Exploring Dissatisfaction in Bus Route Reduction through LLM-Calibrated Agent-Based Modeling

1. Problem Statement

The paper addresses a critical challenge in urban transportation planning: the decline in bus ridership driven by emerging mobility modes, which creates fiscal pressure on cities to sustain underutilized routes. Decision-makers face a dilemma where cutting routes to save costs may inadvertently degrade network efficiency and increase passenger dissatisfaction. The core problem is understanding the nonlinear relationship between progressive bus route reductions and passenger dissatisfaction, specifically how these reductions impact different demographic groups and the overall resilience of the transport network. Traditional models often struggle to accurately capture complex human behavioral responses to service degradation without extensive, costly data collection.

2. Methodology

The study proposes a novel hybrid modeling framework that integrates Agent-Based Modeling (ABM) with Large Language Models (LLMs):

• Data Source: The model is grounded in real-world IC-card data from Beijing's Huairou District, providing a realistic baseline of passenger travel patterns.
• LLM Calibration: A key innovation is the use of an LLM with few-shot learning to calibrate the ABM. Instead of relying solely on historical statistical regression, the LLM is used to estimate passenger sensitivity parameters. These parameters quantify how different user groups react to specific service variables:
  • Travel time
  • Waiting time
  • Number of transfers
  • Crowding levels
• Simulation Scenarios: The calibrated ABM simulates progressive bus route cutbacks to observe the evolution of passenger dissatisfaction across the network. It tracks how the removal of specific routes affects the system's stability and the sentiment of various demographic cohorts, including the elderly and passengers with disabilities.

3. Key Contributions

• LLM-Enhanced ABM: The paper demonstrates a new paradigm for calibrating agent-based models using LLMs, allowing for the rapid estimation of complex behavioral parameters that are difficult to derive from raw data alone.
• Identification of Structural Criticality: It shifts the focus from operational metrics (like capacity) to structural configuration, proving that the topology of the bus network is the primary determinant of system stability.
• Three-Phase Dissatisfaction Model: The study conceptualizes the evolution of dissatisfaction into three distinct phases: Stable, Transitional, and Critical. This provides a theoretical framework for understanding how user sentiment degrades over time as services are cut.
• Equity Analysis: It explicitly quantifies the disproportionate impact of route cuts on vulnerable populations, offering a data-driven basis for equitable transport planning.

4. Key Results

• Structural Dominance: The structural configuration of the bus network exerts a stronger influence on system stability than capacity or operational factors.
• Exponential Dissatisfaction: The elimination of high-connectivity routes triggers an exponential rise in total dissatisfaction. The system does not degrade linearly; rather, it remains relatively stable until specific structural thresholds are breached.
• Disproportionate Impact: Passengers with disabilities and older adults are significantly more sensitive to route reductions, facing higher levels of dissatisfaction compared to the general population.
• Three-Stage Thresholds: The simulation reveals that the continuous reduction of bus routes exhibits three-stage thresholds. Once the system crosses the critical threshold, even minor additional reductions in routes lead to a catastrophic loss of passenger flow and a spike in dissatisfaction.

5. Significance and Implications

This research provides actionable insights for urban planners and policymakers:

• Resilient Planning: It underscores the necessity of identifying and preserving structural critical routes to maintain network stability, rather than simply cutting the least utilized lines based on ridership volume alone.
• Equitable Service: The findings highlight the moral and practical imperative to provide stable services to vulnerable groups, as their dissatisfaction rises sharply with service degradation.
• Nonlinear Risk Management: By identifying the "tipping points" (thresholds) where small cuts cause massive system failure, the study enables cities to avoid crossing critical limits, thereby preventing a collapse in public trust and ridership.
• Methodological Advancement: The successful integration of LLMs for calibrating ABMs offers a scalable, cost-effective tool for future transport modeling, particularly in scenarios where historical behavioral data is sparse or complex.