Namit Shrivastava

Namit Shrivastava

he/him

Graduate researcher in survey methodology, trustworthy AI, and computational social science.

Analyzing Public Perceptions and Sentiment of Electric Vehicles in the Era of Sustainable Transformation

Date:

Presentation Overview

Presented research on public sentiment analysis toward electric vehicles (EVs) at the prestigious American Association for Public Opinion Research (AAPOR) annual conference—the premier venue for survey methodology and public opinion research.

Abstract

This study examines public sentiment toward electric vehicles (EVs) across social media and traditional news platforms from 2020-2024. We engineered a multi-source sentiment analysis pipeline processing 1.1M+ posts from Reddit and New York Times, implementing DistilBERT transformers achieving 91.6% accuracy. Our comparative evaluation of 10 LLM variants revealed systematic positive bias (+0.57) in AI-generated sentiment assessments, with critical implications for AI-assisted survey research.

Key Topics Covered

1. Multi-Source Data Collection

  • Reddit API integration across 5 EV-focused communities (550K+ comments)
  • New York Times article analysis (40 articles, 2020-2024)
  • Robust data pipeline handling rate limits and API constraints

2. Transformer-Based Sentiment Analysis

  • DistilBERT implementation achieving 91.6% accuracy
  • Domain-specific fine-tuning for EV terminology
  • Temporal trend analysis revealing sentiment shifts

3. LLM Comparative Evaluation

  • Systematic comparison of 10 Groq LLM variants
  • Discovery of +0.57 positive sentiment bias in LLM predictions
  • Statistical validation: F(2,549)=28.43, p<0.001
  • Implications for AI-assisted survey research methodology

Key Findings

FindingDetails
Sentiment Shift35% increase in negative EV sentiment post-2022
NYT CoverageConsistently negative (M=-0.23) for core EV terms
LLM Bias+0.57 systematic positive bias vs. actual Reddit data
Best LLMLlama-3.2-90b-vision-preview closest to human patterns

Conference Details

  • Conference: 80th Annual Conference of the American Association for Public Opinion Research
  • Dates: May 15-17, 2025
  • Location: St. Louis, Missouri, USA
  • Session: Survey Methodology & AI Integration
  • Co-author: Snigdha Chakravarty

Resources

View Conference Abstract

Implications

This research has significant implications for:

  • Survey Methodologists: Understanding LLM limitations in sentiment analysis
  • Public Opinion Researchers: Capturing authentic public discourse on green technologies
  • AI Practitioners: Developing bias-aware sentiment analysis tools
  • Policy Makers: Evidence-based understanding of public EV perceptions