Quantification and Modulation of Trust in a Crew Health Integrated Medical Response Agent (CHIMERA)

In long duration space medicine, trustworthy decision-making is crucial for delivering effective patient care. It involves integrating complex data, such as diagnostic information, patient history, and treatment outcomes, with clinical expertise and patient values. Bayesian decision support systems, leveraging their ability to handle uncertainty and provide interpretable results, are particularly suited for enhancing this process1. By quantifying uncertainties and fusing data from various sources, these systems support clinicians in making informed, evidence-based decisions that are crucial for patient outcomes and healthcare efficiency. Active inference is a theoretical framework (AIF) for understanding behavior as a single imperative to minimize free energy. The free energy principle suggests that all living organisms, including humans, strive to minimize their free energy—a measure of the system's uncertainty or surprise about its sensory inputs given its model of the world. This minimization is achieved through perception, where the agent updates its internal model to better predict sensory inputs, and action, where the agent acts upon the environment to make its sensory inputs more predictable In this Phase I proposal, we will leverage our work on cyber-physical-human autonomous systems for astronaut health and performance to demonstrate how an active inference framework allows for the quantification and modulation of trust in a critical space flight system called the Crew Health Integrated Medical Response Agent (CHIMERA)