FRED: Ontology-based Agents for enabling E-Coaching Support in a large Company Peter Smolle York Sure Net Dynamics Institute AIFB, University of Karlsruhe Prinz-Eugen-Str. 68-70 Postfach A-1040 Vienna, Austria D-76128 Karlsruhe, Germany peter.smolle@netdynamics-tech.com sure@aifb.uni-karlsruhe.de ABSTRACT example for a FRED ontology and ends with a description We present FRED, an ontology-based agent and it’s appli- for which purposes ontologies currently are explored in the cation in an E-Coaching scenario at a large company. We FRED platform. We present briefly our cost-benefit analy- illustrate the architecture and underlying technology of our sis for the real world “Coaching FRED” application to show agent platform, e.g. ontologies, and present our methodol- the commercial value of an agent-based system in Section 5. ogy for ontology development as well as a brief cost-benefit Before concluding we give a brief discussion of related work. analysis, thus showing also commercial aspects. 2. E-COACHING SUPPORT FOR A LARGE Keywords COMPANY Agent, E-Learning, Ontology 2.1 Background A large company in the utility area with approximately 1. INTRODUCTION 20.000 employees is in transition phase from state owned to- Intelligent agents have become an important software wards privatization. Most of the employees have long term paradigm over the last two decades. Although there exists civil servant behavior, which slows down transition speed. plenty definitions of what agents are (cf. e.g. [2], [5]), one The general managers addresses a clear people development might focus on major roles for intelligent agents1 including strategy: “Our employees are the most important assets of e.g. (i) the “Human Surrogate” that works autonomously the company. Its our aim to know their skills and to de- without human direction in an actual or simulated environ- velop them in such a way that they become a self driven ment and utilizes thereby the capability of intelligent agents motivated work force. In doing so they will contribute sig- to reason in a simple, rational manner and finally reports nificantly to the success of our company in the future.” The back results to humans, (ii) the “Intelligent Assistant” that people education department was given the responsibility supports humans in complex environments by performing to execute this skill development strategy in the most em- tasks in cooperation with the human, and (iii), more gen- ployee driven way. They decided to use a new agent based eral, the “Architectural Paradigm” for a software system platform which enabled the building of so called personal that must integrate disparate subsystems. development agents which will act like coaches – FRED. There exist numerous agent based applications for vari- 2.2 Objectives of the project ous purposes and an active research community2 . A large research project is the DARPA Agent Markup Language The key objectives were set to reflect the mentioned strat- (DAML)3 effort, it aims at developing a language and tools egy: (i) Support the skill-transition strategy, (ii) bring active to facilitate the concept of the Semantic Web, in particular information towards employee, (iii) improve service level for to provide a language for agents to facilitate communication large employee groups, (iv) support the education staff in through machine processable semantics (cf. [3]) provided by reducing routine-tasks and (v) optimize the education pro- ontologies. cess. In this paper we present FRED, an ontology-based agent 2.3 The coaching process and it’s application in an E-Coaching scenario at a large The “Coaching FRED” is an agent based application that company. The outline of this paper is as follows. We start is accessible through the intranet of the company. It sup- in Section 2 by illustrating our motivational scenario, i.e. ports employees to organize and coordinate their life long E-Coaching support for a large company. We continue by learning process. The Coaching FRED aims at increasing explaining the underlying system architecture of the FRED information dissemination of existing courses through de- platform in Section 3. Section 4 describes the ontology engi- livering the right course offer to the right employee at the neering environment and the applied methodology, gives an right moment. Therefore each employee might access his 1 cf. http://www.agent-software.com.au/ personal FRED through the intranet. Using the Coaching 2 cf. e.g. http://agents.umbc.edu/ and FRED starts with profiling the personal assistant by pro- http://www.agentlink.org/ viding main topics of an employees tasks and interests. The 3 cf. http://www.daml.org/ profiling tasks is mandatory and the profiling of interests Net Dynamics Internet Technologies developed an ontol- ogy based software platform for delegation -FRED- popu- lated by intelligent software agents which act on their own- ers behalf to address the following challenges (ordered from more general to more specific challenges): (i) Web content is by far faster growing than the amount of users, (ii) large parts of the content will not be usable because of the lack of security and easy to understand semantic based access, (iii) content suppliers want better methods to enhance their suc- cess in E-Commerce, (iv) reduce costs by using the power of Figure 1: FRED solution concept agents technology to process tasks and workflows, (v) cre- ate a large and robust, scalable and secure platform which is mandatory. Immediately after this easy-to-go first step is able to execute in production environments and will be the Coaching FRED starts looking for appropriate courses. of use for a wide range of application areas which could Naturally all information given to the Coaching FRED are benefit from the delegation principle and (vi) enable access stored safely and secretly through a security mechanism. to agents through mobile devices and browsers and make The coaching process consists of eight steps resulting in a use of coming up technologies like UMTS or Blue Tooth. cyclic process: The FRED architecture addresses those challenges by using (1) Initialize Coaching FRED. (2) The employee creates standards to create new semantic and ontology based meth- a personal task profile. (3) Coaching FRED offers the em- ods which will then enable the benefits of delegation. The ployee topics for courses. (4) Optionally, the employee cre- FRED architecture also enables a very productive way for ates a personal interest profile. (5) Coaching FRED offers building small reusable FRED applications, which will re- additional topics for courses. (6) Optionally, the employee duce development and integration effort for process oriented gives feedback to the Coaching FRED in form of relevant tasks significantly. This is done by using the development topics for courses that do not appear in Coaching FREDs power of ontology based smart objects to build intelligent offering. (7) Coaching FRED informs the employee about agents which are able to execute their tasks autonomously for him relevant courses from the course offerings of the com- and can communicate with each other in a unambiguous pany. (8) The employee is free to change her profile any time way of mutual understanding using the FIPA ACL4 agent and to start the process again with (1). communication language. Currently the main task of Coaching FRED is to create personalized course offerings according to an employees pro- file. For the future this might be extended easily by addi- 3.2 Key technologies of the FRED platform tional tasks like getting official permissions for attending To establish the FRED Platform with its capabilities, we courses or registering for courses. have developed new concepts and methods: Smart Objects. All information within FRED is stored 2.4 The solution concept and exchanged as Smart Objects. Smart Objects are based All different FRED types were developed using the capa- on ontologies, they are dynamic, reusable and can represent bilities of the platform. In our scenario we have types for their content in various forms. They have built-in privacy users (e.g. the staff members), courses and education tasks. mechanisms to make sure that data will only be passed from Each FRED type has it’s own ontology for communication one FRED to another according to the privacy profile of a (cf. Section 4). FREDs are populated with core data about FRED’s owner. Main features of the Smart Objects are the the users and then given to every employee and to the related following: (i) Implement “Real world view” instead of education staff members. Courses and education actions are “data model view”, which allows for sharing and reuse of represented by FREDs. objects in different domains, (ii) cover instances of objects All FREDs are hosted on the FRED platform implemented and constraints, (iii) have build in security features, which at the computing center of the company. The access to are implemented as Smart Objects Security Policies, for ex- FRED is given via the intranet browser environment. Once changing information between FREDs, (iv) Meta Data a FRED gets initialized, users have to register and the coach- (e.g. “Importance”) for reasoning (v) support of multi- ing process described above starts to work. ple languages, (vi) strategy based persistence supports Figure 1 shows the solution developed for this scenario windowing, delayed serialization, high performance persis- which contains in a nutshell the following items: (i) Each tence, etc., (vii) “High Level Introspection” supports FRED-Type (e.g. Staff or Course) represents a role of an AI-techniques (inference engine, reasoning systems etc.) and user or a process, (ii) a FRED Platform hosts the different (viii) Java based components, suited for graphical manip- FRED types with their Application Plans, (iii) Visualizer ulation. is the standard interface towards users of the system (typi- Meeting Rooms. Interacting with each other, two or cally via a browser by using http) and (iv) Tools Connect more FREDs perform their tasks in meetings, held in FRED manages access to existing databases (e.g. pre-existing em- Meetings Rooms. These meetings rooms ensure controlled ployee and course databases). The technical details of each and secure execution of FREDs tasks, they are scalable and component will be described in the following section. optimized to perform as many meetings as possible to give FREDs the chance to meet as many FREDs as possible to 3. FRED ARCHITECTURE achieve the best results. 3.1 Preface 4 cf. http://www.fipa.org/repository/aclspecs.html FRED Control. The potentially high number of FREDs in a FRED Location needs an efficient control mechanism. FRED Control provides robustness and high availability to the FRED Platform. In addition to these technologies a FRED Location uses standard state of the art technologies. Java. The development framework of FRED and appli- cation specific parts have been developed in JAVA. Critical components have been designed together with Sun Microsys- tems5 . Agent-Technology. The proven concepts of agent tech- nology6 are the base technology for communication and in- teraction of FREDs. Ontology. The Section 4 describes the ontology engineer- ing environment the underlying OntoEdit and the applied methodology for developing ontologies for FRED. 4. ONTOLOGY DEVELOPMENT Ontologies [1] aim at capturing domain knowledge in a Figure 2: Ontology development for FRED with On- generic way and provide a commonly agreed understanding toEdit of a domain, which may be reused, shared, and operational- ized across applications and groups. Thus, ontologies are for ontology development, viz. (i) ontology kickoff (basically well-suited for enabling communication between agents in a requirements specification), (ii) refinement, and (iii) eval- general, including software agents as well as human agents [4]. uation. However, because of their size, their complexity and their Firstly, all requirements of the envisaged ontology are col- formal underpinnings ontologies are still far from being a lected. Typically for ontology engineering, ontology engi- commodity. Developing ontologies is a non-trivial task. We neers and domain experts are joined in a team that works relied on a well-known ontology engineering environment ac- together on a description of domain and goal of the ontology, companied by a methodology for ontology development. design guidelines, available knowledge sources (e.g. reusable ontologies and thesauri etc.), potential users and use cases Ontology engineering environment. OntoEdit7 [6] sup- and applications supported by the ontology. The output ports the collaborative development of ontologies by using of this phase is a semi-formal description of the ontology. graphical means. OntoEdit is built on top of a powerful in- Secondly, during the refinement phase the team extends the ternal ontology model. This paradigm supports representation- semi-formal description in several iterations and formalizes language neutral modeling as much as possible for concepts, it in an appropriate representation language. The output relations, attributes, instances and axioms. Several graphi- of this phase is a mature ontology (aka. “target ontology”). cal views onto the structures contained in the ontology sup- Thirdly, the target ontology needs to be evaluated accord- port modeling the different phases of the ontology engineer- ing to the requirement specifications. Typically this phase ing cycle. serves as a proof for the usefulness of developed ontologies How do our ontologies look like? OntoEdit enables the and may involve the engineering team as well as end users user to edit (i) an is-a hierarchy of concepts or classes (e.g. of the targeted application. The output of this phase is an Employee is-a Person), (ii) relations between concepts (e.g. evaluated ontology, ready for the roll-out into a productive Employee works at Organization), (iii) attributes attached environment. to concepts (e.g. Person has name STRING), (iv) instances of concepts (e.g. Mary instance of Person) and (v) axioms 4.1 Ontologies for FRED build on top. The concepts may be abstract or concrete, Ontologies are explored in the FRED platform for mainly which indicates whether or not it is allowed to make direct two aspects: (i) enabling communication between different instances of the concept. Each concept is uniquely identified FREDs and (ii) defining security guidelines for a FRED but may have several names, which essentially is a way to world. Figure 2 shows an example FRED ontology de- define synonyms for that concept. Also, multiple languages veloped with OntoEdit. On the left side the concept hi- are supported by that feature. The same holds for relations erarchy is shown. On the right side attributes and rela- and attributes. The tool allows similar to the well-known tions (with their ranges) are presented for a selected con- “copy-and-paste” functionality the reorganizing of concepts cept (here: Course). This particular ontology is the basis within the hierarchy. An example ontology is shown in Sub- for communication with “Course FREDs”. It defines all rel- section 4.1. evant concepts and relations known by these FREDs. In Methodology for ontology development Concerning general, each FRED type has it’s own ontology for com- the methodology8 , OntoEdit focuses on three main steps munication. Shared concepts and relations enable different 5 kinds of FRED-Types to communicate with each other. cf. http://java.sun.com/ The security guidelines define which kind of information 6 cf. e.g. http://www.fipa.org/ is allowed for exchange between FREDs according to the 7 OntoEdit is available from Ontoprise GmbH, cf. http://www.ontoprise.com. 10132 project On-To-Knowledge, a detailed description of 8 The methodology was initially developed in the EU IST- the methodology can be found in [7]) profile defined by users. One example are different levels 7. CONCLUSION of authorization through users. A FRED might be autho- We presented FRED, an ontology based agent and it’s rized to look for offerings and return appropriate ones to the application in an E-Coaching scenario of a large company. user or to look for offerings and book an appropriate one. The key objectives of our implemented system are: (i) Sup- Each security profile is instanciated according to a “security port the skill-transition strategy, (ii) bring active informa- ontology” that contains the security guidelines. Different tion towards employees, (iii) improve the service level for platforms might have different security guidelines. large employee groups, (iv) support the education staff in For the future there might be FREDs that travel across reducing routine-tasks, and (v) optimize the education pro- borders of FRED Platforms. Ontologies provide a shared cess. Our system explores ontologies mainly for two pur- understanding of domains of interest and are potentially poses: (i) enabling communication between different FREDs valuable to support the mapping tasks in this even more and (ii) defining security guidelines for a FRED world. On- complex scenario. tologies for FREDs are engineered according to a well-known methodology with the help of the ontology engineering en- 5. COST-BENEFIT ANALYSIS vironment OntoEdit. Our real world application is highly scalable and is tar- Attracting industrial customers for such an application re- geted at serving potentially 20.000 users. A cost-benefit quires a detailed comparison of costs and benefits, typically analysis for our project resulted in a break even during the having a strong positive benefit as a requirement for a pur- first year and approximately 4.6 Mio EUR total benefits af- chase order. A cost-benefit analysis is an approach to show ter 3 years for the entire company. the methodology which has been applied at this customer. For the future the company will expand it’s intranet but The assumptions and numbers are therefore associated with also it’s internet websites with attractive delegation offer- this special case only and cannot be transferred to other sit- ings. Internally, i.e. through the delegation tasks provided uations without having a basic understanding of the special within the intranet, the goal is to optimize the life long learn- circumstances. A tight cooperation with our customer led ing process of employees. Externally, i.e. through the dele- to the following results (a detailed description is not within gation tasks provided on the internet, the goal is to improve the scope of this paper). The benefits are based on cur- the customer relationship management by personalized of- rent known efforts and to achieve improvements which will ferings for each customer and by creating an innovative ser- lead to manpower savings to be expected because of dele- vice image in general. gating tasks to FREDs. The number of education activities or courses are in the magnitude of 1.000 in this company. In particular, benefits are achieved in the following areas: (i) 8. ACKNOWLEDGEMENTS improving the productivity of the education staff, (ii) reduc- Research for this paper was partially funded by EU in the ing the time for finding optimized education, (iii) targeted project IST-1999-10132 “On-To-Knowledge”. We would like information about education and (iv) optimizing course at- to thank all colleagues at Net Dynamics and the Institute tendance. The cost part represents a cumulated number AIFB for their lively discussions. Especially we would like and no detailed calculations, to make the order of magnitude to thank our partner Ontoprise (Karlsruhe, Germany), who of the real savings visible. We took two sets of employees is provider of the underlying ontology based technologies. numbers as a basis: an initial set of 4.000 employees orga- nized in 200 units with 16 members of the educational staff 9. REFERENCES for the first phase of the implementation and an expanded [1] D. Fensel. Ontologies: Silver Bullet for Knowledge set of 20.000 employees organized in 1.000 units with 50 Management and Electronic Commerce. Springer members of the educational staff. The break even of the Verlag, Berlin, 2001. project calculated with a base of 4.000 employees is dur- [2] S. Franklin and A. Graesser. Is it an agent, or just a ing the second year. With a base of 20.000 employees the program?: A taxonomy for autonomous agents. In break even is already during the first year. The total bene- Proceedings of the Third International Workshop on fit after 3 years for the entire company will be approximately Agent Theories, Architectures, and Languages. 4.6 Mio EUR. Springer-Verlag, 1996. [3] J. Hendler. Agents on the web. IEEE Intelligent Systems, Special Issue on the Semantic Web, 6. RELATED WORK 16(2):30–37, March 2001. E-Learning by itself addresses more the use of technology [4] S. Luke, L. Spector, D. Rager, and J. Hendler. for teaching where E-Coaching has the power to represent Ontology based web agents. In Proceedings of the First a “teacher” in the process. Though there is plenty of work International Conference on Autonomous Agents, 1997. available, which describes software agent research and ap- [5] S. J. Russell and P. Norvig. Artificial Intelligence: A plications, the area of using intelligent agents for coaching Modern Approach. Prentice Hall, Englewood Cliffs, NJ, an education process is just about to evolve. The interests USA, 1995. increase due to the achievable productivity for very large [6] Y. Sure, M. Erdmann, J. Angele, S. Staab, R. Studer, communities. Still, this research area is rather new. A and D. Wenke. OntoEdit: Collaborative ontology planned workshop at Carnegie Mellon University9 will be development for the semantic web. In Proceedings of of help to get an overview on current research in the area of the International Semantic Web Conference 2002 using intelligent agents for coaching. (ISWC 2002), June 9-12 2002, Sardinia, Italia., 2002. 9 “Coach Agent-, and Multi-Agent Modeling Workshop”, [7] Y. Sure and R. Studer. On-To-Knowledge Methodology, planned for June 2002 at Carnegie Mellon University, USA. chapter 3. 2002.