978-1-4799-3516-1/13/$31.00 © 2013 41 A Multi-agent Knowledge Expert Locating System: A Software Agent Simulation on Personal Knowledge Management (PKM) Model Shahrinaz Ismail Malaysian Institute of IT Universiti Kuala Lumpur Kuala Lumpur, Malaysia shahrinaz@miit.unikl.edu.my Tran Duc Nguyen College of Graduate Studies Universiti Tenaga Nasional Kajang, Malaysia tnguyen422@gmail.com Mohd Sharifuddin Ahmad College of Information Technology Universiti Tenaga Nasional Kajang, Malaysia sharif@uniten.edu.my Abstract—This paper presents a multi-agent simulation that demonstrates personal knowledge management (PKM) processes among knowledge workers based on a PKM model. Two scenarios are simulated: the mediation of Knowledge Seeker-Knowledge Expert and Knowledge Seeker-Knowledge Collaborator tasks that are animated in an agent-oriented development platform. The results of the simulation validate the PKM model at agent-mediation level. Keywords-software agent; multi-agent system; simulation; personal knowledge management; GUSC I. INTRODUCTION Personal knowledge management (PKM), or KM at individual knowledge workers’ level, has been implemented and practised over the Web 2.0 tools and technologies in the last decade, but the intelligence underlying the PKM technology is somewhat ignored. In contributing to this deficiency, we propose an agent-mediated PKM system based on a model, which we called the Get-Understand- Share-Connect (GUSC) model [1, 2, 3]. In this model, agent’s intelligence is exploited to perform the tasks of managing personal knowledge on behalf of human knowledge workers. This paper presents a part of the work in progress of our research in agent-mediated Personal Knowledge Management. Recent research on PKM suggests that circumstances dictate the need to find knowledge experts within and without the organisations [1]. In technology and knowledge- intensive organisations, the need to seek for new knowledge is compelling, which demands the role of a gatekeeper to identify the relevant and appropriate knowledge experts [2]. While the need persists, locating the right person who knows the right knowledge presents a challenge to the intelligent agent research community. This paper presents the developmental tasks of an agent- based simulation to demonstrate an agent-mediated PKM framework [3]. Intelligent software agents are deployed in the simulation to model the PKM processes, namely the get, understand, share and connect (GUSC) processes in the simulation scenarios. The simulation is based on two scenarios to meet the research objective: To simulate the mediation of tasks from Knowledge Seeker, Knowledge Expert and Knowledge Collaborator to software agents. II. RELATED WORK A. GUSC Model and Personal Knowledge Management Current research in agent-mediated knowledge management (KM) shows promising results of intelligent agents performing tasks on behalf of their human counterparts. At this level of KM, the term used is personal knowledge management (PKM), in which knowledge is managed at individual human knowledge workers’ levels. In a recent study on agent-mediated PKM processes [4, 5], the PKM processes are defined in a cycle of ‘Get/retrieve knowledge’, ‘Understand/analyse knowledge’, ‘Share knowledge’ and ‘Connect to knowledge source’ (i.e. GUSC), which are translated into interactions between humans and agents. These interactions are proven to be possible between human-agent and agent-agent, while easing the human- human interactions [3]. The order of the PKM processes between individuals (i.e. GUSC) is found to be different when the tasks are mediated using software agents. Instead of starting with G (i.e. get/retrieve knowledge), the human-agent interaction starts with C (i.e. connect), and this is followed by S, G and U (i.e. CSGU). The difference in this sequence is due to the “different environments in which knowledge is being translated between tacit and explicit forms” [1]. The first sequence (i.e. GUSC) represents the changes between tacit and explicit forms of knowledge during the interactions within a knowledge worker’s mind in managing personal knowledge, whereas the second sequence (i.e. CSGU) represents the processes when software agents are used to mediate the task of finding knowledge experts on behalf of the human knowledge worker [1]. In deploying the GUSC model, software agents are found to have the capabilities that PKM processes would need. Comparison analysis is made on the capabilities of software agents based on the definitions given by authors of the past two decades against the GUSC processes. Consequently, the challenge of making the agents to meet the expected capabilities is resolved by assigning the agents with the GUSC roles [5]. For example, software agents should be able to “engage in dialogues and negotiate and coordinate the transfer of information” [6] for which the processes of Get, Share and Connect are found to be the required PKM processes. On the other hand, the ability of software agents to “carry out some set of operations on behalf of a user or