Argumentation for Legal Reasoning: Meta-models, Technology and Beyond

This thesis presents a comprehensive exploration of argumentation in the context of legal reasoning, bridging the gap between formal argumentation theory and its technological applications. Central to this work is the enhancement of the ASPIC+ framework, integrating structured meta-argumentation to address limitations in reasoning about rules, conflicts, and preferences, including the concept of the burden of persuasion. This advancement expands the framework’s applicability in legal reasoning and beyond.

A pivotal aspect of this research is the development of the arg2p framework, a robust and versatile environment integrating theoretical advancements in argumentation. The framework marks a significant stride in realising practical, logic-based environments for argumentation in intelligent systems, demonstrating a marked focus on user-friendliness and technical maturity, crucial for bridging theoretical innovation with functional application.

The thesis also delves into the realm of machine learning (ML), illustrating the integration of structured argumentation with automated machine learning (AutoML). This integration is aimed at enhancing the transparency and control in the development of ML systems by offering a symbolic interface for incorporating expertise in ML, exemplifying the convergence of traditional symbolic AI methods with data-driven ML approaches.

This work significantly contributes to argumentation theory and legal AI, providing a nuanced understanding of meta-argumentation and its practical applications. The enhancements to ASPIC+, coupled with the arg2p framework, present new avenues for legal analysis and decision-making. The integration with ML further highlights the potential of structured argumentation in contemporary AI, paving the way for more robust and ethically sound AI systems across various domains.