Abstract:Causal networks are widely used in many fields to model the complex relationships between variables. A recent approach has sought to construct causal networks by leveraging the wisdom of crowds through the collective participation of humans. While this can yield detailed causal networks that model the underlying phenomena quite well, it requires a large number of individuals with domain understanding. We adopt a different approach: leveraging the causal knowledge that large language models, such as OpenAI's GPT-4, have learned by ingesting massive amounts of literature. Within a dedicated visual analytics interface, called CausalChat, users explore single variables or variable pairs recursively to identify causal relations, latent variables, confounders, and mediators, constructing detailed causal networks through conversation. Each probing interaction is translated into a tailored GPT-4 prompt and the response is conveyed through visual representations which are linked to the generated text for explanations. We demonstrate the functionality of CausalChat across diverse data contexts and conduct user studies involving both domain experts and laypersons.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is how to use the knowledge of large - language models (LLMs, such as OpenAI's GPT - 4) to construct and refine causal networks. Specifically, the author aims to overcome two main bottlenecks in traditional causal network construction by combining text - and data - driven methods: 1. **Limitations of the dataset**: Existing datasets are often unable to fully capture all the variables required to build a comprehensive causal model. 2. **Lack of data for newly discovered relationships**: For newly discovered causal relationships, there may be a lack of corresponding data support. To solve these problems, the paper proposes a new method - CausalChat, which explores the causal relationships between individual variables or variable pairs by using the knowledge of LLM through interaction with users, and identifies potential variables, confounding factors, and mediating variables, thereby gradually constructing a detailed causal network. Each interaction is translated into a customized GPT - 4 prompt, and the response is conveyed to the user through a visual representation, accompanied by explanatory text. ### Main contributions - **Recursive text/data - driven causal network development and refinement paradigm**: Provides a practical framework that enables users to gradually build and optimize causal networks through a conversational approach. - **Multi - angle LLM prompts**: Designs a series of prompts to explore hypothesized causal relationships from different angles, in order to enhance confidence in causal relationships and reveal potential conflicts. - **Visualization tools**: Introduces a variety of visual charts to help users efficiently understand complex causal relationships and their potential variables, confounding factors, and mediating variables. - **Feedback loop management**: Has the ability to manage feedback loops in causal reasoning, expanding to multiple causal networks for different decision - making processes. - **Interpretable AI mechanism**: Links visual representations with text generated by LLM to ensure the transparency and interpretability of causal reasoning. - **Interactive visual interface**: Enables analysts to play an active role in recursive text/data - driven causal network development and refinement activities. - **Case studies and usage scenarios**: Demonstrates the application effect and potential value of the system through case studies of experts and ordinary users. In conclusion, this paper provides an innovative and practical method for constructing and refining causal networks by combining the knowledge of LLM and the interaction of human analysts, especially suitable for users without domain - specific knowledge.

CausalChat: Interactive Causal Model Development and Refinement Using Large Language Models

An Explainable AI Approach to Large Language Model Assisted Causal Model Auditing and Development

Towards Human-like Perception: Learning Structural Causal Model in Heterogeneous Graph

Evaluating Large Language Models for Causal Modeling

Causality for Large Language Models

Causal Reasoning and Large Language Models: Opening a New Frontier for Causality

Causal Agent based on Large Language Model

Using GPT-4 to guide causal machine learning

Large Language Model for Causal Decision Making

Large Language Models for Constrained-Based Causal Discovery

Causal Dataset Discovery with Large Language Models

Cause and Effect: Can Large Language Models Truly Understand Causality?

Large Language Models and Causal Inference in Collaboration: A Comprehensive Survey

Narrating Causal Graphs with Large Language Models

LLM4Causal: Democratized Causal Tools for Everyone via Large Language Model

CLEAR: Can Language Models Really Understand Causal Graphs?

Is ChatGPT the Future of Causal Text Mining? A Comprehensive Evaluation and Analysis

Can large language models build causal graphs?

Automating psychological hypothesis generation with AI: when large language models meet causal graph

Counterfactual Causal Inference in Natural Language with Large Language Models

Integrating Large Language Models in Causal Discovery: A Statistical Causal Approach