The Neural Basis of Advice Utilization During Human and Machine Agent Interactions

Abstract

Understanding how individuals utilize advice from humans and machines has become progressively more pertinent as technological advances have pervaded our society. With an increasing shift towards relying on automation, the necessity to understand the complex interactions that exist between humans and automation has emerged. This thesis examines the behavioral, cognitive and neural mechanisms involved with advice utilization from human and machine agents framed as experts. A series of two studies were implemented that consisted of an X-ray luggage-screening task with functional magnetic resonance imaging and effective connectivity analysis. To assess advice taking differences between human and machines across both studies, the agents’ reliability was manipulated with high error rates. To fully ascertain how individuals respond to unreliable advice, the focus of Chapter Two was on false alarms, while in Chapter Three the focus was on misses. In each study, we demonstrated that there were unique behavioral responses and brain activation patterns, but in both studies participant performance levels declined overall. In Chapter Two, we showed that participants interacting with the human agent had a greater depreciation of advice utilization during bad advice and there was activation in brain regions associated with evaluation of personal characteristics, traits and interoception. In addition, the effective connectivity analysis revealed that the right posterior insula and left precuneus were the drivers of the network that were reciprocally connected to each other and also projected to all other regions (right precuneus, posterior cingulate cortex, rostrolateral prefrontal cortex and posterior temporoparietal junction). In Chapter Three, we demonstrated that advice utilization decreased more for the machine-agent group and brain areas involved with the salience and mentalizing networks, as well as sensory processing involved with attention, were recruited during the task. The effective connectivity analysis showed that the lingual gyrus was the driver during the decision phase that projected to all other target regions (anterior cingulate cortex, precuneus and cuneus) and the fusiform gyrus was the driver during the feedback phase that sent an output to the inferior parietal lobule. The contribution of this thesis is a greater comprehension of the decision-making processes involved during advice taking, which may serve as a building block for uncovering the different factors involved with human-machine interactions.