Sign in Subscribe
Social media

AI-Driven Framework for Evaluating Climate Misinformation on Social Media

Strategic Insights

— 1 min read

Researchers at Kristianstad University have developed an AI-based framework to assess the quality of climate-related content on social media platforms. The framework evaluates data quality using key dimensions of credibility, accuracy, relevance, and sentiment polarity, and has been tested on a dataset of over 1 million posts from Twitter and YouTube between 2018 and 2024. The results show a 9.2% improvement in misinformation filtering and an 11.4% enhancement in content credibility detection compared to baseline models.

Key Takeaways:

  • The AI-based framework uses transformer-based NLP models, sentiment classifiers, and misinformation detection algorithms to evaluate data quality on social media platforms.
  • The framework defines data quality using key dimensions of credibility, accuracy, relevance, and sentiment polarity.
  • The system processes user-generated content to detect sentiment drift, engagement patterns, and trustworthiness scores.
  • The dataset used for evaluation consisted of over 1 million posts from three major platforms (Twitter and YouTube) between 2018 and 2024.
  • The results demonstrate a 9.2% improvement in misinformation filtering and an 11.4% enhancement in content credibility detection compared to baseline models.
  • The findings provide actionable insights for researchers, media outlets, and policymakers aiming to improve climate communication and reduce content-driven polarization on social platforms.
  • The research was conducted by Zeinab Shahbazi, Rezvan Jalali, and Zahra Shahbazi from the Research Environment of Computer Science (RECS) at Kristianstad University.

Statistics:

  • 9.2% improvement in misinformation filtering compared to baseline models.
  • 11.4% enhancement in content credibility detection compared to baseline models.
  • Over 1 million posts from three major platforms (Twitter and YouTube) between 2018 and 2024.
  • Evaluation metrics included precision, recall, F1-score, and AUC.
  • The framework developed in this research demonstrated a significant improvement in evaluating data quality on social media platforms.

Sources:

  • AI-Driven Framework for Evaluating Climate Misinformation and Data Quality on Social Media. Future Internet, 2025, 17(6):231.
  • Our news editors report that additional information may be obtained by contacting Zeinab Shahbazi, Research Environment of Computer Science (RECS), Kristianstad University, 291 39 Kristianstad, Sweden.

Read more