用大型语言模型理解因果关系：可行性与机遇
Understanding Causality with Large Language Models: Feasibility and Opportunities
AI总结：
本文评估了LLMs回答三类因果问题的能力，指出其可结合现有知识作答但难以发现新知识或支持高风险决策，并提出显式/隐式因果模块与深度因果感知LLMs等未来方向。
摘要：
我们通过分析大型语言模型（LLMs）在三种类型因果问题上的优势与不足，评估其回答因果问题的能力。我们认为，当前的LLMs能够凭借现有的因果知识，以组合领域专家的方式回答因果问题。然而，它们尚不能为发现新知识或高精度的高风险决策任务提供令人满意的答案。我们讨论了未来可能的方向与机遇，例如实现显式与隐式因果模块以及深度因果感知的LLMs。这些不仅将使LLMs能够回答多种不同类型的因果问题以产生更大影响，还将使LLMs在总体上更值得信赖且更高效。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
