Explicitly Mentioned Causal Fact

Explicitly mentioned causal facts in text data are a central focus in current research on causal inference using large language models (LLMs). Researchers are investigating how the frequency and context of causal statements in LLM training data affect their ability to accurately identify and reason about causal relationships, revealing that LLMs often struggle with inferring causality beyond memorized facts and are prone to fallacies like post hoc ergo propter hoc. This work is crucial for improving the reliability and trustworthiness of AI systems in applications where causal understanding is critical, such as decision-making in healthcare or legal contexts, by addressing the limitations of current LLMs in handling nuanced causal information. The development of methods to represent and utilize causal background knowledge is also a key area of investigation to enhance the accuracy and generalizability of causal inference models.

Papers

November 15, 2024

A Survey of Event Causality Identification: Principles, Taxonomy, Challenges, and Assessment
Zefan Zeng, Qing Cheng, Xingchen Hu, Yuehang Si, Zhong Liu
Timely Survey Technical Challenge Comprehensive Taxonomy General Principle Causal Direction Causal Knowledge Event Causality Identification Event Graph Explicitly Mentioned Causal Fact

November 2, 2024

Causal reasoning in difference graphs
Charles K. Assaad
Causal Direction Causal Discovery Method Causal Reasoning Explicitly Mentioned Causal Fact

July 29, 2024

From Pre-training Corpora to Large Language Models: What Factors Influence LLM Performance in Causal Discovery Tasks?
Tao Feng, Lizhen Qu, Niket Tandon, Zhuang Li, Xiaoxi Kang, Gholamreza Haffari
Large Corpus Causal Discovery Causal Relation LLM Performance Causal Discovery Task Explicitly Mentioned Causal Fact

June 18, 2024

LLMs Are Prone to Fallacies in Causal Inference
Nitish Joshi, Abulhair Saparov, Yixin Wang, He He
Medical LLM Causal Inference Causal Relation Causal Inference Task Causal View Explicitly Mentioned Causal Fact

September 25, 2022

AI, Opacity, and Personal Autonomy
Bram Vaassen
Artificial Intelligence Transparency Index Artificial Intelligence Algorithm Algorithmic Decision Making Opaque Machine Learning Ground Glass Opacity Explicitly Mentioned Causal Fact

July 10, 2022

On the Representation of Causal Background Knowledge and its Applications in Causal Inference
Zhuangyan Fang, Ruiqi Zhao, Yue Liu, Yangbo He
Financial Application Causal Inference Individual Representation Directed Acyclic Graph Causal Knowledge Causal DAG Explicitly Mentioned Causal Fact

March 22, 2022

Out-of-distribution Generalization with Causal Invariant Transformations
Ruoyu Wang, Mingyang Yi, Zhitang Chen, Shengyu Zhu
Distribution Generalization Causal Feature Invariant Causal Invariant Causal Feature Explicitly Mentioned Causal Fact Causal Invariance