-
CausalChat: Interactive Causal Model Development and Refinement Using Large Language Models
Authors:
Yanming Zhang,
Akshith Kota,
Eric Papenhausen,
Klaus Mueller
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 understand…
▽ More
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.
△ Less
Submitted 17 October, 2024;
originally announced October 2024.
-
An Explainable AI Approach to Large Language Model Assisted Causal Model Auditing and Development
Authors:
Yanming Zhang,
Brette Fitzgibbon,
Dino Garofolo,
Akshith Kota,
Eric Papenhausen,
Klaus Mueller
Abstract:
Causal networks are widely used in many fields, including epidemiology, social science, medicine, and engineering, to model the complex relationships between variables. While it can be convenient to algorithmically infer these models directly from observational data, the resulting networks are often plagued with erroneous edges. Auditing and correcting these networks may require domain expertise f…
▽ More
Causal networks are widely used in many fields, including epidemiology, social science, medicine, and engineering, to model the complex relationships between variables. While it can be convenient to algorithmically infer these models directly from observational data, the resulting networks are often plagued with erroneous edges. Auditing and correcting these networks may require domain expertise frequently unavailable to the analyst. We propose the use of large language models such as ChatGPT as an auditor for causal networks. Our method presents ChatGPT with a causal network, one edge at a time, to produce insights about edge directionality, possible confounders, and mediating variables. We ask ChatGPT to reflect on various aspects of each causal link and we then produce visualizations that summarize these viewpoints for the human analyst to direct the edge, gather more data, or test further hypotheses. We envision a system where large language models, automated causal inference, and the human analyst and domain expert work hand in hand as a team to derive holistic and comprehensive causal models for any given case scenario. This paper presents first results obtained with an emerging prototype.
△ Less
Submitted 23 December, 2023;
originally announced December 2023.
-
Patterns of Social Vulnerability -- An Interactive Dashboard to Explore Risks to Public Health on the US County Level
Authors:
Darius Coelho,
Nikita Gupta,
Eric Papenhausen,
Klaus Mueller
Abstract:
Social vulnerability is the susceptibility of a community to be adversely impacted by natural hazards and public health emergencies, such as drought, earthquakes, flooding, virus outbreaks, and the like. Climate change is at the root of many recent natural hazards while the COVID-19 pandemic is still an active threat. Social vulnerability also refers to resilience, or the ability to recover from s…
▽ More
Social vulnerability is the susceptibility of a community to be adversely impacted by natural hazards and public health emergencies, such as drought, earthquakes, flooding, virus outbreaks, and the like. Climate change is at the root of many recent natural hazards while the COVID-19 pandemic is still an active threat. Social vulnerability also refers to resilience, or the ability to recover from such adverse events. To gauge the many aspects of social vulnerability the US Center of Disease Control (CDC) has subdivided social vulnerabilities into distinct themes, such as socioeconomic status, household composition, and others. Knowing a community's social vulnerabilities can help policymakers and responders to recognize risks to community health, prepare for possible hazards, or recover from disasters. In this paper we study social vulnerabilities on the US county level and present research that suggests that there are certain combinations, or patterns, of social vulnerability indicators into which US counties can be grouped. We then present an interactive dashboard that allows analysts to explore these patterns in various ways. We demonstrate our methodology using COVID-19 death rate as the hazard and show that the patterns we identified have high predictive capabilities of the pandemic's local impact.
△ Less
Submitted 7 January, 2023;
originally announced January 2023.
