Compositional Co-Design of Traffic Law Reform: Integrating Citizen-Sourced Data with Strategic Game-Theoretic Analysis
TL;DR
The study applies transformer-based sentiment classification and a domain-specific emotion lexicon to filter and rank traffic law violations and proposes a two-step recommendation framework that guides the contextual screening of feasible enforcement options.
Abstract
Traffic laws work successfully when they are both enforceable and accepted by the public. In many low- and middle-income countries, however, rapid urban change, limited enforcement capacity, and short-term personal incentives make certain rules difficult to uphold. Moreover, short-term personal incentives often dominate the strategic interactions between citizens and law enforcers, leading citizens to violate laws and law enforcers to overlook violations. This study introduces a practical, data-driven framework for illustrating how such laws can be diagnosed and how strategic interactions can contribute to enforcement failure. Using 496 citizen-generated posts from the Facebook group Road Planners Bangladesh, the study applied transformer-based sentiment classification and a domain-specific emotion lexicon to filter and rank traffic law violations. Sixteen critical cases were identified, and Case V176 (autorickshaw driving on a national highway) showed the highest public concern with a Composite Intensity Index of 94/100. The selected case was modeled as a two-player strategic interaction between a violator and an enforcer using ambassador-derived and analytic hierarchy process-weighted payoff matrices. In both models, Nash equilibrium analysis produced a stable Violate–Ignore outcome, showing that the rule persists in a misaligned incentive environment rather than from random disobedience. Backward induction was then used to identify the specific incentive leverage point at which behavior can shift toward compliance—analogous to a codesign intervention within a sociotechnical system. The results show that raising the perceived risk of being caught, rather than increasing fines, was the most effective modeled lever for shifting the equilibrium in the specific case. To support implementation, the study proposes a two-step recommendation framework that guides the contextual screening of feasible enforcement options. Overall, the findings demonstrate how citizen-sourced data and computational reasoning can support a more adaptive, participatory, and system-aware approach to sustainable traffic law reform.