Designing evolvable knowledge graphs - Learning as a mechanism for evolvability
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
Lecture Hall L5 (Oulun Puhelin), Linnanmaa Campus
Topic of the dissertation
Designing evolvable knowledge graphs - Learning as a mechanism for evolvability
Doctoral candidate
Master of Science Anna Teern
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, Software Engineering and Information Systems
Subject of study
Information Processing Science
Opponent
Professor Polyxeni Vassilakopoulou, University of Agder
Custos
Professor Tero Päivärinta, University of Oulu
Experts are drowning in data – an evolvable knowledge graph highlights what matters
Intelligent assistants are already part of many professional settings. They can be used, for example, to monitor the condition of machines and buildings to identify suitable solutions for issues. However, data alone does not make an intelligent assistant useful. To support decision-making in dynamic environments, these digital assistants must be able to learn; to add and modify knowledge in their repositories, such as knowledge graphs (KGs). This doctoral research examines how learning processes can enhance KGs to deliver contextually relevant knowledge to the users.
Hybrid intelligence combines the strengths of humans and artificial intelligence. Machines can process large volumes of data quickly, whereas humans understand broader context and judgement. In many practical settings, these strengths often remain separate: the system processes data, but human experience and judgement do not influence the stored knowledge, making it seem detached or difficult for the user to benefit from. This doctoral research addresses this gap by proposing an evolvable KG that can be expanded and refined on the basis of new data, machine learning, use and expert feedback.
In practice, such solutions are needed in situations where decisions rely on both data and expert interpretation. In organisations, knowledge is generated in systems and documents, as well as through people’s experience and expertise, but these sources do not always converge in decision-making. An evolvable KG can ensure that intelligent assistants have access to relevant knowledge and can update this knowledge as needed.
The doctoral research was conducted as design science research, where solutions are developed through interaction between practical environments and prior research knowledge. The study examined the design objectives, development process and learning approaches of evolvable KGs, and then applied the solution in the context of an intelligent assistant. The empirical settings were industrial maintenance and decision-making in autonomous driving, but the results are intended more broadly for the development of intelligent assistants.
The key message is that the usefulness of an intelligent assistant relies on its ability to learn from its environment and human expertise. When the knowledge base evolves, artificial intelligence can better support human decision-making. The human remains at the centre: an evolvable KG does not replace expert decision-making, but helps to structure and refine stored knowledge for experts to use.
Hybrid intelligence combines the strengths of humans and artificial intelligence. Machines can process large volumes of data quickly, whereas humans understand broader context and judgement. In many practical settings, these strengths often remain separate: the system processes data, but human experience and judgement do not influence the stored knowledge, making it seem detached or difficult for the user to benefit from. This doctoral research addresses this gap by proposing an evolvable KG that can be expanded and refined on the basis of new data, machine learning, use and expert feedback.
In practice, such solutions are needed in situations where decisions rely on both data and expert interpretation. In organisations, knowledge is generated in systems and documents, as well as through people’s experience and expertise, but these sources do not always converge in decision-making. An evolvable KG can ensure that intelligent assistants have access to relevant knowledge and can update this knowledge as needed.
The doctoral research was conducted as design science research, where solutions are developed through interaction between practical environments and prior research knowledge. The study examined the design objectives, development process and learning approaches of evolvable KGs, and then applied the solution in the context of an intelligent assistant. The empirical settings were industrial maintenance and decision-making in autonomous driving, but the results are intended more broadly for the development of intelligent assistants.
The key message is that the usefulness of an intelligent assistant relies on its ability to learn from its environment and human expertise. When the knowledge base evolves, artificial intelligence can better support human decision-making. The human remains at the centre: an evolvable KG does not replace expert decision-making, but helps to structure and refine stored knowledge for experts to use.
Created 16.9.2026 | Updated 17.9.2026