Master's Thesis presentation, Alexander Reusch Eide och Wassim El Haouzi

Abstract: A legal dispute involves conflicts between two or more parties that arise from disagreements over the interpretation or application of the law. Trying to predict and model the outcomes of said disputes is called using a legal dispute model. Previous works on legal dispute models involve using AI and machine learning to use previous outcomes of legal disputes to make predictions. Instead, in this thesis, a decision tree-based approach is taken in collaboration with Eperoto that is purely user-input based, treating every dispute as unique and uncorrelated to previous rulings of similar disputes. This work aims to optimize Eperoto's model to provide lawyers with a valuable and reliable tool for predicting the outcomes of legal disputes, which can help them provide better advice to their clients.

A computational study involving the construction of sampling schemes based on the Horvitz-Thompson estimation model is performed, along with evaluating their effectiveness in accuracy and computational efficiency compared to Eperoto's decision tree-based approach.

The model proposed in conjunction with some of the sampling schemes achieves remarkable accuracy and computational efficiency compared to the baseline decision tree-based approach. Furthermore, insights regarding limitations and future work on the topic are proposed and discussed to improve similar implementation of legal dispute models.

Supervisor: Erik Kristiansson
Examiner: Serik Sagitov