Understanding Causality with Large Language Models: Feasibility and Opportunities
This work addresses the problem of evaluating and improving LLMs for causal reasoning, which is incremental as it builds on existing capabilities without introducing a new method.
The paper assesses the ability of large language models (LLMs) to answer causal questions, finding they can handle existing knowledge but struggle with discovering new knowledge or high-stakes tasks.
We assess the ability of large language models (LLMs) to answer causal questions by analyzing their strengths and weaknesses against three types of causal question. We believe that current LLMs can answer causal questions with existing causal knowledge as combined domain experts. However, they are not yet able to provide satisfactory answers for discovering new knowledge or for high-stakes decision-making tasks with high precision. We discuss possible future directions and opportunities, such as enabling explicit and implicit causal modules as well as deep causal-aware LLMs. These will not only enable LLMs to answer many different types of causal questions for greater impact but also enable LLMs to be more trustworthy and efficient in general.