Poster #133 - Why Communities Say No: An LLM-based Analysis of Community Acceptance in Renewable Energy Projects

• In Event: Friday Poster Reception

Abstract

Public support for renewable energy is widespread, yet many projects face strong opposition from local communities, delaying or blocking implementation. This “social gap” between national consensus and local resistance underscores the need to better understand how communities perceive renewable energy projects. Traditional methods, such as surveys and interviews, offer valuable insights but have clear limitations: surveys rely on predefined questions, and interviews are costly and difficult to scale. Recent large language models (LLMs)-based approaches tend to focus on sentiment without capturing the underlying reasons for support or opposition. Our method uses LLMs to extract structured arguments directly from local news discourse, identifying who supports or opposes a project and why. This enables scalable analysis of community reasoning, grounded in how concerns are publicly framed and expressed. These insights are particularly critical at the local level, where community reasoning often diverges from national attitudes.

To address these challenges, we introduce the CARE (Community Acceptance for Renewable Energy) Framework, a new approach that uses LLMs to automatically analyze community reactions in news articles. CARE identifies three core elements: the energy project (e.g., a wind farm), the local community (e.g., residents or local leaders), and their specific concerns (e.g., noise or environmental impact). It also determines whether the community supports or opposes the project and why, shifting the focus from general sentiment to structured reasoning. By identifying the rationale behind community responses, CARE reveals how local acceptance is articulated in public discourse, complementing broader surveys that primarily capture national attitudes.

We evaluate CARE using a benchmark dataset of 338 U.S.-based news articles, manually annotated by experts. Results show that LLMs can reliably detect community stance and core concerns at the article level. While capturing fine-grained nuances remains challenging, CARE demonstrates strong potential as a scalable and cost-effective tool for analyzing community-level discourse across large datasets. We then apply CARE to over 3,000 news articles on wind energy and construct a Community Acceptance Index (CAI), which integrates engagement volume and net sentiment to generate state-level scores that reflect the intensity and polarity of community discourse.

Our analysis reveals substantial variation in how communities frame their concerns. States with lower acceptance tend to emphasize immediate, localized issues such as “jobs” and “noise.” In contrast, states with higher acceptance frame renewable energy in terms of broader goals like “climate change” and “reliability,” while also voicing system-level critiques such as “cost” and “pollution.” These distinctions—often flattened in sentiment-only analyses—suggest that community acceptance is shaped not just by sentiment, but by how communities frame and reason about energy projects. CARE enables researchers and policymakers to trace these patterns across space and time, offering new opportunities to understand resistance and to design more responsive, context-sensitive renewable energy policies.

Authors

• Chaerim Song, Korea Advanced Institute of Science and Technology
  Presenting Author

• Soh Young In, Korea Advanced Institute of Science and Technology
  Non-Presenting Co-Author

• Gaku Morio, Stanford University
  Non-Presenting Co-Author

• Hayoon Song, Korea Advanced Institute of Science and Technology
  Non-Presenting Co-Author