GreenScreen: Automatic Accessible Presentation Generation from IPCC Reports (Papers Track)
Alice Heiman (Stanford University); Komal Vij (Stanford University); Anjali Sreenivas (Stanford University)
Abstract
The Intergovernmental Panel on Climate Change (IPCC) Summary for Policymakers (SPM) is key for communicating climate assessments to leaders and policymakers. However, these SPMs often have poor readability for their target audiences. Research in cognitive theory suggests that more accessible and visual presentations can improve understanding of complex, dynamic systems, such as climate change. AI-driven extractive summarization and content curation have shown promise in fields such as medicine and the social sciences, leading to calls for their application in climate science, where critical information is often complex to understand despite its urgency. In response, we propose an LLM-driven automated pipeline, GreenScreen, which transforms dense IPCC SPM reports into clear, visual slide decks. This approach makes key climate insights more accessible and actionable. Our results indicate that GreenScreen improves readability from a Grade 18 (College Graduate) level to a Grade 4 level while preserving 83% content accuracy. Code is available at https://github.com/kvcs11/GreenScreen.