The dynamics of trust : AI and human interaction in sleep scoring accuracy and agreement

Abstract

This research explores the integration of artificial intelligence (AI) into sleep research, specifically focusing on trust dynamics between sleep technologists and AI-driven recommendations. Utilizing the MicroNyx platform, we aimed to understand how the source of sleep scoring recommendations-either AI or human-affects the trust and behavior of sleep technologists. The study involved providing sleep technologists with recommendations that were either human-generated but presented as AI-generated, or human-generated and presented as human-generated, assessing their accuracy, decision-making time, and navigational behavior.Our findings highlighted the need for accurate and reliable recommendations. When technologists were given incorrect recommendations, their accuracy dropped. They preferred human recommendations over AI ones, even when the AI recommendations were correct. On average, they took 6 seconds less per epoch with correct human recommendations compared to no recommendations, and nearly 19 seconds less compared to incorrect AI recommendations.These results emphasize the importance of building trust in AI through reliable data and consistent accuracy.