Captains donüt navigate with a keyboard: Developing AI for naturalistic decision-making

Book chapter, 2024

his chapter discusses the challenges and mitigation strategies when developing AI decision-support systems that are technically resilient and helpful in naturalistic environments. Drawing upon a 2-year development project of an AI decision-support system for maritime navigation, three noteworthy insights are highlighted. First, it is not necessary to account for all environmental parameters in AI systems to support naturalistic decision-making. Second, common design principles to make AI-produced information explainable only apply to a certain extent in the context of naturalistic decision-making, where decisions are often made under high time pressure. Third, the chapter illustrates the benefits of employing testing methods with different degrees of contextual immersion to integrate tacit domain knowledge in the AI development process. The chapter emphasizes that appropriate HMI design is crucial to avoid potential information overload and to build trust to ensure future adoption.