Selecting AI Solutions for Managing False Information During Disasters: A Prescriptive Recommendation Framework

Abstract

Social media has become an essential communication tool during disaster situations. Unfortunately, it is also used to spread false information very rapidly. This includes rumors, outdated or misleading information, doctored images, and fake news stories. These can undermine and disrupt public trust and emergency response efforts and further endanger already impacted populations. Although research has addressed how AI can identify false information, much of it has focused on individual AI models. There remains a lack of prescriptive guidance to assist disaster management teams in selecting appropriate AI configurations for various types of disasters and false information. To address this, we propose a framework using a design science research approach. This paper reviews literature on disaster-related false information, AI-enabled misinformation detection, multimodal analysis, real-time adaptive models, cloud-based AI, federated learning, and Edge AI. We then provide a framework for three essential design tasks: mapping false information to disaster types, determining the factors that influence AI use in these situations, and recommending appropriate AI solutions. Our framework shows that different disaster contexts require different AI solutions. For instance, rapidly spreading text-based misinformation requires real-time natural language processing and network analysis, visual misinformation requires multimodal analysis and authenticity verification, privacy-sensitive disasters may require federated learning or Edge AI, and recurring disasters may benefit from cloud-based AI that facilitates the use of shared datasets. Overall, this research proposes a prescriptive artifact to help disaster managers select appropriate AI solutions to manage misinformation on social media during disasters.