Exploring LLM Capabilities to Explain Decision Trees
Date: February 08, 2024
Presenter: Paulo Bruno De Sousa Serafim
Location: Officine Garibaldi - il Cantiere delle Idee, Via Vincenzo Gioberti, 39, 56124 Pisa PI, Italy
Seminar by visiting PhD student Paulo.
Abstract: Decision Trees are widely adopted in Machine Learning tasks due to their operation simplicity and interpretability aspects. Nonetheless, following the decision process path taken by trees can be difficult in a complex scenario or in a case where a user has no familiarity with them. In order to facilitate understanding for non-expert users, there is a recent growing effort to use natural language explanations. In this presentation, we explore the proficiency of Large Language Models (LLMs) to explain decision tree predictions by generating natural language explanations. Based on early experiments, our goal is to identify capabilities that strengthen LLMs as an effective solution as well as highlight potential limitations, foreseeing further exploration possibilities.