Toward a person-environment fit framework for artificial intelligence implementation in the public sector

Abstract

Using an embedded mixed method design, we compared a nationally representative sample of US adults and a sample of US-based emergency managers (EM) on their attitudes toward artificial intelligence (AI) and their intentions to rely on AI in a set of decision-making scenarios relevant to emergency management. Emergency managers reported significantly less positive attitudes toward AI and were less likely to rely on AI for decisions compared to the nationally representative sample. Our analysis of EMs' open-ended responses explaining their choices to use or not use AI-based solutions reflected specific concerns about implementation rather than wariness toward AI generally. These concerns included the complexity of the potential outcomes in the scenarios, the value they placed on human input and their own extensive experience, procedural concerns, collaborative decision-making, team-building, training, and the ethical implications of decisions, rather than a rejection of AI more generally. Managers' insights integrated with our quantitative findings led to a person-environment fit framework for AI implementation in the public sector. Our findings and framework have implications for how AI systems should be introduced and integrated in emergency managerial contexts and in public sector organizations more generally. Public managers' perceptions and intentions to use AI and organizational oversight processes are at least as important as technology design considerations when public sector organizations are considering the deployment of AI.