Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Political Actor Agent: Simulating Legislative System for Roll Call Votes Prediction with Large Language Models

Hao Li, Ruoyuan Gong, Hao Jiang

TL;DR

The paper tackles roll-call vote prediction by modeling legislators as intelligent agents using large language models. It introduces the Political Actor Agent (PAA), which builds scalable profiles, reasons via multi-view planning, and simulates legislative action through an influence mechanism to predict votes. Experiments on the 117th–118th U.S. House demonstrate that PAA, especially the GPT-4o-based variant, achieves superior accuracy and interpretability compared with embedding-based baselines, with ablations showing the profile module as particularly impactful. The approach offers a scalable, human-understandable framework for political science research and can be extended to other countries and data modalities in the future.

Abstract

Predicting roll call votes through modeling political actors has emerged as a focus in quantitative political science and computer science. Widely used embedding-based methods generate vectors for legislators from diverse data sets to predict legislative behaviors. However, these methods often contend with challenges such as the need for manually predefined features, reliance on extensive training data, and a lack of interpretability. Achieving more interpretable predictions under flexible conditions remains an unresolved issue. This paper introduces the Political Actor Agent (PAA), a novel agent-based framework that utilizes Large Language Models to overcome these limitations. By employing role-playing architectures and simulating legislative system, PAA provides a scalable and interpretable paradigm for predicting roll-call votes. Our approach not only enhances the accuracy of predictions but also offers multi-view, human-understandable decision reasoning, providing new insights into political actor behaviors. We conducted comprehensive experiments using voting records from the 117-118th U.S. House of Representatives, validating the superior performance and interpretability of PAA. This study not only demonstrates PAA's effectiveness but also its potential in political science research.

Political Actor Agent: Simulating Legislative System for Roll Call Votes Prediction with Large Language Models

TL;DR

The paper tackles roll-call vote prediction by modeling legislators as intelligent agents using large language models. It introduces the Political Actor Agent (PAA), which builds scalable profiles, reasons via multi-view planning, and simulates legislative action through an influence mechanism to predict votes. Experiments on the 117th–118th U.S. House demonstrate that PAA, especially the GPT-4o-based variant, achieves superior accuracy and interpretability compared with embedding-based baselines, with ablations showing the profile module as particularly impactful. The approach offers a scalable, human-understandable framework for political science research and can be extended to other countries and data modalities in the future.

Abstract

Predicting roll call votes through modeling political actors has emerged as a focus in quantitative political science and computer science. Widely used embedding-based methods generate vectors for legislators from diverse data sets to predict legislative behaviors. However, these methods often contend with challenges such as the need for manually predefined features, reliance on extensive training data, and a lack of interpretability. Achieving more interpretable predictions under flexible conditions remains an unresolved issue. This paper introduces the Political Actor Agent (PAA), a novel agent-based framework that utilizes Large Language Models to overcome these limitations. By employing role-playing architectures and simulating legislative system, PAA provides a scalable and interpretable paradigm for predicting roll-call votes. Our approach not only enhances the accuracy of predictions but also offers multi-view, human-understandable decision reasoning, providing new insights into political actor behaviors. We conducted comprehensive experiments using voting records from the 117-118th U.S. House of Representatives, validating the superior performance and interpretability of PAA. This study not only demonstrates PAA's effectiveness but also its potential in political science research.

Paper Structure

This paper contains 34 sections, 3 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Modeling Political Actors for Voting Prediction
  4. LLM Agent in Decision Simulation
  5. Method
  6. Profile Construction Module
  7. Personal information
  8. Constituency details
  9. Sponsorship activity
  10. Voting records:
  11. Multi-view Planning Module
  12. Trustee view
  13. Delegate view
  14. Follower view
  15. Simulated Legislative Action Module
  16. ...and 19 more sections

Figures (5)

  • Figure 1: Examples of different political actor modeling methods include: The ideal point model represents legislators and bill entities as vectors. The graph-based model embeds nodes from heterogeneous information graphs into vectors using a graph embedding model. Our agent-based model does not rely on distances between embeddings; instead, it directly generates voting outcomes using LLM agents.
  • Figure 2: Framework of PAA
  • Figure 3: The results on the impact of profile length on the performance of the PAA.
  • Figure 4: The consistency experiment results.
  • Figure 5: An example demonstrating how an agent cast a vote and subsequently summarize the reasons for its decision.