2015 4th International Conference on Advanced Computer Science Applications and Technologies Multi-Layered Alert Filtration and Feedback Cycle using Brahms Model Abid Ghaffar1,2 , Mohamed Ridza Wahiddin1 , Asadullah Shaikh3,4 , and Akhlaq Ahmad1,5 1 Department of Computer Science, Kulliyyah of Information and Communication Technology, International Islamic University Malaysia. 2 Department of Computer Science, Foundation Year Program, Umm Al-Qura University, Makkah, Saudi Arabia, 3 Department of Computer Science and Information Systems, Najran University, Najran, Saudi Arabia. 4 Faculty of CS and IT, Institute of Business and Technology, Korangi Creek Karachi, Pakistan. 5 College of Engineering and Islamic Architecture, Umm Al Qura University, Makkah, Saudi Arabia, Email: aaghaffar@uqu.edu.sa, mridza@iium.edu.my, asshaikh@nu.edu.sa, aajee@uqu.edu.sa sions without any solid evidence and proof. Identification of an employee while doing mistakes during his job activities and putting him on a right track would benefit organization in the long term basis[11]. Brahms Modeling and Simulation technique is used to simulate activities in a work practice system[21], [19]. Alert is generated using Brahms and then filtered through multilayered filtration system. Employees receive alerts through feedback cycle at the time of human errors while performing job activities as shown in Figure 1. In this paper we propose multi-layered alert filtration system involving context aware layer, Key Performance Indicator (KPI) layer and Human-Decision-Integration using Brahms Model as shown in Figure 2. Multi-layered alert filtration system would filter alerts generated by Brahms Modeling and Simulation technique and may take rational decisions. Intelligent alerts can be generated through sms or emails directly to workers who make mistakes during job activities. Human Computer Interaction is useful for todays computer world where we are interacting with ubiquitous devices having different set of controls[5]. Our proposed system is specially designed and useful for interactive mode of applications where user performs his job activities. In case of human error activity either computer-based or manual, the system would respond and generate an alert to protect the work practice environment [20]. There are two types of simulators working in the proposed system, one is Brahms Simulator and second one is Actual Time Simulator as shown in Figure 3. We use Actual Time Simulator for the validation process of Brahms Simulator. Both simulators process jobs independently, therefore input is given to the Brahms Simulator through Actual Time Simulator to test the results and output from Brahms Simulator. The results would be compared with the same input given to Brahms Simulator independently. If the results are same, then Brahms Simulator results would be validated easily. Abstract—Customer service improvement is directly related with organizational standards and productivity. Employees activities have certain objectives to be followed but sometimes outcomes are different than expected. Human error while performing regular job activities cause sufficient losses and difficult to address. Management faces real challenges while dealing with employee related issues and sometimes becomes unproductive. We propose Brahms Model with multi layered alert filtration and feedback cycle which would address the human error in the system and generate the filtered alerts in the form of sms or by emails. This is especially useful for HumanComputer Interfacing (HCI) Scenarios. Controlling human error and addressing real issues may protect any institution from severe damage and losses. Efficiency of a customer service department can be improved and maximized by multi-layered alert filtration system using Brahms Model. Keywords-Brahms Model; Customer Service Improvement; Context Aware Applications; Alert Generation and Filtration; Organizational behaviour; Human Computer Interfacing I. I NTRODUCTION Customer service improvement plays a significant role in the finishing product of a system. Employees are the most essential part of an organization that are exceptionally responsible for the failure and success of predetermined goals and objectives[18], [12]. Human error detection and its rectification of a worker is a real challenge in a competitive environment[11], [9], [10]. Huge investment is made in terms of equipment and tools but the results are not encouraging. Sometimes, workers do not know the system rules and make mistakes which need to be addressed on time. On the contrary, some workers are habitual to make mistakes which may compromise overall organizational performances[18], [12]. Employees conduct issues and monitoring feedback analysis need to be investigated thoroughly. Customer service can be improved, if employees behavior is streamlined while performing job activities[18], [12]. Reputable organizations are sensitive about self-esteem and respect of their individuals and do not take premature deci978-1-5090-0424-9/16 $31.00 © 2016 IEEE DOI 10.1109/ACSAT.2015.27 121 II. R ESEARCH M ETHODOLOGY Figure 1. We started our research study simply with bibliographic research and then later on qualitative and design research is applied[22], [17]. Online survey carrying twenty two questions was also conducted from different professionals belongs to different organizations in order to ensure human behavioral problems exist in their system[8]. Alerts sent to workers, who do not make mistakes may challenge the authoritys decision making procedures. Therefore, there is a need to verify and validate the decision making process for sending alerts to the identified workers to avoid chaos or discredit the system. Multi-layered alert filtration system is suggested which would filter the alerts using Brahms modeling and simulation technique. Employees in an organization would be identified and received filtered alerts based on their mistakes. Quantitative research methodology is helpful for seeking suitable tools for capturing human activities in a work practice system[4]. It also guides us to choose and analyses the best tool available in this regard. Finally, formalism for the best available tool is checked and analyzed[8]. On the other hand, Design Research is also used and considered as improvement research comprised of Knowledge Flows, Process Steps and Output[22], [17]. Process Steps mainly involve with Suggestions, Development, Evaluation and Conclusion. Similarly Output is concerned with Proposal, Artifact, Performance Measure and Results[22], [17]. Data collection is ensured using ethnography study to observe work practice system. It involves human observation through Voice Recording, Video Recording, Pictures, Questionnaires and Interviews wherever appropriate. Different scenarios at different situations are written in the form of a report having all the components of Brahms Model[21], [19], [13]. We generated alerts using Java API and input is given to the context aware layer, carrying details of employees KPI. Finally, alerts reach to the Manager Desktop, who would take the final decision, either alerts need to be sent to employees through communication layer or it should be further reconsidered. Implementation of Alert Generation using Brahms Model[11]. Figure 2. Context Aware Layer, KPI- Layer and Human-DecisionIntegration in the system[11]. We conducted two research studies at Malaysian University using ethnography study to observe human activities in the customer service department[13]. One research study was organized at International Student Affairs Division (ISD), Immigration Unit and other one was conducted at Graduate School of Management (GSM). III. R ESEARCH C ONTRIBUTION Brahms Modeling and Simulation is used basically to capture human activities in any work practice system. It can be used in the customer service improvement by introducing human error detection and alert generation. These alerts can be filtered through multi-layered filtration system using context aware filtration and human in the loop cycle. No alert would be generated unless it is not filtered through multilayered filters as shown in Figure 1. Normally, alerts may be generated and may be delivered to employees without any check or scan process. Once alerts are filtered and scanned properly, it can be delivered to those employees, who make mistakes in the work practice environment. This paper is divided into different sections while addressing human error detection in the work practice environment and generating filtered alerts through feedback cycle. Section II comprised of Research Methodology used in the proposed system, Section III focuses on the contribution part. Section IV explains the basic concept of Brahms Modeling and Simulation concept, Section V explains the functioning of different layers involved in filtration of alerts. Section VI and VII addresses Brahms Automation and Implementation segments. Section IX explains the Related Work and background of our references and finally Section X is related with Conclusion and Future Work. 122 V. M ULTI -L AYERED A LERT F ILTRATION Figure 3. Multi-layered alert filtration system comprised of Context Aware Filtration layer, KPI Layer and Human-DecisionIntegration as shown in Figure 1 and Figure 2. Alerts generated from Brahms Simulation have to be verified and checked in these three layers sequentially. Once decision is taken by the manager in the form of Human-DecisionIntegration, alerts are passed on to the communication layer to interact with employees through sms or emails. Filtered alerts are generated through Brahms Modeling and Simulation process. Context aware filtration and KPI layer would be used together or separately as shown in Figure 1 and Figure 2 to filter the alerts. Manager receives these alerts and takes an appropriate decision based on employee track record [14], [16]. Every employee in the customer service department has certain environment and conditions where he is supposed to deliver his job. The context of an employee involves job timings, activities performed, human errors done, KPI, customer complains, appreciation letters received and efficiency rating [1]. All these parameters can be captured in a context aware filtration layer. On the other hand, KPI filtration layer would capture the employee key performance indicator exclusively[7], [15]. Later on, manager receives alerts and takes the decision either these alerts need to be verified again or sent directly to communication layer. Two separate databases are created for these two layers which can be accessed and updated as per requirements as shown in Figure 2. Brahms Simulator validation using Actual Time Simulator. Alerts filtered by the manager are received by the communication layer which is specially designed to handle sms or emails as shown in Figure 1. This particular layer is responsible for the source of communication between employees and management at the time of human errors. Once human error is fixed by the employees, the message would be sent back to the manager through the same communication layer in the form of sms or emails. The feedback cycle can be easily viewed through the communication layer responsible for handling SMS or email alerts. Finally, an overall response of Brahms Simulator can be checked and verified through another simulator which mimics actual work practice system as shown in Figure 3. Once Brahms simulation is finalized and we get our results, there should be some check and balance outside Brahms Model, either Brahms Simulator is behaving in the same way as it is designed or there are some other differences. Therefore, we introduce another simulator called Actual Time Simulator which would capture the activities of employees in the actual work practice system and pass it on to the Brahms Simulator. If the results are same then before, then it proves that our Brahms Simulator is generating proper filtered alerts to those employees who make mistakes in the system as shown in Figure 3. VI. P ROPOSED F RAMEWORK AUTOMATION The proposed framework may be connected to the actual work practice system online in order to track down the human errors on real time basis [7]. The main objective is to automate the activities in the work practice environment and generate the filtered alerts automatically [23]. The activities which are performed through online databases, email communication, photocopier activity, fax communication and online fee payment system could be easily automated[7]. Manual activities need to be entered as an input manually into the proposed framework to generate the filtered alerts[7]. IV. B RAHMS M ODELING AND S IMULATION Brahms is a multi-agent business processing modeling tool which is basically used for representing work practice system through simulation[21], [19]. There are different aspects in Brahms which need to be considered while modeling work practice system, for example; Agents, Objects and Classes, Timings, Geography, Activities, Communication and Knowledge[21], [19]. Data is collected from the work practice environment using ethnography study[13]. Scenarios collected from an organization are converted directly into Brahms Model and simulated through Brahms environment. Results are stored in the databases and output is displayed using Agent Viewer component of Brahms[21], [19]. VII. R EASON F OR C HOOSING B RAHMS F OR A LERT G ENERATION Brahms Language is useful for modeling interactions among people, devices, objects, documents and communication tools. Locations of the people and objects are important to perform certain activities, which is available in Brahms. The simulation of activities performed in different scenarios in the customer service department could undergo a field test during implementation process. Interaction is possible for multiple Brahms systems using proxy agents. 123 Figure 4. Brahms Simulator without using Actual Time Simulator. Communication layer provides a dynamic architecture which allows changes in the components at runtime[6]. The Subsumption feature in Brahms language provides the facility to start, resume and interrupt any activity based on the condition-action rule. Agent Viewer tool of Brahms language provides us the facility to display all the activities in the work practice system for visualization purpose[6]. The main reason for choosing Brahms for generating filtered alerts may not be possible through other language platforms. The unique features of Brahms containing different sub-models like Agents, Timing, Activity, Geography, Object, Communication and Knowledge are integrated together at one single platform [6]. Most of the multi-agent based languages mainly focused on few parameters like agents or activities while Brahms provides complete picture of an activity with required parameters as mentioned above. Figure 5. Scenario for Pre-determined mistakes in case of new Student Visa Application. Brahms Simulator. Brahms simulator responds in the same way as it was responding with input scenarios as shown in Figure 3. This process relates Brahms simulation with the actual work practice system. Consider an example of a new student visa application procedure in a customer service department of immigration unit at Malaysian University where human errors are committed. Brahms Simulator takes many scenarios like Figure 5 as an input and processes it; filtered alerts are generated as shown in Table I. Brahms Simulator is trained by series of scenarios to get mature enough decisions using framework in Figure 1 and Figure 2. Mistakes are divided into two categories that is, predetermined and random. Pre-determined mistakes are those mistakes which are already known to the employees or an observer while random mistakes are those mistakes which are unknown. Initial results are generated based on predetermined mistakes in the series of different scenarios as shown in Table I. The status of employee mistakes and relevant details could be easily reviewed through Table I using different column details like Employees, Date, Time, Alerts and Status. VIII. B RAHMS S IMULATION I MPLEMENTATION Brahms Modeling and Simulation is tested through actual simulation of work practice system[21], [19]. Brahms Simulator is trained by multiple processing scenarios and all the activities are monitored and well recorded, consequently filtered alerts are generated through feedback cycle. We need to analyze the behavior of Brahms generated results either alerts are generated properly at the time of human error in the system or misdirected. Alerts which are not filtered or misguided may lead the system unstable and put a question mark for institution credibility[3], [2]. The proposed framework for alert generation would be effective, if it is connected with the actual work practice system otherwise the whole effort for the creation of alert generation would be compromised. Brahms Simulator is working independently and generating filtered alerts as shown in Figure 4. We need to introduce another Simulator in parallel for verification of results called as “Actual Time Simulator” as shown in Figure 3. Sometimes, Brahms Simulator may generate different results than expected which can be verified and validated through Actual Time Simulator. The second simulator located outside Brahms receives input from external source independently and pass it on to IX. R ELATED W ORK Existing research carrying the idea for capturing human activities during work practice system and generating the alerts without detail checks and scan process. The study involves Context Aware Filtration Layer, KPI and Human in the loop factor to protect the employees interest in the proposed framework. Abid Ghaffar et al. Feb. 2015 presented an idea about filtered alerts using the concept of Context Aware Security 124 Employees Sister-1 Sister-2 Brother-1 Brother-2 Brother-3 Date July 1, 2014 July 3, 2014 July 8, 2014 July 1, 2014 July 14, 2014 Time 09:50:10 A.M 11:22:15 A.M 03:15:12 A.M 10:15:42 A.M 10:30:28 A.M Alerts Documents accepted without Picture Processing Fee missing Passport accepted going to be expired shortly Hard Copy for students information missing Passport copy missing Status Active Inactive Active Active Inactive Table I F ILTRATION OF ALERT GENERATION USING P RE - DETERMINED MISTAKE SCENARIOS views about quantitative research methodology and the procedures involved while doing research. They focused the relationship of numbers with actual research parameters[4]. Chris Johnson 1999 focused on the relationship between workers behavior and organizational improvement. He suggested institutional failures are strongly connected with people attitude and dealings while delivering their jobs[12]. Brigitte Jordan 1996 presented her ideas about understanding work flow system in the work practice system using Ethnography Study. She used different methods for data collection like Photography, Video Recordings, Voice Recording, Questionnaires and Interviews[13]. Marc J. Epstein and Marie-Josee Roy 2001 while David Parmenter April 2015 spotlights the parameters involved in Key Performance Index for the job performance. They explained those factors which are responsible to increase the employees efficiency in an organization[7], [15]. Maarten Sierhuis et al. October 2009 presented an idea for the decrease of work load by 90% by employing intelligent agent software technology in the form of Brahms Modeling and Simulation[20]. and Brahms Model. Alerts generated through Brahms Modeling and Simulation may be filtered through Context Aware Security before it is delivered to workers who make mistakes in the work practice system[11]. Abid Ghaffar et al. Jan. 2015 suggested an idea about implementing alert generation system using Brahms Model. Warning codes can be written clearly to target workers, who make mistakes during job activities[9]. Abid Ghaffar et al. Feb. 2015 evaluated and compared different multi-agent systems and proved that Brahms has distinguished features to model and simulate work practice systems[8]. Abid Ghaffar et al. May 2013 discussed about warning alert generation system using Brahms Model. Alerts are beneficial for the customer service improvement in an organization[10]. Mohamad Fauzan Noordin 2013 highlighted the relationship between heart-ware and information technology. He focused on the importance of a personal behavior while implementing software systems[14]. Stephen Robbins 2012 discussed about organizational behavior in terms of productivity and output. He focused on the basic requirements and motivations for the workers which enable them to be more productive[18]. Saad Almutairi et al. 2012 focused on the context aware systems based on time and location. He explained the importance of security systems based on time and space[1]. Geong Sen Poh et al. 2012 proposed a security framework which addresses the flaws in an information flow[16]. Cristian Bravo-Lillo et al. 2011 presented an idea about the importance of warning dialogue communication. The messages for the user through dialogue boxes in the event of security issues are extremely important which may lead to disaster, if misunderstood[3]. Jim Blythe USC et al. 2011 discussed about the importance of communication in the event of a cyber-attack. One can take wrong decisions, if messages are not understood properly[2]. Maarten Sierhuis et al. 2002 and 2007 expressed his views about the simulation of work practice system using Brahms Model. He proved that different parameters like Agents, Objects, Activities, Geography, Timing, Knowledge and Communication could be represented in the form of job scenarios using Brahms[21], [19]. Mark Balnaves and Peter Caputi 2001 expressed their X. C ONCLUSION AND F UTURE W ORK Employees make mistakes in an organization and its rectification on time is essential. Generating alerts to the right person with proper communication skills would be a great advantage in long term basis to restore sanctity of an employee and organization together. Miscommunication to employees or giving them unauthorized alerts may worsen the situation and would raise a question mark for the credibility of workers. We propose multi-layered filtration for alerts generated through Brahms Modeling and Simulation. The main components of alert filtration involve context aware filtration layer carrying KPI index of employees and human in the loop. Alerts have to be filtered through context aware layer and then manager has to take a final decision either alerts should be sent to the corresponding employees or not using communication layer. Validation process could be achieved by introducing Actual Time Simulator along with Brahms Simulator. Different inputs would be given to the Brahms Simulator through Actual Time Simulator. Brahms Simulator behavior and results would be the same as it is designed to generate the 125 filtered alerts. Customer service department can be improved and human errors can be detected on time. Future work involves alert categorization based on urgency and importance. Sometimes employees do not respond even having certain reminders or warnings, there should be some alternative plans available to implement the system. For example, Brahms Model may be connected with restriction of employees online access to internal system databases and salary deduction could be the best option because they do not follow the reminders. Customer service would be more effective, if problems are addressed in its true letter and spirit. [10] A. Ghaffar, M. R. Wahiddin, and A. Shaikh. Computer assisted alerts using mental model approach for customer service improvement. Journal of Software Engineering and Applications, 6(05):21, 2013. [11] A. Ghaffar, M. R. Wahiddin, A. Shaikh, and A. Ahmed. Generating alerts using context aware security and brahms model for customer service improvement. In Communication Technologies, Information Security and Sustainable Development, International Multi Topic Conference, IMTIC 2015, Jamshoro, Pakistan, Feb 11-13, 2015, pages 47–54, 2015. [12] C. Johnson. Visualizing the relationship between human error and organizational failure. In Proceedings of the 17th International Systems Safety Conference, The Systems Safety Society, Unionville, Virginia, United States of America, pages 101–110, 1999. ACKNOWLEDGMENT We are thankful to the Umm Al-Qura University, Makkah, Saudi Arabia and Malaysian Ministry of Higher Education for supporting us through Grant No. ERGS 11-010-0010. Dr. Ghassan Nauman encouragement for the research support is highly appreciated. [13] B. Jordan. Transforming ethnography-reinventing research. Groupware-Software fur die Teamarbeit der Zukunft: Grundlegende Konzepte und Fallstudien, pages 200–212, 1996. [14] M. Noordin. Ict and islam. IIUM Press, 2013. R EFERENCES [15] D. Parmenter. Key performance indicators: developing, implementing, and using winning KPIs. John Wiley & Sons, 2015. [1] S. Almutairi, H. Aldabbas, and A. Abu-Samaha. Review on the security related issues in context aware system. International Journal of Wireless & Mobile Networks (IJWMN), 4(3):195–204, 2012. [16] G. S. Poh, N. N. Abdullah, M. R. Zaba, and M. R. Wahiddin. Reasoning of collaborative human behaviour in securitycriticalwork practices: A framework. In Trustworthy Ubiquitous Computing, pages 99–106. Springer, 2012. [2] J. Blythe, J. Camp, and V. Garg. Targeted risk communication for computer security. In Proceedings of the 16th international conference on Intelligent user interfaces, pages 295–298. ACM, 2011. [17] S. Purao. Design research in the technology of information systems: Truth or dare. GSU Department of CIS Working Paper, 2002. [3] C. Bravo-Lillo, L. F. Cranor, J. Downs, and S. Komanduri. Bridging the gap in computer security warnings: A mental model approach. IEEE Security & Privacy, (2):18–26, 2010. [18] S. Robbins, T. A. Judge, B. Millett, and M. Boyle. Organisational behaviour. Pearson Higher Education AU, 2013. [4] P. Caputi and M. Balnaves. Introduction to Quantitative Research Methods: An Investigative Approach. London, United Kingdom: Sage, 2001. [19] M. Sierhuis and W. J. Clancey. Modeling and simulating practices, a work method for work systems design. Intelligent Systems, IEEE, 17(5):32–41, 2002. [5] J. M. Carroll. Human–computer interaction. Encyclopedia of Cognitive Science, 2009. [20] M. Sierhuis, W. J. Clancey, and T. Hall. intelligent software agent technology. 2009. [6] W. J. Clancey, M. Sierhuis, C. Kaskiris, and R. Van Hoof. Advantages of brahms for specifying and implementing a multiagent human-robotic exploration system. In FLAIRS Conference, pages 7–11, 2003. Nasa deploys [21] M. Sierhuis, W. J. Clancey, and R. J. Van Hoof. Brahms: a multi-agent modelling environment for simulating work processes and practices. International Journal of Simulation and Process Modelling, 3(3):134–152, 2007. [7] M. J. Epstein and M.-J. Roy. Sustainability in action: Identifying and measuring the key performance drivers. Long range planning, 34(5):585–604, 2001. [22] R. H. von Alan, S. T. March, J. Park, and S. Ram. Design science in information systems research. MIS quarterly, 28(1):75–105, 2004. [8] A. Ghaffar, M. R. Wahiddin, M. F. Noordin, and A. Shaikh. Evaluation of tools and techniques for the generation of warning alerts: A survey paper. Transactions on Machine Learning and Artificial Intelligence, 3(2):10, 2015. [23] Wikipedia. Automation. [9] A. Ghaffar, M. R. Wahiddin, M. F. Noordin, and A. Shaikh. A framework to improve customer service using brahms model. IJEIR, 4(1):99–106, 2015. 126 Automation. https://en.wikipedia.org/wiki/