Measuring the state of open science in transportation using large language models
Using a large language model pipeline validated against human annotations, the study evaluated code and data sharing across 10,724 articles published in Transportation Research Parts A through F from 2019 to 2024. Among quantitative papers, only 5% shared code repositories, 4% shared data repositories, and 3% provided both. The analysis detected no statistically significant differences in citation counts or peer-review durations between open-science papers and non-sharing papers.
Why it matters — It provides the first large-scale empirical baseline for open science practices in transportation, demonstrating that individual author incentives like higher citations or faster reviews are currently absent in the field.
Caveat: The dataset is restricted to six journals within the Transportation Research Part series, which may not represent all subfields or publishing venues in transportation studies.