Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FÆRDXEL: An Expert System for Danish Traffic Law

Luís Cruz-Filipe, Jonas Vistrup

TL;DR

FÆRDXEL addresses the challenge of applying AI in the legal domain by offering an explainable, symbolically reasoned tool for Danish traffic law. It encodes the law in an enriched Datalog knowledge base and uses SLD-resolution to generate refutations with navigable natural-language explanations, aiding legal professionals rather than delivering judgments. Empirical evaluation shows partial alignment with court outcomes and expert feedback points to substantial potential while highlighting usability and knowledge-capture challenges. The work points to practical paths for deployment and future extensions, including sentencing reasoning with fuzzy logic and improved Natural Language to Datalog translation techniques.

Abstract

We present FÆRDXEL, a tool for symbolic reasoning in the domain of Danish traffic law. FÆRDXEL combines techniques from logic programming with a novel interface that allows users to navigate through its reasoning process, thereby ensuring the system's explainability. Towards the goal of better understanding the value of FÆRDXEL, two evaluations of the system have been performed: (1) An empirical evaluation showing that for a selection of court cases, the conclusions of FÆRDXEL align with those of Danish judges. (2) A qualitative evaluation from legal experts indicating that this work has potential to become a foundation for real-world AI tools supporting professionals in the Danish legal sector.

FÆRDXEL: An Expert System for Danish Traffic Law

TL;DR

FÆRDXEL addresses the challenge of applying AI in the legal domain by offering an explainable, symbolically reasoned tool for Danish traffic law. It encodes the law in an enriched Datalog knowledge base and uses SLD-resolution to generate refutations with navigable natural-language explanations, aiding legal professionals rather than delivering judgments. Empirical evaluation shows partial alignment with court outcomes and expert feedback points to substantial potential while highlighting usability and knowledge-capture challenges. The work points to practical paths for deployment and future extensions, including sentencing reasoning with fuzzy logic and improved Natural Language to Datalog translation techniques.

Abstract

We present FÆRDXEL, a tool for symbolic reasoning in the domain of Danish traffic law. FÆRDXEL combines techniques from logic programming with a novel interface that allows users to navigate through its reasoning process, thereby ensuring the system's explainability. Towards the goal of better understanding the value of FÆRDXEL, two evaluations of the system have been performed: (1) An empirical evaluation showing that for a selection of court cases, the conclusions of FÆRDXEL align with those of Danish judges. (2) A qualitative evaluation from legal experts indicating that this work has potential to become a foundation for real-world AI tools supporting professionals in the Danish legal sector.
Paper Structure (15 sections, 1 equation, 4 figures, 2 tables)

This paper contains 15 sections, 1 equation, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. This work.
  3. Structure of the paper.
  4. Related Work
  5. Background
  6. Datalog
  7. SLD-resolution
  8. Specifics of the Danish court system
  9. Translation from Law to Rules
  10. Inference System and Explainability
  11. Interface
  12. Evaluation and Target Users
  13. Evaluation of correctness.
  14. Evaluation of social value.
  15. Conclusions and Future Work

Figures (4)

  • Figure 1: Explanation in færdXel. $\alpha$-nodes are explanation-nodes and $\beta$-nodes are reasons-nodes.
  • Figure 2: Four proof trees created by querying on Example \ref{['ex:case']}.
  • Figure 3: Explanation generated by færdXel for query on Example \ref{['ex:case']}.
  • Figure 4: Illustrations of the interface. The three horizontal dots at the bottom of a display box informs the user that additional items can be found by scrolling.

Theorems & Definitions (3)

  • Example 1
  • Example 2
  • Example 3