LAMDA: Large Language Model as Decision Analyst
Talk, Oral presentation, 2025 INFORMS annual meeting, Atlanta, Georgia
Influence diagrams address the challenges of decision-making under risk by structuring information, decisions and values, while clearly depicting uncertainties and probabilistic dependencies. However, constructing an influence diagram requires expertise in decision analysis and is further complicated by the need to process large amounts of contextual information. This work automates the construction of influence diagrams from natural language input, leveraging large language models (LLMs). We design a workflow to prompt LLMs to output elements of an influence diagram and resolve issues through verification and regeneration. We also construct a new dataset of typical decision problems under risk. Evaluations using this dataset demonstrate that our framework effectively identifies key factors and relationships in natural language, making better decisions than standalone LLMs and LLMs enhanced with standard techniques such as chain-of-thought (CoT). The dataset and code are available at this URL. Furthermore, we extend the framework to aggregate opinions from multiple sources and apply the aggregation method to discussions by groups of disease control experts on a hypothetical pandemic scenario.