InsurAgent: A Large Language Model-Empowered Agent for Simulating Individual Behavior in Purchasing Flood Insurance

TL;DR

The paper addresses the underutilization of flood insurance among at-risk populations and proposes a framework to model individual decision-making using large language models (LLMs). It builds a benchmark dataset from prior survey work, evaluates LLMs' qualitative and quantitative capabilities, and introduces InsurAgent, a five-module LLM-empowered agent that grounds reasoning in region-specific data via retrieval-augmented generation and encodes temporal dynamics with a memory module. InsurAgent achieves strong alignment with empirical marginal and bivariate probabilities (e.g., $R^2=0.778$; $MAE=0.024$) and demonstrates robust extrapolation to contextual factors beyond the regression baseline, outperforming state-of-the-art LLMs. The memory-enabled, context-aware agent offers a practical tool for behavioral modeling and policy analysis in disaster risk management, while acknowledging limitations in generalizability and the need for broader regional data and validation.

Abstract

Flood insurance is an effective strategy for individuals to mitigate disaster-related losses. However, participation rates among at-risk populations in the United States remain strikingly low. This gap underscores the need to understand and model the behavioral mechanisms underlying insurance decisions. Large language models (LLMs) have recently exhibited human-like intelligence across wide-ranging tasks, offering promising tools for simulating human decision-making. This study constructs a benchmark dataset to capture insurance purchase probabilities across factors. Using this dataset, the capacity of LLMs is evaluated: while LLMs exhibit a qualitative understanding of factors, they fall short in estimating quantitative probabilities. To address this limitation, InsurAgent, an LLM-empowered agent comprising five modules including perception, retrieval, reasoning, action, and memory, is proposed. The retrieval module leverages retrieval-augmented generation (RAG) to ground decisions in empirical survey data, achieving accurate estimation of marginal and bivariate probabilities. The reasoning module leverages LLM common sense to extrapolate beyond survey data, capturing contextual information that is intractable for traditional models. The memory module supports the simulation of temporal decision evolutions, illustrated through a roller coaster life trajectory. Overall, InsurAgent provides a valuable tool for behavioral modeling and policy analysis.

Figures (14)

• Figure 1: Illustration of correct responses of LLM in the third-person discriminative tasks: LLM demonstrates a qualitative understanding of factors influencing individual flood insurance decisions.
• Figure 2: Illustration of LLM failure in first-person role-playing tasks: LLMs fail to capture the differences in insurance purchase probabilities across individual profiles.
• Figure 3: Architecture of the InsurAgent, comprising five modules: perception, retrieval, reasoning, action, and memory.
• Figure 4:
• Figure 5: Representative examples of InsurAgent-simulated marginal probabilities of purchasing flood insurance.