A Framework for User Interaction for Adaptive Web-Based Information Systems

A Framework for User Interaction for Adaptive Web-Based Information Systems Completed Research Paper Dirk Peters Marc Petersen Carl von Ossietzky University Oldenburg Carl von Ossietzky University Oldenburg dirk.peters@uni-oldenburg.de marc.petersen@uni-oldenburg.de Jorge Marx Gómez Carl von Ossietzky University Oldenburg jorge.marx.gomez@uni-oldenburg.de ABSTRACT Information systems (IS) are a fundamental part of the curricula of today s higher education institutes (HEI) especially in fields of business informatics, economics and computing sciences. But the current methods used to teach IS in HEI are not sufficient when compared to the capabilities of today s learning technologies. The learning material which is provided by software vendors or teachers itself, is neither technologically-enhanced nor adapted to the individual knowledge background and learning preferences of thr ingle, individual learner. Adaptive learning environments (ALE) offer a sufficient way to teach (and learn) IS. They can track the user interaction and adapt the learning material to the learner s performance. Thus, the lacking integration of an ALE with the IS is a major issue. In this contribution, we are presenting a user interactionframework for web-based IS which allows the generation of individual excercises for each learner depending on her/his specific performance in the future. Keywords Web-based Information Systems, Adaptive Learning Environments, IS Education, User Interaction, GRAPPLE. INTRODUCTION The importance of enterprise systems or IS in general, increases more and more. At the same time, the complexity and the amount of people getting in contact with these systems also rises since the last decades. The demand of companies in the employment market has changed in a way, that today s students need to bring skills in planning, implementing and using such systems on a very high (often strategic) level (Boyle and Strong, 2006). The pervasion of computer supported business processes in today s enterprises affects nearly all business sections and their employees (Konradin, 2011). This is especially the case for Enterprise Resource Planning (ERP) systems, which integrate data, processes and functions of different business departments, such as financial accounting, production or sales and distribution (Gronau, 2010). Due to that reason, the resulting, raising need of appropriate learning material and employee trainings needs to be met. In companies, as well as in tertiary education institutes, like universities e.g., it is highly relevant to bring people in contact with these systems in order to ensure an effective use of the different systems (Konstantinidis, Kienegger, Wittges and Krcmar, 2010). This is important for people coming from different fields of studies. Primarily business economics or business informatics people need to be educated in this area, but also engineers or computer scientists should be addressed. Especially the practical use of these systems is an important part of getting people familiarized with the software. Therefore, this handson experience needs to be addressed by future learning methods. This trend is also discussed in the literature very well (Hawking, McCarthy and Stein, 2005; Strong, Fedorowicz, Sager, Stewart and Watson, 2006; Cameron, 2008; Winkelmann, Holler, Püster and Heide, 2012). Unfortunately, the existing training methods don t fulfill this higher and specific demand. The current and most common way to teach and learn the practical issues of a system is to conduct case studies. But these case studies lack due to different reasons: First of all, the common case studies do often not have a deeper didactical background. This is because they are mostly created by the software vendors themselves and therefore only include technical or functional issues. Another problem is that most of the case studies are paper-based and not technology-enhanced in general. Therefore, the learner cannot benefit from well-known advantages coming from methods like distance learning or community-based approaches. The reuse of learning material is much easier within electronic learning environments (Baumgartner and Kalz, 2005). This is Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 1

very important, because the systems are changing a lot due to software updates and new software releases. Thus, the user interface can have a different look in a newer version or further functions are implemented, which also should be part of future, easy adaptable learning material. But the biggest shortcoming we identified in regular training methods is the ignorance of specific characteristics of each individual learner. Most of the time, the learners have a different background in knowledge and skills, but existing learning material often addresses a group of people or students, which are very heterogeneous in their style of learning. Some learners are working on a case study very active and reach the learning target very quick. Some others need more theoretical information about the specific functions in order to complete a task with satisfaction. Within this contribution we will present a developed user interaction-framework (UIF) for web-based IS. This UIF allows displaying tool-tips, tasks or additional information within a web-based IS on top of the used system, like an ERP system for example. Furthermore, the UIF is able to track the user s interactions in the system itself. Based on this information, it is possible, to give individual instructions to the user/learner depending on her/his specific behaviour. Designing adaptation rules in the backend allows the creation of individual training courses, where its tasks can differ in the mode of presentation (visual, textual or multimedia-based) or the level of difficulty (beginner, intermediate or advanced). After this introduction with its problem definition, the major advantages of adaptive learning environments in the context of IS education are given. On the basis of this motivation which also highlights the relevance of the presented topic, the main objectives and the approach of our research are described. In the following, the theoretical concept including the architecture of the developed UIF is presented and the UIF prototype based on an existing SAP ERP system is described. At the end, the potential use of the UIF and the future work are discussed in the conclusions. ADAPTIVE INFORMATION SYSTEM-EDUCATION According to the literature, intelligent learning systems exist since the beginning of the 1980s (Schulmeister, 2007). Surely, the term intelligent has a very vague definition, but starting with the first computer assisted instructional programs, the first reactive or adaptive systems were developed in that decade. The combination of computer technology with methods from the artificial intelligence (AI) aiming at the improvement of educational instructions can be summarized under the most important software category, called Intelligent Tutoring Systems (ITS) (Gharehchopogh and Khalifelu, 2011). In a general understanding, ITS consist of a model of the knowledge of a specific domain (domain model), a model of the learner who uses the learning system (learner model), a model of the pedagogical strategy (tutor model) and a component which is responsible for the communication of the system with the learner (interface) (Schulmeister, 2007). Therefore, ITS can also be classified as adaptive learning systems, because the system reacts on the basis of the learner behavior (Brusilovsky, 2001). Reference models of adaptive learning environments such as GALE (cf. De Bra, Pechenizkiy, van der Sluijs and Smits, 2008), which was developed in the frame of the GRAPPLE project and is used as the basis of the UIF, are implementing the basic concepts of ITS in order to create a broader environment for the learner and her/his educational life. Normally, existing ITS are used in very different domains (programming, mathematics, geography, medicine, etc.) and are also designed for a very broad target group ranging from first grader to domain experts. Therefore, the success of the different instances of ITS also varies in like manner (Schulmeister, 2007). The major advantages of ITS or adaptive learning environments in general are the creation of individual learning material for each single learner, the time- and location independence and the possibility to enhance existing teaching courses with an interactive learning experience (e.g. blended learning). Besides these advantages, ITS and adaptive learning environments also come with some disadvantages which are mostly focusing on the very time-consuming and extremely complicated development of these systems. It is necessary to have programming experts for the development, professional tutors to design effective learning material and domain experts which can deliver the relevant content for the learning material (Csizmadia, 2003). As described in the introduction, the domain of IS education is very relevant for a large group of learners in the field of HEI. Furthermore, the IS education lacks in efficiency due to different reasons. At the moment, there are no adaptive learning environments existing which focus on this specific IS-domain and target group. Therefore it is highly relevant, to make use of the named major advantages of adaptive learning environments in the context of IS education in order to improve the efficiency of today s IS education. OBJECTIVES AND APPROACH In general, the primary objective of our research was to develop a framework which allows the integration between an adaptive learning environment and a real environment of a web-based IS. This is mainly because we want to give the learners a hands-on experience. Furthermore, the integration into an existing system allows the support for trainings on-the-job in order to deliver an approach which is also applicable in companies to their employees. But due to the changing demands in Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 2

higher education institutes (HEI) (Peters, Haak and Marx Gómez, 2012) we are mainly focusing on this HEI field. This also allows us, to evaluate the developed prototype in existing university courses and compare the learning outcomes with the results of traditional teaching methods like case studies, etc. The usage of the developed UIF in the field of on-the-job or professional trainings is a topic for further research. In order to achieve the primary objective we developed a technical solution in a form of a learning system which firstly supports the tracking of user interactions and secondly offers the possibility to display additional information, such as learning instructions or tasks, within that environment. This all should be done without modifying the source code of the target web-based IS. In the research process, existing related systems and approaches were reviewed to identify ways of user interaction tracking and ways to display additional information. After this literature review, we found out, that the proposed UIF can be based on an existing reference model called GALE, which stands for GRAPPE Adaptive Learning Environment and already delivers some major functionality for an adaptive learning system, like the design of an adaptation engine or the definition of adaptation rules. Based on that reference model, the theoretical concept of the UIF was designed. In a last step, a UIF prototype was developed in order to proof the underlying concept. The results are presented and discussed in the following sections. USER INTERACTION-FRAMEWORK This section proposes the UIF which is an outcome of enhancing GALE towards displaying additional information. This main section of our contribution firstly introduces the proposed information flow to be implemented within a learning environment to adapt the learning process in terms of giving the learner new tasks or additional help information. Thereafter, analyses of the GRAPPLE components are conducted and an extension to GALE is presented in order to expand the functionality of the adaptive learning environment. In the end of this main section, a requirements definition for each core component will be identified. Basic Idea and Information Flow Nowadays, learning environments have no direct connections between the learning system and the IS. As a result, the students work with two completely isolated systems. But both systems contain relevant data which can be used to improve the learning process. By integrating both systems, the learning environment - and finally the student - can benefit from the existing data stored in an IS by collecting information about the students' performance. Figure 1: Information Flow Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 3

The basic idea of the UIF can be described by explaining the information flow, shown in Figure 1. The design introduces existing components of a learning environment, namely the students, the web-based IS, the learning system (LS) and the teacher. In order to overcome the lack of integration, the UIF adds three more components to the information flow, namely the configuration file, a browser plug-in and the tracking and displaying script, to directly connect the LS with the web-based IS. This allows a way of learning where students are able to receive their learning tasks, additional information and guidance as well as support directly within the targeted web-based IS. From the user interaction point of view, the student operates with the web-based IS regularly via web browser which has the UIF-browser-plugin installed. This plugin receives information and tasks from the LS over a displaying script and displays the tasks at the right place directly on top of the web-based IS. In addition, the plugin communicates with the LS over the tracking script. This tracking script forwards the student interactions directly through the LS. This enables the LS to instantly evaluate the learning results and students behavior in order to adapt the learning process in terms of giving the student a new task or additional help information. In order to make use of the UIF the creation of a configuration-file is necessary as an initial step. This file includes the structure of all relevant html-elements of the GUI of the web-based IS. This creation needs to be done once for all relevant GUI-elements, which want to be used in the context of a learning scenario. Through this configuration-file the LS knows about the structure of the web-based IS and is able to allocate html-elements to a function of the web-based IS. The underlying (adaptation) logic is delivered by GALE as part of the mentioned GRAPPLE framework. There, it is possible to define so called adaptation rules, which describe the exact behavior of the LS. More precisely, the adaption rules deliver a reasoning which new exercise will be presented to the student based on the tracked information of the students' performance. As an example, the system can provide positive feedback, if the learner clicks on the correct button within the web-based IS according to her/his current task. However, if the user clicks on a wrong button, she/he might get additional information about what was wrong and how the problem can be solved in a next step. Besides these very simple examples, the UIF also allows the definition of more complex adaptation rules in order to provide a more efficient learning experience. As another example, the difficulty level of the shown task can vary depending on her/his performance at the system. Therefore it is imaginable to display more advanced tasks for experienced learners or, on the other side, give more theoretical information for learners who get stuck during a task. Architecture of User Interaction-Framework The UIF extends the core components of GALE. Therefore the GALE architecture was analyzed and transformed into components. Components necessary to integrate the LS and the web-based IS were extracted from the introduced information flow and added to the component diagram, shown in Figure 2. The new components in the architecture are colored. To highlight the generic character of the developed architecture, the UIF can also be called a Generic User Interaction- Framework (GUIF). The first extension to the GALE architecture is made to the existing web browser component. Previously this component was only able to render html sites which were completely assembled by the GALE Servlet. The extension enables the web browser to render content of the web-based IS and in addition to render content from the GALE servlet on top of the webbased IS interface. The web-based IS component is not part of the UIF architecture itself, because it is an external system which is connected to the LS through the added components. For the sake of clarity it is mentioned in the overall architecture in Figure 2 and therefore has a different color. The GUIF Web Browser includes three components, the displaying and tracking script as well as a file named GALE configuration. This file includes dynamic settings for the two scripts, e.g. the type of interactions which should be tracked. Dynamic settings means that the file is generated based on the GRAPPLE Authoring Toolkit (GAT) configuration; which can be changed during runtime. All three components are part of the web browser plug-in and are loaded or updated from the UIF server when the user starts a new learning session. The displaying and tracking script will directly be injected into the XHTML page whereas the GALE Configuration file runs in the background of the plug-in and can be accessed with readrights by both scripts at any time. The GALE Servlet also has two new components, the GUIF Display Connector and the GUIF Tracking Connector. Previously, the GALE Servlet component was only able to deliver XHTML websites from its own server to the client based on HTTP requests. The GALE Servlet then waits for the HTTP response. In order to enable tracking of user interactions on a website, which runs on a different server, the GALE Servlet needed to learn a Web Service protocol. It must then be able to receive Web Service calls from the tracking component. This extension is provided by the GUIF Tracking Connector, which also transforms the received data into a format that is understandable for the Update Processor component. Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 4

The Concept Manager delivers the conceptual information based on an URL, while being part of the original GALE model. The purpose of the Display Connector is to manipulate the concept manager to deliver additional information instead of full websites. The delivered additional information is made available by the Display Connector component until the Displaying component requests the next message. Just as it was highlighted in the Tracking Connector, the request from the Displaying component to the Display Connector is also performed over a Web Service protocol. Figure 2: Extended GALE Architecture The GUIF IS Configuration and GUIF Configuration Adapter are two extensions which were made to the original framework. The IS Configuration component is a configuration file which is generated based on the specific web-based IS and provided to the Configuration Adaptor. The Configuration Adaptor component is needed in order to facilitate the teacher in creating new exercises and adaptation rules by using the GAT. Each web-based IS might have a completely different structure. The relevant parts of this structure is needed to be understood by the teacher and are therefore stored in the configuration file which has been generated by a technical expert or automatically. Requirements This section investigates the functional requirements of the components of the proposed framework for the development of the proof-of-concept prototype. The IS configuration component provides the necessary information for task and adaptation rule creation and thus it reflects the technical structure of the IS web based interface. The general structure of such an interface can be divided into pages which themselves can be divided into elements. The configuration adapter adapts the GAT towards handling the IS configuration. This component reads in the IS configuration and enables the creation of learning instructions for specific elements of a specific page. An author can use GAT to create domain concepts (a learning instruction is used as an equivalent to a domain concept) and to group domain concepts into domain models. Furthermore the author can create rules to update the user model and rules to trigger the display of learning instructions. These rules are based on the fact that a domain concept or a complete domain model is known or unknown to a specific user. In fact a rule triggers the display of a learning instruction this instruction is forwarded to the display connector. The display connector provides a Web Service to access the forwarded learning instruction. The displaying script uses the Web Service of Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 5

the displaying connector and requests the next learning instruction. After processing the instruction a tooltip shows its content on top of the IS web based interface in relation to the current user relevant element. The tracking of user interaction is performed based on a configuration file on the client side which contains all relevant pages and elements as described in the previous section. The tracking script tracks three types of interactions namely mouse interaction, keyboard interaction and no interaction. No interaction may indicate that a user is stuck in the learning process and needs assistance. Tracked user interactions are forwarded to the tracking connector which provides a Web Service for receiving user interactions. Received user interactions are forwarded to the update processor component which finally updates the user model. UIF PROTOTYPE Implementation and Technologies The defined requirements from the previous section are implemented in a first version of a prototype in order to show the general functionality of the framework. This section introduces the technology stack chosen to implement the framework in order to map the products, frameworks and technologies to their application area within the UIF prototype. The chosen web-based IS is SAP NetWeaver Portal installed in version 7.02 ABAP Trial Version on a virtual appliance of a Windows Server 2003 R2. The SAP NetWeaver Portal can only be accessed with the Microsoft Internet Explorer, nevertheless the client-side browser plug-in was developed with Crossrider, a framework to develop cross browser extensions. The browser plug-in itself was developed using JavaScript together with jquery. The injection of JavaScript, XHTML and CSS into the targeted web page performed the tracking and displaying functionality. The communication between the client and the server was realized with Representational State Transfer (REST) Web Services and the server returns data in a JavaScript Object Notation (JSON) format. The server-side implementation is based on the standard Java EE 6 technology stack. Providing RESTful Web Services was developed with the use of the RESTEasy project, which is a full implementation of the JAX-RS specification. The GALE context is represented by the Context and Dependency Injection (CDI) framework and the Java Persistence API (JPA), an object-relational mapping tool, is used to persist data in the database or query data from the database. The prototype uses an embedded database, called HyperSQL. The overall application was configured and built using the Apache Maven Project in version 3 and is hosted using the JBoss Application Server 7. In order to get an idea, how this technology in the UIF prototype is used in a use case scenario, the next subsection describes an SAP example. SAP NetWeaver Portal Case Study This section demonstrates the use of the proof-of-concept prototype within a use case scenario. The use case is documented with two screenshots in Figure 3 and 4. In order get an overview about the main idea of the use case scenario, not every single screen of the described use case scenario is printed in this contribution. The learning task of the use case scenario is to show the learner how to create a new appointment in the personal calendar of the web-based IS in the SAP Easy Access menu. Therefore the UIF will guide the learner through two different ways of doing this (a way represents a certain domain model). The first approach is designed for novice users, showing them how to perform the tasks by using a common navigation path. The second approach is designed to be executed by advanced users and shows how to perform the tasks using the quickest path of navigation. The use case scenario is designed for the SAP NetWeaver Portal and starts tracking and displaying after the user logged in into the system. After the login the learner is located on the SAP Easy Access page. The first domain model is activated and shows the information Open the SAP menu: Office Appointment Calendar and double-click on the SSC1 Owner menu entry. Figure 3 exemplifies how this information is displayed in the learning environment. The UIF tracks when the learner has performed this task and the UIF will display Click on the Create Appointment button to create a new appointment on top of the next page. This leads the user to a dialog with a form to create an appointment. After the user has entered some information into the form the next advice of the UIF is Save the appointment by clicking on the button with the green check sign. The final advice Close the calendar by clicking on the Exit button leads the user back to the SAP Easy Access page. Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 6

Figure 3: UIF SAP Example 1 The second domain model is now activated, because the first domain model is marked as known to the learner. As shown in Figure 4, on the SAP Easy Access page the UIF instructs the learner Type in SSC1 into the transaction box. After this interaction has been tracked the next advice given by the UIF is Right click on the cell in the calendar, which indicates when your appointment should start and select New Appointment from the context menu. This leads the learner to the dialog with the form to create an appointment. This time the UIF waits until the learner has saved the appointment. Back on the calendar page the UIF advices the user Press Shift+F3 to go back to the SAP Easy Access page. Figure 4: UIF SAP Example 2 During this learning session the UIF has shown seven instructions (four in the first and three in the second domain model) and tracked ten interactions (six in the first and four in the second domain model). Finally the learner has created two appointments in her/his personal calendar in two different ways and mastered two domain models. SUMMARY AND CONCLUSIONS In this contribution we presented a user-interaction framework which allows the allocation of individual learning tasks to different learners who want to get familiar with the usage of web-based IS. Therefore, we presented an integrated approach in order to track the learner s behavior and display upcoming learning tasks within the IS itself. Based on the learner interaction, the UIF offers the possibility, to allocate different kinds of tasks with regards to previous knowledge and/or learning style, to different types of learners. Starting with an introduction in the first section, we highlighted the existing problems in today s IS education at university level. In order to present the current state-of-the-art in the field of adaptive learning environments, we also gave a short insight in the relevance of ITS which can provide the basis for a new adaptive approach. As a result, we found out that there are existing technologies which can be used for the development of an enhanced adaptive learning environment in the context of IS education at HEI. In the following section the objectives and the approach were described. After presenting the major theoretical background of the UIF concept, we also explained and discussed the implemented prototype which is a first version of the adaptive learning system. In this last part, we also presented two screenshots of the existing UIF prototype in order to get a better understanding of the functionality of the prototype. The main idea of this contribution is the integration of an adaptive learning environment with a real, existing IS. This system and data integration generates the main benefit of our approach which is mainly the possibility of tracking, collecting and analyzing data about the learner s tasks and her/his performance in the system. Furthermore, through this integration, tasks can be displayed context-sensitive on top of the graphical user interface of the web-based IS. Due to the architectural basis coming from the existing GRAPPLE framework, the UIF approach allows the generation of individual learning material Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 7

which can be presented to the learners directly through the learning system. In summary, we provide an approach which gives the opportunity to learners, to gather from hands-on experience in a more efficient way of learning in the field of IS education. REFERENCES 1. Baumgartner, P. and Kalz M. (2005) Wiederverwendung von Lernobjekten aus didaktischer Sicht, in Djamshid Tavangarian and Kristin Nölting (Eds.) Auf zu neuen Ufern! E-Learning heute und morgen, 34, NY, München, Berlin: Waxmann, 97-106. 2. Boyle, T.A. and Strong, S.E. (2006) Skill Requirements of ERP Graduates. Journal of Information Systems Education, 17(4), 403-413. 3. Brusilovsky, P. (2001) Adaptive Hypermedia. User Modeling and User-Adapted Interaction, 11, 1-2, 87-110. 4. Cameron, B.H. (2008) Enterprise Systems Education: New Directions & Challenges for the Future. AMCIS Proceedings, 119-126. 5. Csizmadia, V. (2003) Constructing an Authoring Tool for Intelligent Tutoring Systems with Hierarchical Domain Models. Worcester Polytechnic Institute. 6. De Bra, P., Pechenizkiy, M., van der Sluijs, K. and Smits, D. (2008). GRAPPLE: Integrating adaptive learning into learning management systems. Available at: http://www.editlib.org/d/29093/proceeding_29093.pdf. 7. Gharehchopogh, F.S. and Khalifelu, Z.A. (2011) Using Intelligent Tutoring Systems in Instruction and Education. 2nd International Conference on Education and Management Technology, 13, Singapore: IACSIT Press, 250-254. 8. Gronau, N. (2010) Enterprise Resource Planning: Architektur, Funktionen und Management von ERP-Systemen. (2nd ed.). München: Oldenbourg. 9. Hawking, P., McCarthy, B. and Stein, A. (2005) Integrating ERP s Second Wave into Higher Education Curriculum. Proceedings of the 9th PACIS Conference 2005, 1001-1008. 10. Konradin Business GmbH (2011) Konradin ERP-Studie 2011: Einsatz von ERP-Lösungen in der Industrie. Leinfelden- Echterdingen: Konradin Mediengruppe. 11. Konstantinidis, C., Kienegger, H., Wittges, H. and Krcmar, H. (2010) Planspiele in der ERP-Lehre: Eine empirische Untersuchung deutscher Bildungseinrichtungen, in Matthias Schumann, Lutz M. Kolb, Michael H. Breitner and Arne Frerichs (Eds.) Multikonferenz Wirtschaftsinformatik 2010, Göttingen: Universitätsverlag, 1709-1721. 12. Peters, D., Haak, L. and Marx Gómez, J. (2012) Adaptive Learning Cycle to improve the Competence-Building for Enterprise Systems in Higher Education, in João Eduardo Quintela Alves de Sousa Varajão, Maria Manuela Cruz-Cunha and Antonio Trigo (Eds.) Organizational Integration of Enterprise Systems and Resources: Advancements and Applications, IGI Global, 76-99. 13. Schulmeister, R. (2007) Grundlagen hypermedialer Lernsysteme: Theorie - Didaktik - Design (4th ed.). München: Oldenbourg. 14. Strong, D., Fedorowicz, J., Sager, J., Stewart, G. and Watson E.E. (2006) Teaching with Enterprise Systems. Communications of the Association for Information Systems, 17, 33, 728-755. 15. Winkelmann, A., Holler, J., Püster, J. and Heide, T. (2012) Lernerfolg durch ERP-System-Vergleich Einsatz von SAP Business ByDesign und SAP ERP in der Lehre, in Dirk C. Mattfeld and Susanne Robra-Bissantz (Eds.) Multikonferenz Wirtschaftsinformatik 2012, Berlin: GITO, 525-536. Proceedings of the Nineteenth Americas Conference on Information Systems, Chicago, Illinois, August 15-17, 2013. 8