Oral Presentation

Studies suggest that damaging earthquakes lead to abundant news content and people turn to news sources for vital information during disasters. Among questions journalists seek to answer are: What happened and what should we expect in the immediate future? Where and when was the earthquake? Who and what is impacted? Are there injuries? Why did this happen? When will it happen again? When is “THE BIG ONE?” Explanations are often crafted by well-meaning journalists with varying levels of knowledge and experience with earthquake science, but who seek to answer these vital questions for the public. Journalists covering an event use well-documented journalistic processes, such as consulting expert sources who are in the position to speak to the event and referencing evidence that both tells the story and validates the information provided by scientists.
However, in the age of digital publishing, newsroom resources are scarce and journalistic work is evolving. Media outlets combine traditional journalistic methods with novel content production. The Los Angeles Times uses Quakebot to pull USGS data on an earthquake’s location, magnitude, and other critical information into an article that is reviewed by an editor and can be published in minutes. Other publications use generative AI to create realistic, but fabricated, images showing scenes of future earthquake damage.
At the same time, the use of generative AI programs are becoming more popular as a source of basic information, particularly by people in the workforce and young adults. As AI pulls information from a variety of sources to deliver earthquake information, we ask: What earthquake information do generative AI programs offer? What is the quality of earthquake information generative AI programs deliver? What sources do they use in their synthesis of earthquake information? This case study explores the earthquake-related content generated by four AI-driven programs. A thematic analysis of the data gleans surprising insights into the relationship between journalism and AI, shows the types of problematic AI content can be delivered to users, and points to the importance of consistent, accurate online earthquake communication.

This study has direct implications for science communication and can inform how scientists interact with news organizations using AI, as well as how science communication can contribute to more accurate and helpful AI generated content.