Transcreation for Disaster Information: Simulation of Multicultural Support Using AI-Generated Personas

Today, the rapid and accurate dissemination of information is necessary. However, conventional disaster alerts often fail to effectively communicate with populations of various background. Simple translation is often insufficient, and universal design approaches cannot cater to fit diverse individual needs. Therefore, this research hypothesizes that "transcreated" messages reflecting the recipient's cultural and personal background should be used. However, transcreation methods need to take into account various factors for the varieties of the recipients are numerous. We explore the solutions through a two-stage methodology. First, a foundational study was conducted using AI-generated personas to simulate the efficiencies of varieties of personalized transcreation of disaster alerts. Second, a real-time generative AI based system is developed to be tested by human subjects. The results from simulated persona suggest that AI-generated personas can be useful for simulating real human reactions. The human-subject surveys indicate that AI-generated transcreated disaster alert is more comprehensible and effective than conventional alerts. The comparison between AI-generated persona and human survey provide us of some insights about the simulation is robust to predict the real-world behaviour of human beings. This study demonstrates the effective use of generative AI to improve disaster communication. It also provides insights for developing next-generation disaster communication systems.