Autonomous vehicles in the public eye: A global social media analysis for smarter policies

With the widespread popularity of social media, it has emerged as a mirror of public sentiments. This provides planners and policy makers the opportunity to leverage this large and diverse data source to assess the sentiments towards futuristic travel alternatives like autonomous vehicles (AVs) and investigate their temporal evolutions. Further, the geo-tagged social media datasets and the advancements in topic modelling also provide the opportunity to analyse the spatial variation of topics of interest alongside the polarity of sentiments in different parts of the world. This motivates this study, where we extracted and utilised 94,469 AV-related Tweets between the 2010-21 period across eleven countries to develop multiple machine learning (ML) based and large language model (LLM) based sentiment analysis models to detect spatiotemporal heterogeneity in sentiments among the countries and cluster them based on similarity in sentiment polarity. Subsequently, dynamic topic models (BERTopic) are used to identify the underlying themes of public discourse in the identified clusters and contextualise their evolution. The sentiment analysis results reveal substantial differences in sentiment polarity across countries and uncover five distinct clusters. People in cluster 1 (USA and South Africa) and 2 (UK, Canada and Australia) are found to be more opinionated (both positive and negative) than the other three clusters. In terms of temporal trend, the results indicate higher positive sentiments prevailing for the most part of the last decade, except for fatal accidents, which created a strong negative ripple effect. Further investigation of the key topics driving the public sentiment reveals five major inherent themes based on Tweet frequency as well as commonality across countries: safety concerns, industry and investment, excitement, clean energy, and trucks. The dynamic topic model results also provide insights into the effect of the context-specific interventions, while the relative order of the topics indicates the differential acceptance or expectation of the autonomous future among the general masses. These insights include recommendations to policymakers and AV manufacturers to suggest future directions for a sustainable transition to autonomous transportation systems.