LegiGPT : Party Politics in Transport Policy with a Large Language Model

Hyunsoo Yun, Eun Hak Lee · 2025

Read LegiGPT : Party Politics in Transport Policy with a Large Language Model by Hyunsoo Yun, Eun Hak Lee (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-16692.txt.

Introduction

Hyunsoo Yun and Eun Hak Lee build a tool for reading party politics in transport bills in this 2025 paper. Lawmakers' ideologies shape the laws they back, and transport policy is no exception. The authors combine a large language model with explainable AI methods.

Working with bills from South Korea's twenty-first National Assembly, they use the language model to pick out transport bills through keyword, sentence, and context filters. Explainable methods then show which sponsor traits track with outcomes. The shares of conservative and progressive sponsors mattered most, along with district size and voter population. Both camps, the authors find, fed into bipartisan bills in different ways, one side by proposing and the other by supporting.

Catalog record

CREATORHyunsoo Yun, Eun Hak Lee
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-16692

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

LegiGPT : Party Politics in Transport Policy with a Large Language Model Hyunsoo Yun, Eun Hak Lee Abstract Given the significant influence of lawmakers’ political ideologies on legislative decision-making, understanding their impact on policymaking is critically important. We introduce a novel framework, LegiGPT , which integrates a large language model (LLM) with explainable artificial intelligence (XAI) to analyze transportation-related legislative proposals. LegiGPT employs a multi-stage filtering and classification pipeline using zero-shot prompting with GPT-4. Using legislative data from South Korea’s 21st National Assembly, we identify key factors—including sponsor characteristics, political affiliations, and geographic variables—that significantly influence transportation policymaking. The LLM was used to classify transportation-related bill proposals through a stepwise filtering process based on keywords, phrases, and contextual relevance. XAI techniques were then applied to examine relationships between party affiliation and associated attributes. The results reveal that the number and proportion of conservative and progressive sponsors, along with district size and electoral population, are critical determinants shaping legislative outcomes. These findings suggest that both parties contributed to bipartisan legislation through different forms of engagement, such as initiating or supporting proposals. This integrated approach provides a valuable tool for understanding legislative dynamics and guiding future policy development, with broader implications for infrastructure planning and governance. Keywords: party politics; legislative bill; transport policy; lawmaker; large language model; explainable artificial intelligence 1 Introduction Transportation legislation significantly shapes infrastructure, economic growth, and public welfare. In democratic systems, political ideologies influence these policies by guiding funding decisions and strategic priorities—e.g., rural vs. urban investments or highways vs. public transit sorensen2014political . In South Korea, this ideological divide is spatially evident: conservatives dominate in Gyeongsang, progressives in Jeolla chae2008conservatives ; yun2024 . Within Seoul, Gangnam supports conservatives, while other districts lean progressive. These divides manifest not only in policy content but also in legislative language, reflecting party values and electoral goals. Recent studies highlight how political dynamics shape transportation legislation, particularly in funding priorities and project approvals klein2022political ; legacy2019consensus ; mcandrews2015politics ; sciara2024state . For example, federal budget analyses show that states aligned with the ruling party often receive more infrastructure funding. A 2024 report found that 73% of funds (~$7.8B) went to Democrat-led states, despite their being fewer in number christenson2024 . Ideological divides also affect support for sustainable initiatives: conservative administrations tend to boost rural infrastructure, while progressive ones favor electric vehicles and urban transit usafacts2024 . Voting patterns further illustrate this divide. The 2021 Infrastructure Investment and Jobs Act, which allocated $1.2 trillion for projects like high-speed rail, was widely supported by progressives but opposed by conservatives prioritizing highway expansion and fossil fuel infrastructure usdot2023 . As such, statistical trends show that political ideologies shape both the content and direction of transportation legislation, leading to disparities in resource allocation and reflecting deeper ideological divisions. Prior studies have examined this relationship through surveys. For instance, Klein et al. klein2022political found that conservatives tend to support highway expansion, while progressives favor public transit. Christiansen christiansen2018public further showed that opposition to restrictive policies like local tolls can erode satisfaction with democratic institutions, highlighting the link between policy design, public attitudes, and political legitimacy. While informative, these survey-based approaches face challenges, including low response rates, bias, and resource constraints. Given these challenges, many studies have used text mining to extract insights from large-scale unstructured data in the transportation sector. Applications range from analyzing passenger reviews to inform resilient policies park2022toward , to identifying key planning factors from policy documents chowdhury2023investigation , and measuring user satisfaction via social media sentiment zou2024b . Chen et al. chen2024exploring also traced shifts in port policy using text mining. However, traditional methods often rely on simplified linguistic features, limiting their ability to capture complex structures and evolving terminology. Large language models (LLMs) provide an alternative. Their pre-trained understanding enables zero-shot analysis of legislative texts without labeled data kuila2024deciphering ; leong2024metroberta ; lim2025large ; tornberg2024large ; kim2025linear . LLMs offer three main advantages: (1) efficient processing of large unstructured datasets, (2) reduced human bias, and (3) the ability to detect latent patterns and contextual nuances wang2025 ; wei2022b . They also generalize across tasks without task-specific training wei2022a ; zhang2024 . Despite these benefits, the use of LLMs to systematically analyze how political ideologies influence transportation legislation remains limited—highlighting the need for further research in this area. This study aims to understand how political ideology shapes the proposal patterns, sponsor composition, and approval outcomes of transportation-related legislation. To achieve this, we developed LegiGPT (Legislative GPT), which integrates two key components: (1) a LLM is used to extract and refine transportation-related bills from legislative texts through a multi-stage filtering process; and (2) explainable AI (XAI) techniques are applied to model and interpret the relationships between political party affiliation and key legislative, demographic, and geographic attributes. The results reveal the most influential attributes and offer insights into their interactions. This integrated approach demonstrates the potential of AI-driven methodologies for uncovering latent patterns in transportation policymaking influenced by political ideologies. • We develop LegiGPT , a novel framework that integrates LLMs with explainable artificial intelligence (XAI) techniques to analyze transportation-related legislative proposals. • We propose a multi-stage classification approach that leverages zero-shot prompting to filter transportation bills from raw legislative texts without requiring labeled data. • We reveal key political and geographic factors, such as party affiliation, sponsor ideology, and electoral population, that shape legislative outcomes in transportation policymaking. 2 Data The open data portal in South Korea provides a centralized platform for accessing a wide range of government datasets data_go_kr . It supports transparency and innovation by offering open data across various sectors, including legislation, transportation, and public administration. Among the available datasets, two were used in this study. First, the National Assembly’s legislative activity information was utilized. The legislative process in the Korean National Assembly comprises four main steps: (1) bill introduction, (2) review by the relevant standing committee, (3) review by the Legislation and Judiciary Committee (LJC), and (4) floor debate and voting. Initially, a bill is introduced by a legislator, referred to as the sponsor. Other lawmakers who support the bill are designated as co-sponsors. The proposed bill is then reviewed by the standing committee specializing in the subject to assess its necessity and relevance. Bills approved by the committee are passed on to the LJC, which examines their legal structure and terminology. If approved, the bill proceeds to floor debate and is voted on. A bill is enacted into law if it receives a majority vote and is signed by the president. At each stage, a bill may advance or be dropped. The legislative dataset includes attributes, i.e., bill ID, bill title, description, sponsors, and review outcomes for each stage. Second, election data—comprising voting and ballot-counting results—was used. This dataset provides granular insights into electoral outcomes by district, including voter turnout and party-level vote shares. It also includes candidate names, the number of electors, voters, votes received, votes not received, and abstentions, aggregated at district, regional, and national levels. Figure 1: Political parties by district. Both datasets focus on the legislative activities of South Korea’s 21st National Assembly, covering the period from June 2020 to May 2024. This timeframe encompasses all bills proposed, reviewed, and processed during the 21st Assembly,

Read the complete text