Author:

Abstract:

In the last few decades our society has moved to a knowledge society in which continuous innovation has become more important than ever. Innovation is not the responsibility of R&D departments alone anymore, but is also driven by the knowledge of the company's expert employees. In this thesis we investigate how expert knowledge can be used to develop powerful applications, and how applications should be designed to leverage the expert knowledge to the fullest. Following Liebowitz we look at this expert knowledge at three levels: the domain knowledge level that contains the factual knowledge of the domain, the inference knowledge to know how to execute a specific job, and task knowledge that refers to the ability of decomposing a complex task in subtasks (Liebowitz, 2003). This distinction aligns well with the lens of the Knowledge Base Paradigm that we use. This paradigm advances a strict separation between the factual domain knowledge of a problem, and the inference tasks that use this knowledge to find a solution for a problem (e.g., Van Hertum, 2016). In this distinction, the basic inference tasks are an example of inference knowledge, while selecting the correct inference tasks to solve a specific problem is an example of task knowledge. Gathering these different kinds of knowledge to build a knowledge base system, requires different knowledge acquisition activities. With the help of six case studies we create a partial decision support system classification and derive knowledge acquisition recommendations for each of them. The first category of systems are legal applications. These applications support users in making correct decisions based on statutory law. Of course, correctness and explainability of results are important requirements of legal systems. Contrary to common belief, dealing with legislation is not mainly characterised by deductive reasoning, but mostly by default reasoning with possibly incomplete information. This impacts the development and setup of the application, which should display a high level of interactivity. Product configuration systems are the second kind of applications. They are not limited to the design of tangible products, but can also be used to develop intangible artefacts, such as a course program or communication templates. The help of domain experts to develop a configurator is indispensable, and a configuration project should explicitly focus on taking domain experts in the loop. The third category of applications is scheduling applications. The scientific body of knowledge on personnel scheduling is vast, but its use in practice is limited. We find that not only the quality of the schedule is important, but also the social role that many planners occupy. The last category is an overarching category of third-party applications, in which the company that develops the application does not have access to all knowledge. A more or less significant part of the knowledge is only available for the third party -often an external client- that uses the application. Applications in this category should devote a great deal of attention to making the knowledge base accessible to the user. This thesis clarifies the kinds of projects for which knowledge base systems are best suited, demonstrates their power and makes recommendations on how to develop them efficiently. Liebowitz, J., & Megbolugbe, I. (2003). A set of frameworks to aid the project manager in conceptualizing and implementing knowledge management initiatives. International Journal of Project Management, 21(3), 189-198. https://doi.org/https://doi.org/10.1016/S0263-7863(02)00093-5 Van Hertum, P. (2016). New Language Constructs and Inferences for the Knowledge Base Paradigm: A Business and Multi-agent Perspective. Retrieved from https://lirias.kuleuven.be/bitstream/123456789/551798/1/thesis_0710_Lirias.pdf