An Innovative Measure of Cognitive Function in the Human-Autonomy Partnership

Abstract

Understanding how the user will interact with the system is fundamental to ensuring success in achieving a given goal. Therefore adopting a human-centered design approach will assist in integrating the human as a key component of the system during the design process. With the increased use of autonomy across different domains, the role of the human will inevitably change; in that how the user interacts with the system is dependent on the level of delegated authority the system has been assigned. To understand these interactions and the impact this has on the user, it is important to assess how the human interacts with the system. However, as these systems become more complex we must ask whether the measures we currently use are sufficient in allowing us to better understand the underlying cognitive functions involved in human-autonomy interaction. Evaluating this partnership we can not only assess the effectiveness and efficiency of human-autonomy interaction, but also provide guidance for future designs. Novel techniques such as functional Near Infrared Spectroscopy (fNIRS) offer a direct measure of cortical blood flow changes related to brain activity. This paper discusses findings from an experiment that examined human-autonomy interaction in a simulated Autonomous Vehicle (AV) whilst exploring the neural correlates of trust and workload. Participants were asked to complete a series of primary driving scenarios with secondary distraction tasks using both manual and autonomous vehicles. fNIRS was used to assess driver cognition across both conditions. Participants were also confronted with different levels of system transparency to determine whether the level of information presented by the system effected driver trust. Findings suggest that when autonomy was presented then the cognitive activity in the right and left dorsolateral prefrontal cortex (dlPFC) and the left ventrolateral prefrontal cortex (vlPFC) was reduced, whilst secondary task performance improved. These regions are associated with effortful decision-making based on working memory (WM) and reasoning, suggesting that using autonomy helps to reduce cognitive effort by removing the user’s need to make these decisions. During the system transparency scenarios, areas of the right and left vlPFC and left dlPFC showed significantly increased activity when the system provided very little information. These regions have previously been associated with uncertainty of decision making and increased visual processing, suggesting that a lack of information provided by the system meant the driver attempted to process the decisions of the vehicle through monitoring the environment. These findings demonstrate how novel measures of cognitive function could inform the design of future systems and facilitate a more effective human-autonomy partnership.