Abstract
This dissertation presents KERAIA, a framework for symbolic knowledge engineeringdesigned to address the challenges of representing, reasoning with,
and executing knowledge in dynamic and complex environments. The central
research question is: How can unstructured human expertise be transformed
into algorithms that AI can efficiently utilize? This research advances symbolic
knowledge engineering by addressing this question through innovative
representation and reasoning techniques.
The contributions of KERAIA are twofold. Firstly, it introduces a robust software
platform that supports extensive knowledge acquisition and a language for representing
human mental models. This platform enables seamless collaboration
between human experts and AI systems, bridging the gap between abstract AI
concepts and their practical applications. Secondly, its applicability is demonstrated
through three case studies: naval warfare scenarios, diagnostics in water
treatment plants, and strategic decision-making in the game of RISK. These case
studies highlight the versatility and relevance of KERAIA in both the real-world
and simulated domains.
The core innovations of KERAIA include Clouds of Knowledge, which integrate
diverse knowledge sources, and Knowledge Lines (KLines) and Lines
of Thought (LoTs), which enhance transparency and traceability in decision
making. Dynamic Relations (Drels) extend traditional ontologies by enabling
context-sensitive property sharing, while Forks, Cloud Dimensions, and Junctures
facilitate multidimensional knowledge structures. Ultragraphs enrich
the knowledge acquisition process through interactive multimedia, and the Knowledge Elicitation and Validation Index (KEVI) benchmarks AI systems for
effectiveness and accuracy.
By addressing some of the limitations of earlier AI initiatives and adhering to
explainable AI principles, KERAIA aims to contribute meaningfully to symbolic
knowledge engineering by fostering human-machine collaboration and
enhancing the practical utility of AI systems.
| Date of Award | 18 Jun 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | The Anh Han (Supervisor), Alessandro Di Stefano (Supervisor) & Claudio Angione (Supervisor) |