Analyzing Public Perceptions and Sentiment of Electric Vehicles in the Era of Sustainable Transformation
Engineered a multi-source sentiment analysis pipeline processing 1.1M+ social media posts from Reddit and NYT, achieving 91.6% classification accuracy using DistilBERT and comparing 10 LLM variants for sentiment bias.
Published in 80th Annual Conference of the American Association for Public Opinion Research (AAPOR), 2025
Overview
This research examines how public sentiment toward electric vehicles (EVs) has evolved from 2020-2024 across social media and traditional news platforms, revealing critical insights about public perception gaps and systematic biases in LLM-based sentiment analysis.
Methodology
Data Collection & Processing
- Engineered a multi-source sentiment analysis pipeline processing 1.1M+ social media posts
- Collected data from 5 Reddit communities (550K comments) and 40 New York Times articles
- Implemented DistilBERT transformer model achieving 91.6% classification accuracy on electric vehicle discourse
LLM Comparative Analysis
- Conducted systematic evaluation of 10 Groq LLM variants (Llama 3.1/3.2 series, Mixtral)
- Discovered systematic bias in large language model sentiment prediction
- Found LLMs exhibited +0.57 positive sentiment bias compared to actual Reddit data (M=-0.18, SD=0.52)
- Statistical significance demonstrated: F(2,549)=28.43, p<0.001
Key Discoveries
Temporal Trends
- Identified 35% increase in negative EV sentiment on Reddit post-2022 despite rising adoption rates
- NYT coverage remained consistently negative (M=-0.23) for core EV terms
- Positive sentiment observed for Tesla and autonomous vehicle keywords
LLM Performance
- Llama-3.2-90b-vision-preview best approximated human sentiment patterns
- Revealed systematic optimism bias in AI-generated sentiment assessments
- Critical implications for using LLMs in survey research and public opinion analysis
Research Implications
This study has significant implications for:
- Survey methodologists using AI-assisted data collection and analysis
- Public policy researchers analyzing public sentiment toward green technologies
- AI researchers developing more accurate and unbiased sentiment analysis tools
- Climate communication strategists understanding public EV perception
Recommended citation: Chakravarty, S., & Shrivastava, N. (2025). "Analyzing Public Perceptions and Sentiment of Electric Vehicles in the Era of Sustainable Transformation." 80th Annual AAPOR Conference. St. Louis, MO.
Download Paper