflow Modelling Language The application of web technologies enables the development of modern user interfaces independent of the system platform or device. It allows the development of user interfaces which can be easily adapted to new requirements and opportunities offered by mobile devices. For modelling social workflows, we developed a graphical workflow modelling language we call ‘CAKEFLOW Cloud Notation’ (CFCN) which is derived from UML activity diagrams. The example workflows shown in Figure 1 to Figure 3 are depicted in CFCN. In CFCN one can model workflow tasks with control flow as well as data objects (e.g. documents, URIs, numerical and boolean data) with the involved dataflow. It supports sequential, parallel, and conditional execution of tasks, as well as loops. The control flow structures may be nested but not interleaved like it is possible with graph-oriented workflow languages like BPMN. As a specific instrument for implementing agility, we introduced a stop sign icon (i.e., a kind of break point) that can be placed into the control flow of a running workflow instance to pause workflow execution in case the stop sign is reached [13]. Placing the stop sign enables the adaptation of the subsequent sub-workflow. Figure 5 shows a sketch of the three control flow structures (parallel execution (A), conditional execution (B), loops (C)) we support. Although we only replaced the original UML symbols, we think it will receive more acceptance by business and private users. This assumption is based on current Human Computer Interaction (HCI) research where “enchantment” is an important aspect and beauty is part of enchantment in using UIs [14]. Furthermore, many HCI researchers state that the visual appeal influences factors as (perceived) reliability, usability, information quality, trustworthiness and usefulness [15]. An important aspect of the general block-oriented nature of UML activity diagrams is that the correctness-by-construction principle holds and consistency checks during changes of the dataflow can be performed based on the connected control flow activities (tasks). 8

Figure 5: Some Control-Flow Elements.

4.4. CAKE Knowledge Engine for Process-Oriented Case-Based Reasoning Workflow modelling usually requires significant skills and experience in the respective domain (e.g. which steps are important when moving to a new city) and in modelling principles in general. Additionally, when performed using some modelling environment, skills in using this environment and knowledge about the available services that can be integrated into a workflow are important as well. As far as workflow quality is concerned, a well-designed workflow is hard to produce by a person with no or little experience in workflow modelling. Hence, workflow modelling from scratch is tedious and will probably not be accepted by private individuals who usually do not benefit from a significant number of repetitive executions of one and the same workflow, like companies do. To address this problem and to increase the acceptance by new users, we propose a process-oriented CBR approach to support the reuse of previously modelled shared workflows[16]. Case-based rasoning is an established Artificial Intelligence methodology to problem solving based on the assumption that similar problems have similar solutions [17, 18, 19]. Experience items (called cases) capture solutions to previous problems and are collected and stored in a case base, i.e., a repository of successful problem solving episodes. When a new problem must be solved, cases addressing similar problems are retrieved from the case base and their solution is adapted to become a solution of the new problem. Thus, problem solving from scratch is avoided. General problem solving knowledge is replaced by a collection of concrete cases, which will grow over time and thereby improve the problem solving performance. Process-oriented CBR systems are particularly tailored to support the creation and adaptation of cases in the form of workflows [20, 21, 22, 23, 24]. By applying process-oriented CBR to the social workflows, users should be relieved from the burden to develop workflows from scratch. A community of users, with a similar mind-set, shall be supported in collecting their workflows in a case repository. This repository is the collective experiential knowledge in social workflows of this user community, which is maintained by the CAKE system. The content of this repository is shared among the users and the resource model controls the access [3]. Thus, the available workflow repository for each user is a subset of all existing workflows to which she/he has access to. Further, every resource in the CAKE resource model [3] (particularly every workflow and every task) can be tagged with annotations, which leads to an enrichment with semantic descriptions. Such social tagging can produce a social workflow folksonomy [25] for user community. Once the bootstrapping of such a workflow repository (e.g. through content provided by very committed key users) succeeds, less experienced users may obtain their own social workflows by selecting an appropriate workflow from the repository and modifying it if necessary. For navigating in the workflow repository, we envision two options: • The users may search for an appropriate workflow by using the tags from the folksonomy. The retrieval engine then proposes workflows with similar tags from the repository for reuse. • Alternatively, users may start describing their problem by creating a brief sketch of the workflow (e.g. a sequence of a few tasks) they are interested in, thereby specifying already some structure of the intended solution. 9

The case retrieval component will then propose workflows that contain a sub-workflow similar to what the user has specified [20]. If the user is satisfied with the found workflow, she/he can transfer a copy of it to her/his personal workspace in the social workflow service and prepare it for execution. After all, if necessary, the user can also manually adapt the current workflow by editing it via the workflow editor. As a step ahead of this manual adaptation, we also envision an automatic adaptation of workflows supported by workflow adaptation cases [21]. In the festival scenario, Peter uses the knowledge engine when initially constructing the festival workflow. He uses the retrieval engine to find workflows from others who have already attended a festival. In the course of modelling of his festival workflow, he eventually uses the retrieval function to look for workflows with a structure similar to his current workflow sketch. He might find a workflow from a group of people going together to Disneyland over a weekend and discovers the automatic cost accounting service and recognizes how he can integrate this service into his workflow. Also Tamara uses the knowledge engine to find Peters festival workflow as a source for planning her Spanish festival attendance. 4.5. Storage Layer Integrating new resources, such as workflows into a social network requires a dedicated model of access control for workflows. This is particularly important because workflow sharing requires that users are enabled to actively control the access rights to their social workflows. In the proposed social workflow service, all resources can be managed and controlled by their users. The integrated resource model for collaboration and reuse [3] is part of the storage layer. This layer has to ensure that any stored resource (a workflow, a task, a document, and any further workflow related resource) is accessible and possesses a clear ownership. In principle, the access control mechanism is a decentralized Discretionary Access Control [26] (DAC) with subject-object relationships specified in Access Control Lists (ACLs). Setting the access rights on a file for a certain user or group is a well-known example of a normal DAC. In this context decentralized means that a user can transfer access rights to another subject or a community. The basic idea is that every resource (a workflow, a task, a document, and any further resources) in the system has a dedicated owner who is allowed to manage the access rights for the resource. The access control mechanism of CAKE is a workflow specific concept. It has been developed with the intention in mind that everyone can register easily with the CAKE system. Therefore a person can also register with her/his Facebook account and integrate her/his list of friends for the execution of tasks in social workflows. In the festival scenario, the storage layer controls access to the festival workflow as well as to the exchanged documents, e.g. the purchase list. While Peter as the workflow owner retains the modification permission for the whole workflow, the other friends only receive the right to execute the tasks they are assigned to. Finally, Peter grants a read access right to the workflow definition (not the particular instance) to the public. Thus other users, such as Tamara, can read and copy this workflow into their own workspace and use it for their own purposes. 4.6. Interface Layer and Clients The CAKE Interface Layer provides various APIs for the communication with different kinds clients for workflow modelling, sharing, adaptation, and execution. It includes a push event mechanism which is linked to the workflow engine in order to propagate all changes on workflows to the affected clients and thus enabling collaborative workflow modelling and execution in real-time [12]. The overall CAKE software is implemented in JAVA as Web-based system. The client user interfaces are implemented using the Google Web Toolkit (GWT)3 , enabling access to all workflow related functions such as workflow modelling, execution, similarity-based retrieval, and adaptation using a standard browser (see Figure 6). Further, the CAKE Server API also allows mobile applications to directly connect to CAKE, e.g., to support the mobile Apps. 3 http://www.gwtproject.org/

10

Figure 6: Browser-Based Access to CAKEFLOW.

5. Related Work Till today, the terms “social workflow” and “social workflow service” are not defined in the literature. An online search using Google4 for these terms turns out that they are not frequently used. The idea of social workflows, presented here, is closely related to different research areas and application domains and therefore these different views are regarded. This consideration serves on the one hand as delimitation of the term ‘social workflows’ from ‘personal’ or ‘private workflow’. On the other hand it shows the intersections of this work with different related research areas. The idea to employ workflows in a non-business context is not a new one. Evidence is given in research as well as in commercial systems. These approaches will be presented and their differences are discussed. The focus of this work lies on the creation, management and reuse of procedural knowledge for private individuals and communities. In this context, research has just started to describe Social BPM (S-BPM) which tries to merge traditional BPM with collaborative social software. Hence, related work in S-BPM and workflow systems and management is regarded first. 5.1. Related Research 5.1.1. Social Business Process Management Along with the development of the idea of “Social Business” or ”Business 2.0”, companies start to develop social business software which integrates social aspects. The aim is to foster flat hierarchies and a stronger collaboration across all levels in a company, e.g. by means of communication tools and common social software - lend from the Web 2.0 - like wikis. This is an actual topic in BPM research [27] where social BPM (S-BPM) is sometimes called 4 The term “social workflow” resulted in 7.780 result items and “social workflow service” in 8 items. The search was performed in December 2014.

11

Subject-Oriented Business Process Management or Social Business Process Management. In the end S-BPM means that business process management itself becomes a collaborative process, including all stakeholders (employees, suppliers, partners, etc.). The knowledge and the experience of the stakeholders shall contribute to the overall quality of a business process. Two examples of S-BPM research are now considered. In 2010, Rito-Silva et al. [28] propose a human-centred approach following the principles of agile software development, called AGILIPO. Their research aim is to define an agile business process methodology that shall overcome weaknesses of traditional BPM, especially the lack of stakeholder involvement. By including different stakeholder groups different perspectives can influence the process design. Managers have an abstract top-level view while normal employees have a rich detailed view on the same processes. Their idea is to empower all community members to participate in the design of new processes. This way, the delay between process design and enactment shall be shortened because organizations and business environments often change faster than the available processes. Agility in AGILIPO is performed by allowing under-specified process models with “generic” activities which can be specified during the run-time of an instance. When an activity has been completed, its executor (workflow participant) may annotate the activity with tags to reuse the instance in a next evolution of this process. AGILIPO promotes a social software environment in which the folksonomy of tagged activities can be discussed and rated. So far, Rito-Silva et al. [28] only sketch the idea of AGILIPO but an implementation of the proposed features has not been described. In 2012, Sch¨onthaler et al. [29] presented the Horus methodology. Horus provides comprehensive support for the BPM life cycle including all stakeholders of the business community. A core component of Horus is the “Social BPM Lab” [30] that strives for a re-thinking how BPM is done. During a business process many stakeholders are involved in strategic, tactical, and operational issues, sharing their knowledge in a networked structure. Horus fosters knowledge management via an integrated social business network so that all involved people can work collaboratively. Horus is a mature BPM tool that allows to model, simulate, and execute workflows in a cloud-based environment. However, knowledge management in Horus does not address process reuse and learning from experience of others. 5.1.2. Workflow Management Systems In a business context, workflow management systems are employed for the creation, management, and reuse of business processes. One of the first articles elaborating the idea that individuals have private workflows is written by Gary et al. (1997) [31]. They state that organizational workflows are overlapping with the workspace of an employee and with the implicit personal workflows of a person. They claim that these personal workflows are little regarded considering the value for the organization. Therefore they want to enable employees to develop their own productive work practices in support of an organization’s processes. They postulate a more integrated view for the processes of the organization and the personal workflows of the employees. In 2002, Keller et al. [32] described a personal workflow system. Here a personal workflow management system is a means for an individual to organize the participation in multiple projects. When working in different projects, a person may be confronted with the problem of having to perform several tasks concurrently, while she/he might only work on one task at a time. Social workflows support private individuals in a similar way as they can organize tasks executed in parallel and in a sequential order. Parallel tasks can occur if a person pursues several complex goals within one or several social workflows. In contrast to the works above, Hwang and Chen (2003) [33] define personal processes as processes that constitute personal tasks and data for private individuals. A personal workflow management system (PWFMS) is a system that coordinates personal processes. Their work is the first, we know of, that explicitly shifts the use of a workflow management system to a non-profit private scenario. In contrast to a traditional workflow management system, which enforces task execution, the main objective of their PWFMS is to remind or provide suggestions to a mobile private individuals. In the evaluation of their work Hwang and Chen describe a simple data-driven workflow system to execute personal workflows. Their PWFMS only regards individuals and how they can be supported in their everyday life. Therefore groups or collaboration within communities are not within the scope of their work. The work of Hwang and Chen is partly closer related to pervasive/ubiquitous computing than to workflow management because not the structure of workflows is in the focus but the idea of a pervasive/ubiquitous computing environment which is supported by a workflow system.

12

5.1.3. Pervasive and Ubiquitous Computing Mark Weiser coined the term ‘ubiquitous computing’ in 1991 [34]. He envisioned a scenario in which the environment proactively interacts with a human and the presence of computers becomes invisible. Pervasive computing is more pragmatic and does not try to hide the presence of computers but rather tries to embed technology or intelligence in small and easy-to-use devices, like smartphones. Therefore the term pervasive computing is more often used by the industry [35]. Taking into account these aspects of pervasive computing, it is reasonable that current research on pervasive/ubiquitous computing are up to further include social interaction in their systems [36, 37, 38]. Ben Mokhtar and Capra (2009) [36] argue that pervasive computing is unavoidably moving towards “Pervasive Social Computing”, with the pervasiveness of handheld devices and the popularity of social network sites. For them, pervasive social computing aims to take advantage of human social relationships, expressed as social networks, to enable the fulfilment of private user’s tasks. Their aim is to assist private individuals in realizing their daily tasks by reasoning about their social links and recommending them to other users that share similar interests. The pursuit of complex private goals is not considered and so there is no flow of activities and there are no dependencies among them. In 2004 Ranganathan and McFaddin [39] stated that private individuals perform different kinds of tasks in order to achieve their goals. They point out that one problem in pervasive computing environments is the discovery and coordination of different web services for achieving the users goals. They conceive pervasive computing environments in which persons walk into and the environment interacts with the persons to attain their goals. The environment will have a number of web-services performing various tasks like providing location or map information, giving product or exhibit information, receiving payment, etc. Their approach of private workflows aims at E-commerce scenarios which use predefined knowledge implemented in web services to guide a customer. Montagut and Molva (2005) [40] suggest a distributed/pervasive computation paradigm in which mobile users do not act merely as clients but can also actively participate in the execution of the overall services. They outline technical issues for such a pervasive execution environment like overall execution control and dynamic task assignment. However, their approach omits synchronization issues for parallel executed tasks and constraints for the task execution for reasons of simplicity. The pervasive execution environment has a distributed control, so no single device is in charge of managing the workflow execution and the overall control is assured through the collaboration of the devices involved in the execution of the workflow. Although the authors consider collaboration, no human activities are regarded in their work. An extension for their approach could be WS-BPEL4People [41] which bridges the gap between BPEL workflows and the interaction with humans. Kramer et al. (2007) [42] developed a context-driven mobile tourist guide. They pursue the idea to have a local guide, who understands the individual interests and timeframe, knows the local situation and gives a personal tour to each tourist. Their system is able to interpret personal interests according to a user profile and has a dynamic knowledge base which uses web services to collect information about sights and places. If a user wants the system to compute a personal tour, the system computes a sequence of attractive places to be visited according to the user’s interests, the current GPS position, and a given timeframe. Their approach is similar to the Personal Workflows of Hwang and Chen (2003) [33] because the flow of activities is dynamically computed by provided mobile data. However, the scope of this work is limited to travels and tour guides. A further interesting aspect of their work is the evidence that concrete interests or goals of a private individuals in the scenario of travels are not predetermined and that tourists need to be inspired to get aware of their own interests. Urbanski et al. (2009) [43] propose a distributed workflow system with dynamic, adaptable processes to enact personal workflows (PerFlows) for private individuals. The concept of the PerFlows system is built upon the described scenario of Mark Weiser. A PerFlow allows specifying conditions, e.g., based on location that cause tasks to start, to skip or to abort. Like in many of the above described approaches, the data that is the cause for the execution of the workflow is delivered by mobile devices. PerFlows provides the infrastructure with additional information about the context which triggers the individual tasks. However, the focus of PerFlows is not the procedural knowledge of workflows, but the creation of an ubiquitous environment backed by a distributed workflow system. This can also be seen in the granularity of the tasks in the example workflow, like ‘Make Coffee’ or ‘Go Home’, which cannot be considered as a complex goal or activity. PerFlows focusses on supporting individuals and not communities or groups. Hence every user needs to model her/his own workflows as collaboration is not considered. 13

5.1.4. Personal Information and Knowledge Management Another research field of interest is Personal Information Management (PIM) and Personal Knowledge Management (PKM). This is evident, because PIM aims to support people to have the right information at the right time, in the right place, in the right form, to perform their current activity [44], whereas PKM is a collection of processes that a person uses to gather, classify, store, search, retrieve, and share knowledge in her/his daily activities [45]. PKM is more about E-Learning and about taking personal responsibility in an organization. Hence, a person is enabled to work better with knowledge [46]. It is a bottom-up approach to traditional knowledge management directed at the needs of individual employees [47]. Surprisingly, in none of these two research fields, related work could be found which is applied in a non-profit scenario using procedural knowledge. 5.2. Non-Technical Systems Detached from a certain research area, Dorn et al. (2012) [48] describe flexible social workflows. They address collaborative problem solving and its integration with process-support systems as an architecture comprising human components and connectors. They state that human collaboration consists of human components and human connectors, e.g. a secretary (connector) might respond on behalf of the principal (component). With this assumption they model a process which reflects three collaboration patterns (social networking, wiki creation, and crowdsourcing) based on Web 2.0 applications in order to pursue a complex goal (Writing a scientific project proposal). However, the authors do not propose an execution service or an executable workflow representation. The procedural knowledge of a social workflow is not regarded but only the collaboration within a dedicated group of persons to pursue a goal. 5.3. Workflow Communities The current state of research has shown that the idea of social workflows, as stated in this work, has not been pursued yet. Of course, there are intersections and similarities to different works and research areas. Especially, knowledge sharing and the idea of building a tied community, linked by their interests, were left beyond. In the following, three selected applications are described which haven proven a successful implementation of these aspects concerning scientific workflows. The Taverna website5 says that “Scientific workflows are widely recognised as a useful paradigm to describe, manage, and share complex scientific analyses”. They are also a means to provide unambiguous outcomes of experiments because the whole experimental setup is repeatable and the workflow serves as execution protocol [49]. The following three systems show that there is an interest to collect and organize procedural knowledge within a community in a non-business context. The WINGS6 (Workflow Instance Generation and Specialization) [5] platform is used to create and execute workflows which represent computational experiments. It was initially developed to create large-scale workflows for physics-based simulations to construct seismic hazard maps for the prediction of earthquakes. Additionally, it has been used in several scientific applications including biomedical image processing and genomics. WINGS aims at communities of researchers, so they can use, create, and execute data mining workflows. To assist its users, WINGS uses Artificial Intelligence (AI) planning and semantic reasoners to support users in creating workflows while validating that the software components are properly used and the requirements are fulfilled to execute the components. Besides that, WINGS supports its community in sharing and reuse of their workflows. Therefore, it implements a discovery and exploration service to find suitable experiments. ClowdFlows7 [50] is a cloud-based workflow management system that supports the creation, execution, and sharing of data mining workflows among a community. It is built as a service-oriented architecture that allows integrating third-party services in workflows at runtime. The service discovery of ClowdFlows offers a wide range of data mining and utility components for data processing. All user interfaces are web based and also intended to run on browsers of mobile devices. As a social aspect, every registered user may share and reuse workflows of other users by copying a public workflow in her/his private repository. 5 http://www.taverna.org.uk/introduction/why-use-workflows/ 6 http://wings-workflows.org/ 7 http://www.clowdflows.com/

14

MyExperiment8 [51] is a social networking environment for sharing research objects and especially scientific workflows. It offers a collection of scientific workflows which consists most notably of Taverna9 workflows, which are executable in the Taverna workflow workbench. Taverna was designed from the outset to assist bioinformaticians on their laboratory work. The Taverna workflow system should release bioinformaticians from the burden to repeatedly perform tedious tasks of plumbing together data resources and analytical tools. Nevertheless, the MyExperiment environment supports workflows of various workflow engines. For the community of MyExperiment the workflow execution engine does not matter but the knowledge in the workflows and the used services. For this reason, it offers a market place where users can share, reuse, and repurpose scientific workflows to avoid unnecessary reinvention. To foster collaboration and the community feeling it also integrates communication means to enable users in discussing workflows and their related scientific objects. 5.4. Summary and Classification of Related Reseach In the following, the different research approaches are visualized in two spider chart diagrams. These charts are a subjective rating on how far the described approaches cover the idea of social workflows addressed in this paper and are a means to show intersections between the different approaches. For the rating, additional information have been considered that can be checked in the given references. This includes formative evaluations or realized implementations of the postulated approach. The relationship of an approach to our work is assessed by six aspects. The first aspect is called automatic execution. It shows the focus of the authors considering the automatic execution of workflows and how far their systems or approaches are developed to enable an automatic execution. The second aspect process-orientation shows how much the presented work focuses on a procedural execution of tasks. The third aspect collaborative work expresses the degree of human involvement in the execution and creation of workflows. The forth aspect knowledge sharing illustrates if reuse and sharing of procedural knowledge is regarded in an approach. The fifth aspect community involvement displays if a certain or several communities shall be involved and supported by the approach. The last aspect private context shows if the approach aims to be applied in a private environment opposed to an business environment. automatic execution

process-oriented

private context

Gary et al. Keller et al. Hwang et al. Rito-Silva et al.

community involvement

collaborative work

Schönthaler et al.

knowledge sharing Figure 7: Comparison of Process-Oriented Systems.

Figure 7 summarizes the introduced areas of related work in workflow management systems and S-BPM. It can be seen that community involvement and knowledge sharing are only in the focus of the S-BPM approaches of Sch¨onthaler et al. [29] and Rito-Silva et al. [28]. Sch¨onthaler et al. got a higher ranking in both categories because they have a working implementation while Rito-Silva et al. [28] only sketch their approach. Sch¨onthaler et al. [29] pursue knowledge sharing in their approach but it does not aim at the reuse of procedural knowledge but at providing 8 http://www.myexperiment.org/ 9 http://www.taverna.org.uk/

15

additional social software like a social network to their BPM tool. Mainly, these papers focus on the automatic execution of modelled workflows and mostly regard individuals in a corporate organization. Only the work of Hwang et al. [33] completely focusses on a private context. Figure 8 compares the described approaches in pervasive computing including the idea of Dorn et al. [48]. It is obvious that research on pervasive computing in general is less process-oriented although Montagut et al. and Ranganathan et al. prove that process-orientation can be advantageous for pervasive systems as they use BPEL in their approaches. automatic execution Mokhtar et al. private context

process-oriented

Ranganathan et al. Montagut et al. Kramer et al.

community involvement

collaborative work

Urbanski et al. Dorn et al.

knowledge sharing Figure 8: Comparison of Pervasive Computing Systems.

Pervasive computing has a broad scope of application and also takes private application scenarios into account. As a consequence some approaches like Mokhtar et al. [36] also regard whole communities and its members for assistance. The article of Urbanski et al. [43] on personal workflows (PerFlows) is strongly related to social workflows, although it does not cover all aspects, because the process-oriented view as well as the execution of workflows in a private context is considered as very important. Emphasis should be put on the theoretical article of Dorn et al. [48] and the work of Mokhtar et al. [36] as they are the only ones who regard collaborative work and the involvement of a community as valuable. Unfortunately, none of the works considers sharing or reuse of procedural knowledge. 6. Potential Study In order to get first insights in the potential acceptability of a social workflow service as part of a social network and as a means to define a road map for the integration with existing services, a formative evaluation of the overall concept just described has been performed. 6.1. Exploratory Study First, an exploratory study was conducted, considering the questions which online services for communication and collaboration are used by private individuals today and in which application scenarios a social workflow service could provide a high benefit to the users. Then, an additional online study was conducted (see Section 6.2) to investigate the hypotheses based on three very specific application scenarios derived from the exploratory study. 6.1.1. Hypotheses of the Exploratory Study For the exploratory study, three main hypotheses have been identified which are essential to further guide the development of a social workflow service. The first hypothesis relates to the overall assumption behind the idea of social workflows. The hypothesis H1 is that persons today already use existing Internet services to perform complex goals that require a structured flow of activities and that these activities may involve personal friends as well as professionals. Only if this hypothesis holds, a social workflow service can provide a benefit to the user. 16

The second hypothesis addresses a social workflow service in general. The hypothesis H2 is that there are application scenarios in which social workflow services can provide a benefit to the user. Additionally, we want to know from the experts participating in the study what such scenarios could be and which online services are considered most relevant. The third hypothesis relates to our specific concept of a social workflow service. The hypothesis H3 is that the following five core features of the CAKE system are relevant, i.e., the ability 1. 2. 3. 4. 5.

to discover new workflows, to adapt workflows, to invite new friends as workflow participants, to automatically execute a workflow, and to share workflows.

6.1.2. Method of the Exploratory Study The topic of social workflows and social workflow services is quite novel and hardly investigated. For this reason the study was created as an exploratory study. As instrument the Delphi method was used [52]. In its core, this method constitutes of a turn-based, structured expert consultation. It can be used to assess future technological trends. The characteristic of this method is that the given answers are reviewed after each iteration and the experts are informed about the result. With this feedback included, the next iteration of the survey starts. However, the Delphi method leaves much scope considering the procedure. So, the quantity of asked experts is not determined by the Delphi method as well as the number of iterations [52]. The setting of these parameters should be oriented towards the problem with the respective application domain and research goals in mind. This exploratory study started with 20 experts and was limited to two iterations. Because one purpose of this study was the identification of potential application domains for social workflows, the asked experts belonged to two different groups. The first group of 10 experts consisted of people with academic experience on process management. It was assumed that the academic expert knowledge of this group would help to identify trends. The second group consisted of 10 persons with a high technical appeal, who are heavily using social media and web technologies. These people had no or little expertise on process management. The assumption for this group was that they could provide a non-biased view on technology. Of these 20 experts five were female and 15 male. The first iteration of the survey was finished by 14 experts, while the second iteration was finished by 13 experts, still containing four female experts. Both iterations were performed as an anonymized online survey. 6.1.3. Structure of the Exploratory Study The purpose of a Delphi study is the identification of trends and ideas. For this reason, it is not mandatory to reach a consensus among the experts about a particular fact. A challenge of this study10 was questioning experts about a theme that does not yet exist and to bring them close to the idea of a social workflow service. Therefore, the study was divided in two rounds: 1. An open questioning to obtain new ideas. 2. A closed questioning to concretize the obtained ideas and to rate them. The fact that social workflows and social workflow services are hardly known and little researched, made it necessary to collect as many ideas as possible in an open questioning in order to concretize these ideas for the next iteration. Another reason for the initial open questioning was the intention to discover unpredictable and unforeseen ideas. In the second iteration, the prepared results of the first iteration were presented to the remaining 14 experts for evaluation. For the preparation of the results, duplicate ideas were removed and the ideas were aggregated towards their core approach. The aim of the second iteration was the identification of the usefulness of the ideas. Because of the second group of experts with no or little expertise on process management, the formulation of the open questions regarding potential application scenarios avoids technical terms related to workflow technology. 10 The questionaire is available in german. For further information the reader is referred to http://www.wi2.uni-trier.de/shared/ technical/tp_pwsn12.pdf. It also contains the absolute number of votes for each rated result.

17

Instead, we asked the experts for which purposes, requiring complex sequences of actions, they use Internet services today. An example was given to demonstrate the kind of structured activities we were looking for. Further, we also asked the experts to name the Internet services they use today and whether other persons are involved in the activities as well. The questionnaire of the first iteration ended with a question about the estimated benefit of a social workflow service. As background information regarding this question, a general explanation of the concept of a social workflow service was provided. The experts had to rate the estimated benefit of such a service on a five point ordinal Likert scale (very high benefit, high benefit, modest benefit, little benefit, no benefit). The second iteration started with a detailed use case of planning a vacation trip by support of a social workflow service. This scenario was chosen because it was most frequently named by the experts in the first iteration. The use case covers the five technical features of the CAKE system described in hypothesis H3 of Section 6.1.1. Each technical feature was then individually rated by the experts on a four point ordinal Likert scale (very important, important, less important and useless). The second question was focussed around a description of a concrete social workflow for planning a trip, containing dedicated manual and automated tasks. The experts had to rate each task on a four point ordinal Likert scale. In a third follow-up question the experts were asked for additional important tasks that should be considered for planning a trip. The fourth question introduces a new scenario, namely moving to a new apartment. Again, a social workflow is described containing manual and automated tasks. This scenario was chosen because it was mentioned with the second highest frequency in the first iteration. The experts had to rate each task for this social workflow on a four point ordinal Likert scale. Again, the fifth question was a follow-up question asking for additional relevant tasks. In the sixth question, the experts were asked to rate the potential of eight concrete application scenarios for social workflows mentioned in the experts’ answers obtained in the first iteration using a five point ordinal Likert scale. Finally, the last question asks for any further application scenarios for social workflows that the experts could think of. 6.1.4. Results of the Exploratory Study The open questioning of the first iteration showed that the experts mainly use the Internet for communication, information search, amusement, buying and selling of items, planning of trips, planning of events, data storage, and for job and flat finding. To pursuit complex goals the experts involved friends, colleagues, unknown experts (e.g. members of forums), or family members in their planning process. As means of communication the experts used social networks (Facebook, Google+), instant messaging (ICQ, WhatsApp, Skype), and e-mail communication. The last question of the first iteration on the estimated potential for a social workflow service showed no clear result. Eight experts remained neutral in their opinion, four experts attested a high benefit, while two experts only see a low benefit. One reason for the conservative assessment of the potential might be the fact that there are no working social workflow services today and that it is very difficult to assess a potential system without previous practical experience with it. Table 1: Rating of Technologic Features. Scenario: Vacation Trip

Social Workflow Service Feature

Relevance Scoring Important Less Imporor very tant or UseImportant less

Discover workflows Adapt workflows Invite workflow participants Execute workflows Share workflows

76.9% 76.9% 53.8% 92.3% 69.2%

23.1% 23.1% 46.2% 7.7% 30.8%

The detailed description of the technical features of a social workflow service was given in the second iteration of the study by means of the described use case of a vacation trip. The results for the rating of the technical features is presented in Table 1. Besides the invitation of workflow participants, all technical features have been mainly seen 18

Table 2: Rating for Tasks. Scenario: Vacation Trip

Tasks for a Social Workflow of a Vacation Trip

Relevance Scoring Important Less Imporor very tant or UseImportant less

Inform about destination Check weather on destination Inform about accommodation Compare price for accommodation Book accommodation Compare price for rental car Book rental car Plan the route Online check-in Online visa request

84.6% 46.2% 100% 92.3% 100% 61.5% 69.2% 69.2% 84.6% 76.9%

15.4% 53.8% 0% 7.7% 0% 38.5% 30.8% 30.8% 15.4% 23.1%

as important or very important by the experts. This result is not in complete accordance with the result of the first iteration, where the experts clearly stated that they invite friends, colleagues, or family members in their planning process. The result for the second question concerning concrete tasks in a social workflow for a vacation trip is shown in Table 2. Additionally, the following tasks were mentioned as answers to the third question: Prepare first-aid kit, plan immunization, plan baggage, and inform about currency rate. Table 3: Rating for Tasks. Scenario: Apartment Moving

Tasks for a Social Workflow of a Apartment Moving

Relevance Scoring Important Less Imporor very tant or UseImportant less

Search new apartment Rent a transporter Engage moving company Register new address Ask persons for help Search follow-up tenant Modify subscriptions Register new license plate

92.3% 76.9% 38.5% 69.2% 76.9% 76.9% 76.9% 61.5%

7.7% 23.1% 53.8% 30.8% 23.1% 23.1% 23.1% 38.5%

The result for the fourth question concerning concrete tasks in a social workflow for an apartment moving is depicted in Table 3. In the fifth question one expert added the apartment check-out as additional, possible task for this social workflow. Based on the given answers of the first iteration the sixth question aimed at a rating for eight application domains and their estimated potential in a social workflow service. The rating of the experts is illustrated in Table 4. 6.1.5. Interpretation of the Results The exploratory study basically confirms all three hypotheses. Hypothesis H1 was clearly confirmed as the experts mentioned several scenarios of Internet use that require a complex sequence of actions and for which they use several existing Internet services. The answers also showed that they involve different persons or organizations in their social workflows. Hypothesis H2 could neither be confirmed nor rejected after the first iteration of the study. However, with the more detailed description of the use case scenarios provided in the second iteration, this hypothesis can 19

Table 4: Application Domains and their Estimated Potential

Application Domain

Planning of vacation trips Job search Planning of city trips/short trips Search for apartment Buying/Selling of cars Planning events Buying/Selling of clothes Planning of craft activities

(Very) High Benefit

Relevance Scoring Modest Little and Benefit No Benefit

76.9% 46.2% 46.2%

15.4% 23.1% 53.9%

7.7% 30.8% 0%

53.9% 30.8% 53.9% 7.7% 46.2%

30.8% 30.8% 46.2% 30.8% 53.9%

15.7% 38.5% 0% 61.5% 0%

be confirmed. It has been shown that the organization of vacation trips and city trips/short trips, events and craft activities have been rated with the highest benefit by the experts. All these application scenarios describe typical social workflows with distinct activities which have to be executed in a distinct order while involving services and various other persons and organizations. Hypothesis H3 is confirmed given the relevance scores shown in Table 1. Only the relevance for the feature to invite friends to participate in a social workflow is rated not as high as the relevance for the other features. This could be an indicator that the communication between individuals during the execution of a human task is not considered an essential feature of a social workflow. Another explanation for this result could be that the collaborative aspects of social workflows were perhaps not clearly described in the study. It was an exploratory study that had to transfer the idea of social workflows to a group of experts. Hence, the experts in the study might perceive the concept in a way that suggests that all tasks have to be done without the support of other persons. 6.2. Extended Online Study On the basis of the first exploratory study a second online study was conducted. The aim of the online study was to find out if real elaborated social workflows gain acceptance in a group of persons not involving process management experts. Additionally, it was investigated what kind of data and services are desired by the users of a social workflow service. For this reason, three concrete social workflows were designed with different focus for the creation of benefit for private individuals. In the following the method, the scenarios, and the findings of the extended online study are described. 6.2.1. Method of the Online Study For the online study three concrete social workflows were designed. The selected application areas are based on the relevance scorings of the exploratory study (see Table 4). The first scenario corresponds with the described scenario in Section 3 and is about the attendance of a rock festival. In contrast to the CFCN illustration of the scenario shown in this paper the participants of the study were not confronted with process models, but with the textual description of the corresponding model. The second social workflow was about searching an apartment in a new city. The third workflow was about moving to a new city. The online study was conducted within a period of one month. The questionnaire on the first scenario was answered by 20 persons, the second scenario by 19 persons, and the third scenario by 15 persons. The online study was spread by a mailing list mainly to students of business information systems at the University of Trier and via Facebook. Hence it is a qualitative study with a snowball sampling. Each social workflow was addressed in a separate

20

survey. Because of the extent of each survey, it was not supposed that a person answers all three surveys11 . The main questions in each questionnaire concerned the assessment of the benefit of the proposed features of the social workflow and the overall rating for the described scenario. The respondents were given a six point Likert-Scale (1 ‘no use’ to 6 - ‘very high benefit’) to answer the questions. 6.2.2. Scenarios of the Online Study In the following the scenarios are described. The description is shortened, omitting the details of the online services and excessive descriptions for the respondents. The respondents were first introduced to the idea of social workflows and the possibilities of a social workflow service as mentioned before in this paper. Attend a Festival. Due to the detailed description of this scenario in Section 3 only the postulated features for creating benefit are summarized here. Focus. This social workflow focused on the collaboration between friends to attend a festival. Thereby, the workflow used third-party services like Spotify, Dropbox, or Google to support the friends. Moreover, the festival workflow contained a shared list which was used for packing and for the integrated cost calculation. Search an Apartment. Starting point for this workflow is a person, who needs an apartment in a new city or in the near location of it. The workflow starts with a task that asks for the preferences of the person. Criteria may be the desired size of the department, the price, the number of rooms, and the location. When the person has entered her/his preferences the information is passed to an automated task which executes an online search. This automated task considers real estate columns of local newspapers (according to the desired location) as well as national newspapers and online portals. This is a relief for the person because the selection of a useful real estate column requires regional experiential knowledge, because most sources for local real estate advertisements are not known to people outside the particular city. After reading and parsing all sources, the automated task creates an apartment list which aggregates all information and advertisements. This aggregated list is then enriched by three following service tasks which provide additional local expertise. The first task provides information about the connection to the public transport near the apartment. The second one adds the rental index for this area. The third checks online ratings about the quality of life in this area. With this additional information the apartments in the aggregated list are more comparable. The person can now select apartments on the list and decide if a viewing is worthwhile. If so, for each selected item on the aggregated list a “contacting and visiting” workflow branch is executed in parallel, which includes all tasks from contacting the landlord until the signing of the contract. The first task of such a “contacting and visiting” workflow is calling the landlord. Therefore the telephone number is shown and the person can pick it up. If the person and the landlord agree to a viewing, the date and time is directly entered in an online calendar with reminding functionality, otherwise the workflow branch is aborted. Then, the “view apartment” task starts. According to the current GPS coordinates of the person this tasks offers the shortest way by car or public transport to the apartment. If taking public transport the appointed time is considered as arrival time for the viewing. Arrived at the apartment the person may take pictures and add them to the item on the aggregated list. At the end of all parallel executed “contacting and visiting” workflows this helps to recall each apartment and to make a decision. Finally, the person has a list of apartments with all necessary information. She/he can then make a decision and call the landlord of the desired apartment. Focus. This social workflow focused on the analysis of different Internet sources in order to create an aggregated list of objects which contains all necessary information to form a decision. Thereby, different online services for public transport were integrated as well as predefined documents which can be helpful for the person. 11 The structure and all results of the online study are available in german and can be found under http://www.wi2.uni-trier.de/shared/ technical/TP_fopra_social_ext.pdf

21

Organize Move. This workflow accompanies the workflow “search an apartment”. This workflow focuses on the reuse of experiential knowledge, because local knowledge can be valuable for a person in that phase. This local knowledge, e.g. addresses of the town hall, contact persons, forms, and cultural programs, is provided as data objects in the workflow to the person. This workflow starts with support for the change of registration at the phone company and a redirection order for the postal company. Therefore the person receives prepopulated forms in the first task. For the organization of the move the next task reads the Facebook friend list of the person and shows the friends which live nearby in order to ask them, if they could help. While the Facebook friends are asked the person gets a list of local moving and rental car companies. When the move is organized the person gets a predefined checklist of all necessary documents for the departments in the city, so that she/he does not spend much time in collecting and searching all documents by herself/himself. Focus. This social workflow mainly focused on documents and local information which were provided to a person in order to faster collect all necessary information. 6.2.3. Results on Features for Social Workflows First, the results for the proposed features are described. The tables 5-7 contain the described scenarios and their features in detail. The features are basically the tasks of the according social workflow model. The features are ordered by their perceived benefit and the ratings are put in three categories (rating 1-2: bad rating, 3-4: modest and 5-6: good rating). Table 5: Feature Rating: Attending a Festival.

Attending a festival Weather information Shared shopping list Integrated cost calculation Integrate Facebook Map service Provided packing list Share pictures Buy tickets Synchronize appointments Spotify playlist

Bad

Rating Modest

High

5% 10% 15%

25% 20% 20%

70% 70% 65%

5% 0% 5% 38% 30% 43%

40% 50% 45% 15% 30% 26%

55% 50% 50% 46% 40% 32%

65%

25%

10%

Table 6: Feature Rating: Search an Apartment.

Search an apartment Generate aggregated list Public transport Map service Provided documents Add pictures and notes Synchronize appointments

22

Bad

Rating Modest High

0% 6% 6% 28% 6% 22%

17% 17% 28% 17% 50% 39%

83% 78% 66% 56% 44% 39%

Table 7: Feature Rating: Organize Move to New City.

Organize move

Bad

Rating Modest High

Provide information about local administration Provide insider information about the location Organize transporter Change phone/Internet provider and address order Public transport Provide information for students Provide information about local supermarkets and restaurants Provide links to relevant facebook groups Provide information about cultural events Ask friends/professionals for help

20%

7%

73%

20%

14%

67%

20% 27%

20% 20%

60% 54%

27% 20% 14%

20% 26% 40%

54% 53% 47%

33% 26% 73%

20% 47% 7%

47% 26% 20%

A repeating commonality in all three social workflows is that private individuals require a list data structure which contains data like strings, but also aggregated objects (e.g. the packing list for the festival and the list of available apartments of the ‘search apartment’ scenario). Additionally, these lists directly influence the control flow. So in one case all items on the list have to be bought before the next activities are activated and in another case the data items trigger the execution of parallel executed sub-workflows (for each apartment a sub-workflow for contacting the landlord). This is an indication that social workflows need an additional control flow construct. This construct would trigger a parallel execution for each item on the list without a synchronization of the parallel executed branches. The completion of one branch would cause all other branches to immediately stop12 . To summarize, it can be observed that all scenarios require the integration of cloud services. These cloud services are already frequently used by private individuals and may return simple data objects (e.g. the temperature on a location) but also lists, like the personal friends, a packing list, or a list of appointments. 6.2.4. Results on the Perceived Overall Benefit After the textual illustration of a scenario the respondents were asked if they would use such a social workflow service if it supports a scenario exactly the way described. For all outlined social workflows the majority said that they would use such a service. The results obtained from the answers of this question are shown in Table 8. Table 8: Results of the Extended Online Study.

Scenarios

Given Answers No Yes

Attend Festival Search an Apartment Organize Moving

35% 11% 21%

65% 89% 79%

The best rating gained the social workflow on the apartment search, where 89% of the respondents stated they would use a social workflow service in such a scenario. The “organize moving” workflow gained 79% and the “attend 12 According to the workflow pattern categories by van der Aalst et al. [53], this control flow construct would combine pattern 14 (“Multiple instances with a priori run-time knowledge”) and pattern 20 (“cancel case”).

23

festival” workflow gained 65%. This overall rating for the social workflow scenarios confirms the hypothesis H2 (“there are application scenarios in which social workflow services can provide a benefit to the user”). 7. Conclusion and Future Work This paper has shown that social workflows services are a new type of social software that have the potential to be well accepted and used by private individuals. The investigation of related work revealed that no comparable work exists so far that covers all aspects of social workflow services. The integration into a social network facilitates private individuals to use social workflows in their everyday life. This easy accessability together with the ability to automatically execute a social workflow is a considerable relief for a group of friends or a community. Complex tasks or goals can be processed in structured manner and it is always transparent how the current tasks are distributed and how far the process has gone. The presented empirical study showed that the ability to execute social workflows is regarded as important or very important by 92.3% of the respondents. The collective knowledge of the social workflow community will grow every time a shared social workflow is reused and adapted by another person to new requirements. This adaptability feature is seen as important or very important for 76.9% of the respondents, the ability of sharing social workflows is considered important or very important by 69.2%. Due to the fact that people can search for new or similar social workflows, they get connected with people with a similar mind-set and interests, because the interests are directly expressed in the social workflows. This way, people can get in touch with experts, as their knowledge is frequently used in the community and inexperienced users do not need to plan all their social workflows from scratch. So an individual can get in-depth knowledge about a certain problem in which she/he has never been involved before. Social workflows comprise human tasks as well as service tasks which execute online services. The identified services of the scenarios in the online study showed a variety of different services that can be used in social workflows (see Table 5 to Table 7). Such online services occur in social workflows as service tasks. They can form a flow of online services that are connected via the data items they consume or produce. Thus, more complex and modular social online applications can be configured, executed, and shared with the community. By today, the development of a research prototype of the social workflow service CAKEFLOW, described in Section 4, is mostly completed. The CAKE Agile Workflow and Knowledge Engine are fully implemented and the remaining components are almost complete. Social workflows, like the ones described in this paper, can be modelled and executed [54]. The current implementation is only limited due to small set of integrated online services and the restricted set of data types available. The empirical study shows that the spectrum of required online services is very large and heterogeneous. For this reason, we currently develop a generic service description approach which is able to cover the authentication, communication, and integration in the social workflow service CAKEFLOW. This will enable the dynamic composition of online services during the runtime of a social workflow. A second area of future research is related to the semantic descriptions of social workflows. The currently applied process-oriented CBR approach is based on pre-defined ontologies that provide the terminology for the semantic annotation of workflows, tasks, and data items [20]. This approach is appropriate for the annotation of pre-defined online services and well-known human tasks within a closed application area. However, the identified scenarios of the exploratory study (see Table 4) are an indication that many different application areas for social workflows exist and might emerge when people start using a social workflow service for their particular purposes. Thus, pre-defined ontologies lead to the limitation that workflows and related resources cannot be described appropriately due to the lack of terms, which leads to a reduced recall during case-based retrieval. To address this issue, we will research on applying new methods from textual CBR [55]. Alternatively, methods for the automatic extension of light-weight ontologies by free-text descriptions, based on available resources on the web, such as the Open Directory Project13 are interesting. First steps into this direction can be found in research dealing with CBR applications in open domains [56, 57]. In order to grow a community of social workflows users for a certain type of application, an initial repository is needed to provide immediate benefit to new users. To achieve this, a committed set of key users is required, who model their personal experience as social workflows and provide them to the community. We consider the deeper 13 http://www.dmoz.de/

24

integration into a social network such as Facebook or Google+ as an important step to reach and attract large volumes of private individuals. Moreover, we want the exploit the body of experiential knowledge in already existing user communities (e.g. Internet forums). Therefore, we investigate in another line of research how to automatically extract workflows from textual descriptions [58, 59]. Acknowledgement This work is part of the WEDA project. WEDA is funded by Stiftung Rheinland-Pfalz f¨ur Innovation, grant no. 974. We also want to thank Stefan M¨uller, Tonio Kr¨oner, and Hai Ngoc Cu for their help in conducting the studies of the evaluation. References [1] H. Schonenberg, R. Mans, N. Russell, N. Mulyar, W. M. P. van der Aalst, Process flexibility: A survey of contemporary approaches, in: J. L. G. Dietz, A. Albani, J. Barjis (Eds.), Advances in Enterprise Engineering I, 4th International Workshop CIAO! and 4th International Workshop EOMAS, held at CAiSE 2008, Montpellier, France, Vol. 10 of Lecture Notes in Business Information Processing, Springer, 2008, pp. 16–30. [2] W.M.P. van der Aalst, Business process management: A comprehensive survey, ISRN Software Engineering 2013 (2013) 37. [3] S. G¨org, R. Bergmann, S. Gessinger, M. Minor, A Resource Model for Cloud-Based Workflow Management Systems - Enabling Access Control, Collaboration and Reuse, in: Proceedings of the 3rd International Conference on Cloud Computing and Services Science, Aachen, Germany, 2013, SciTePress 2013, 2013. [4] D. d. Roure, C. Goble, R. Stevens, The design and realisation of the myExperiment virtual research environment for social sharing of workflows, Future Generation Computer Systems 25 (2009) 561–567. [5] Y. Gil, V. Ratnakar, J. Kim, J. Moody, E. Deelman, P. A. Gonz´alez-Calero, P. Groth, Wings: Intelligent workflow-based design of computational experiments, IEEE Intelligent Systems 26 (1) (2011) 62–72. [6] G. M¨uller, R. Bergmann, A cluster-based approach to improve similarity-based retrieval for process-oriented case-based reasoning, in: 20th European Conference on Artificial Intelligence (ECAI 2014), IOS Press, 2014. [7] G. M¨uller, R. Bergmann, Workflow streams: A means for compositional adaptation in process-oriented CBR, in: Case-Based Reasoning Research and Development, 22th International Conference on Case-Based Reasoning, ICCBR 2014, Springer, 2014. [8] M. Minor, A. Tartakovski, D. Schmalen, R. Bergmann, Agile Workflow Technology and Case-Based Change Reuse for Long-Term Processes, IJIIT 4 (1) (2008) 80–98. [9] M. Reichert, P. Dadam, ADEPT-Supporting Dynamic Changes of Workflows Without Losing Control, in: Journal of Intelligent Information Systems, 1998, pp. 93–129. [10] B. Weber, W. Wild, An Agile Approach to Workflow Management, in: Bernhard Rumpe, Wolfgang Hesse (Eds.), Modellierung 2004, Proceedings zur Tagung, 23.-26. M¨arz 2004, Marburg, Proceedings, Vol. 45 of LNI, GI, 2004, pp. 187–201. [11] D. Hollingsworth, The workflow reference model, Tech. rep. (1995). URL http://www.wfmc.org/standards/docs/tc003v11.pdf [12] S. G¨org, R. Bergmann, S. Gessinger, M. Minor, Real-Time Collaboration and Experience Reuse For Cloud-Based Workflow Management Systems, in: A. Oberweis, F. Sch¨onthaler, G. Vossen (Eds.), Workshop on Social Business Process Management on the 15th IEEE Conference on Business Informatics, IEEE Computer Society, 2013, pp. 391–398. [13] M. Minor, D. Schmalen, A. Koldehoff, R. Bergmann, Structural adaptation of workflows supported by a suspension mechanism stand by case-based reasoning, in: Enabling Technologies: Infrastructure for Collaborative Enterprises, 2007. WETICE 2007. 16th IEEE International Workshops on, 2007, pp. 370–375. [14] P. Wright, J. Wallace, J. McCarthy, Aesthetics and experience-centered design, ACM Transactions on Computer-Human Interaction (TOCHI) 15 (4) (2008) 18:1–18:21. [15] G. Lindgaard, C. Dudek, D. Sen, L. Sumegi, P. Noonan, An exploration of relations between visual appeal, trustworthiness and perceived usability of homepages, ACM Transactions on Computer-Human Interaction (TOCHI) 18 (1) (2011) 1:1–1:30. [16] M. Minor, S. Montani, J. A. Recio-Garc´ıa, Process-oriented case-based reasoning, Inf. Syst. 40 (2014) 103–105. [17] R. Bergmann, Experience Management: Foundations, Development Methodology, and Internet-Based Applications, Vol. 2432 of LNAI, Springer, 2002. [18] A. Aamodt, E. Plaza, Case-Based Reasoning: Foundational Issues, Methodological Variations, and System Approaches, AI Commun 7 (1) (1994) 39–59. [19] R. C. Schank, J. L. Kolodner, G. DeJong, Conceptual information retrieval, in: Proceedings of the 3rd annual ACM conference on Research and development in information retrieval, SIGIR ’80, Butterworth & Co, Kent and UK and UK, 1981, pp. 94–116. [20] R. Bergmann, Y. Gil, Similarity assessment and efficient retrieval of semantic workflows, Information Systems 40 (2014) 115 – 127. [21] M. Minor, R. Bergmann, S. G¨org, Case-based adaptation of workflows, Information Systems 40 (2014) 142 – 152. [22] D.B. Leake, J. Kendall-Morwick, Towards Case-Based Support for e-Science Workflow Generation by Mining Provenance, in: Klaus-Dieter Althoff, Ralph Bergmann, Mirjam Minor, Alexandre Hanft (Eds.), Advances in Case-Based Reasoning, 9th European Conference, ECCBR 2008, Trier, Germany, September 1-4, 2008. Proceedings, Vol. 5239 of LNCS, Springer, 2008, pp. 269–283. [23] M. Therani, J. L. Zhao, M. Byron, A case-based reasoning framework for workflow model management, Data Knowl. Eng 50 (1) (2004) 87–115.

25

[24] E. Chinthaka, J. Ekanayake, D. B. Leake, B. Plale, CBR Based Workflow Composition Assistant, in: 2009 IEEE Congress on Services, Part I, SERVICES I 2009, Los Angeles, CA, USA, July 6-10, 2009, IEEE Computer Society, 2009, pp. 352–355. [25] I. Peters, Folksonomies: Indexing and retrieval in Web 2.0, Vol. 1, KG Saur Verlag Gmbh & Co, 2009. [26] D. Downs, Discretionary Access Control Guideline, Aerospace Report, The Aerospace Corporation. [27] A. Oberweis, F. Sch¨onthaler, G. Vossen (Eds.), Workshop on Social Business Process Management on the 15th IEEE Conference on Business Informatics, IEEE Computer Society, 2013. [28] A. R. Silva, R. Meziani, R. Magalh˜aes, D. Martinho, A. Aguiar, N. Flores, Agilipo: Embedding social software features into business process tools ulm, germany, september 7, 2009. revised papers, in: Stefanie Rinderle-Ma, S. W. Sadiq, F. Leymann (Eds.), Business Process Management Workshops, BPM 2009 International Workshops, Ulm, Germany, September 7, 2009. Revised Papers, Vol. 43 of Lecture Notes in Business Information Processing, Springer, 2010, pp. 219–230. [29] F. Sch¨onthaler, G. Vossen, A. Oberweis, T. Karle, Business processes for business communities: Modeling languages, methods, tools, Springer, Berlin, 2012. [30] T. Caporale, M. C ¸ itak, J. Lehner, A. Schoknecht, M. Ullrich, Social bpm lab - characterization of a collaborative approach for business process management education austria, july 15-18, 2013, in: A. Oberweis, F. Sch¨onthaler, G. Vossen (Eds.), Workshop on Social Business Process Management on the 15th IEEE Conference on Business Informatics, IEEE Computer Society, 2013, pp. 367–373. [31] K. Gary, T. Lindquist, H. Koehnemann, L. Sauer, Automated process support for organizational and personal processes, in: Proceedings of the international ACM SIGGROUP conference on Supporting group work: the integration challenge, GROUP ’97, ACM, New York and NY and USA, 1997, pp. 221–230. [32] M. Keller, R. Scherer, K. Menzel, A personal planning approach for the integration and coordination of multi project process information, in: Z. Turk, R. Scherer (Eds.), Proceedings of the European Conference on Process and Product Modeling, Portoroz, Slowenia, Swets & Zeitlinger, Lisse The Netherlands, 2002, pp. 549–554. [33] S.-Y. Hwang, Y.-F. Chen, Personal Workflows: Modeling and Management, in: Ming-Syan Chen, Panos K. Chrysanthis, Morris Sloman, Arkady B. Zaslavsky (Eds.), Mobile Data Management, 4th International Conference, MDM 2003, Melbourne, Australia, January 21-24, 2003, Proceedings, Vol. 2574 of LNCS, Springer, 2003, pp. 141–152. [34] M. Weiser, The computer for the 21st century, Scientific american 265 (3) (1991) 94–104. [35] M. Langheinrich, F. Mattern, Digitalisierung des Alltags. Was ist Pervasive Computing?, Aus Politik und Zeitgeschichte (B 42/2003) (2003) 6–12. [36] S. Ben Mokhtar, L. Capra, From pervasive to social computing: algorithms and deployments, in: Proceedings of the 2009 international conference on Pervasive services, ICPS ’09, ACM, New York and NY and USA, 2009, pp. 169–178. [37] J. Zhou, J.-Z. Sun, K. Athukorala, D. Wijekoon, M. Ylianttila, Pervasive Social Computing: augmenting five facets of human intelligence, J. Ambient Intelligence and Humanized Computing 3 (2) (2012) 153–166. [38] M. Endler, A. Skyrme, D. Schuster, T. Springer, Defining situated social context for pervasive social computing, in: Pervasive Computing and Communications Workshops (PERCOM Workshops), 2011 IEEE International Conference on, 2011, pp. 519–524. [39] A. Ranganathan, S. McFaddin, Using Workflows to Coordinate Web Services in Pervasive Computing Environments, in: Proceedings of the IEEE International Conference on Web Services (ICWS’04), June 6-9, 2004, San Diego, California, USA, IEEE Computer Society, 2004, pp. 288–295. [40] F. Montagut, R. Molva, Enabling pervasive execution of workflows, in: Proceedings of the 1st International Conference on Collaborative Computing: Networking, Applications and Worksharing, San Jose, CA, USA, IEEE, 2005, pp. 19–21. [41] M. Kloppmann, D. Koenig, F. Leymann, G. Pfau, A. Rickayzen, C. v. Riegen, P. Schmidt, I. Trickovic, WS-BPEL Extension for People – BPEL4People: A Joint White Paper by IBM and SAP (2005). [42] R. Kramer, M. Modsching, K. t. Hagen, Development and evaluation of a context-driven, mobile tourist guide, International Journal of Pervasive Computing and Communications 3 (4) (2008) 378–399. [43] S. Urbanski, E. Huber, M. Wieland, F. Leymann, D. Nicklas, PerFlows for the computers of the 21st century, in: Proceedings of the 2009 IEEE International Conference on Pervasive Computing and Communications, PERCOM ’09, IEEE Computer Society, Washington and DC and USA, 2009, pp. 1–6. [44] W. Jones, H. Bruce, S. Dumais, Keeping found things found on the web, in: Proceedings of the tenth international conference on Information and knowledge management, CIKM ’01, ACM, New York and NY and USA, 2001, pp. 119–126. [45] J. Grundspenkis, Agent-Based Framework for Modelling of Organization and Personal Knowledge from Knowledge Management Perspective, in: K. Wang, G. Kovacs, M. Wozny, M. Fang (Eds.), Knowledge Enterprise: Intelligent Strategies in Product Design, Manufacturing, and Management, Vol. 207 of IFIP International Federation for Information Processing, Springer US, 2006, pp. 62–70. [46] M. V¨olkel, Personal Knowledge Models with Semantic Technologies, Ph.D. thesis, Karlsruher Institut f¨ur Technologie (KIT), Fakult¨at f¨ur Wirtschaftswissenschaften (Fak. f. Wirtschaftswiss.) Institut f¨ur Angewandte Informatik und Formale Beschreibungsverfahren (AIFB) (2010). [47] D. Pollard, PKM: a bottom-up approach to knowledge management, Knowledge management in practice: connections and context (2008) 95–114. [48] C. Dorn, R. N. Taylor, S. Dustdar, Flexible Social Workflows: Collaborations as Human Architecture, IEEE Internet Computing 16 (2) (2012) 72–77. [49] Y. Gil, E. Deelman, M. H. Ellisman, T. Fahringer, G. Fox, D. Gannon, C. A. Goble, M. Livny, L. Moreau, J. Myers, Examining the challenges of scientific workflows, IEEE Computer 40 (12) (2007) 24–32. [50] J. Kranjc, V. Podpeˇcan, N. Lavraˇc, Clowdflows: a cloud based scientific workflow platform, in: Machine Learning and Knowledge Discovery in Databases, Springer, 2012, pp. 816–819. [51] C. A. Goble, D. C. d. Roure, myExperiment: social networking for workflow-using e-scientists, in: Proceedings of the 2nd workshop on Workflows in support of large-scale science, ACM, New York, NY, USA, 2007, pp. 1–2. [52] M. H¨ader, Delphi-Befragungen: Ein Arbeitsbuch, Springer DE, 2009. [53] Wil M. P. van der Aalst, Arthur H. M. ter Hofstede, B. Kiepuszewski, A. P. Barros, Workflow Patterns, Distributed and Parallel Databases 14 (1) (2003) 5–51.

26

[54] R. Bergmann, S. Gessinger, S. G¨org, G. M¨uller, The collaborative agile knowledge engine CAKE, in: Proceedings of the 18th International Conference on Supporting Group Work, GROUP ’14, ACM, New York, NY, USA, 2014, pp. 281–284. URL http://doi.acm.org/10.1145/2660398.2663771 [55] R. O. Weber, K. D. Ashley, S. Br¨uninghaus, Textual case-based reasoning, Knowledge Engineering Review 20 (3) (2005) 255–260. [56] S. Dieterle, R. Bergmann, Case-based appraisal of internet domains, in: Case-Based Reasoning. Research and Development, 20th International Conference on Case-Based Reasoning, ICCBR 2012, Lecture Notes in Computer Science, Springer, 2012, the original publication is available at www.springerlink.com. [57] W. He, Improving user experience with case-based reasoning systems using text mining and web 2.0, Expert Systems with Applications 40 (2) (2013) 500–507. [58] P. Schumacher, M. Minor, K. Walter, R. Bergmann, Extraction of procedural knowledge from the web: a comparison of two workflow extraction approaches France, April 16-20, 2012 (Companion Volume), in: Alain Mille, Fabien L. Gandon, Jacques Misselis, Michael Rabinovich, Steffen Staab (Eds.), Proceedings of the 21st World Wide Web Conference, WWW 2012, Lyon, France, April 16-20, 2012 (Companion Volume), ACM, 2012, pp. 739–747. [59] V. Dufour-Lussier, F. Le Ber, J. Lieber, E. Nauer, Automatic case acquisition from texts for process-oriented case-based reasoning, Information Systems 40 (2014) 153–167.

27