UCLA

Advancing our understanding of the dynamics of collaboration between human users and autonomous systems is critical for the safe operation of complex sociotechnical systems. As applications extend from safety-critical domains to consumer-level technologies, the inclusion of decision-making machine agents motivates the reassessment of human-system interaction expectations and their implications on safety. Automated Driving Systems (ADS) are expected to play a significant role in future mobility, for both commercial and personal applications. In addition to addressing the complex technical challenges this implies, developers and regulators must consider the role humans play in ADS safety, whether on-board drivers, remote driving assistants, or as fellow road users.This dissertation presents an integrated operational safety methodology to model Human-Autonomy Teams (HATs) in ADS contexts, with a focus on teamwork-related failures and system-level safety, aiming to evolve from isolated safety analyses towards a holistic perspective of human and machine agents. Building on the Information, Decision, and Action in Crew context (IDAC) cognitive framework, this research develops structured representations of the complex dynamics involved in ADS-related teamwork, establishing the foundations of Human Reliability Analysis methods to support the Probabilistic Risk Assessments of ADS operations.The methodology models HATs by developing risk scenarios, defining safety-critical tasks, and identifying contributing factors that influence the performance of both human and automated agents. Three case studies are assessed: (1) drivers interacting with low-level driver assistance features, (2) transitions of control between driver and ADS under conditional automation paradigms, and (3) remote operators supervising highly automated vehicle fleets.A causal driver-system team model is developed and experimentally validated through human subject driving simulator experiments. The influence of selected Performance Shaping Factors on team dynamics in conditional automation settings is analyzed using both objective and subjective performance indicators, which inform and validate modeling choices, as well as providing a pathway to improve human-system interaction studies in simulation environments. The resulting model enables the derivation of design principles and functional constraints for HATs operating in ADS contexts. These findings aim to inform practical design guidelines for enhancing teamwork, mitigating failures, and improving the operational safety of ADS operations.