ABSTRACT

Human – Artificial Intelligence (AI) Teaming (HAT) is a rapidly emerging research field that seeks to combine human and AI agent capabilities to enhance performance, efficiency, and safety of mutual tasks across various domains. In the driving context, the idea of a driver ‘teaming’ with a semi-autonomous vehicle (AV) is a relatively novel concept. However, this framing could significantly help to improve driver’s interactions with highly autonomous vehicles, thereby promoting safety in surface transportation systems. To date, there have been frameworks and models introduced studying HAT within driving, though there have been no investigations into whether current drivers of semi-AVs indeed perceive their vehicles to be teammates or simply intelligent tools, or how, empirically, future semi-AV design can better foster a sense of teaming. This dissertation proposes to take first steps towards addressing this research gap in three phases. In the first study, an online survey was developed and deployed using a crowdsourcing platform to understand the perceptions current owners of Society of Automotive Engineers (SAE) L2 and L3 (partially and conditionally automated vehicles, respectively) have toward seeing themselves as teammates with their semi-AVs. In the second study, taking what was learned from the first phase, an in-lab driving simulation study was conducted to examine agent availability (driver/vehicle) in the pursuit of operationalizing human – AV teaming (HAVT) in a way that encourages safe and comfortable driving behaviors via agent communication (via in-vehicle displays) and goal obtainment. The final study more closely examined the communication modality for agent communication. The goal was to understand if the way the vehicle communicates with the driver (text vs speech) would have any influence on the HAVT outcome. This work can help provide a fundamental understanding of the psychology of semi-AV drivers, which can inform the design of in-vehicle displays that intelligently support coordination, cooperation, and collaboration within HAVT contexts.

BIOGRAPHY

Maya Luster is a current Ph.D. Candidate in Industrial Engineering with a focus in Cognitive Engineering at Purdue University. She is an Associate Human Factors Professional (AHFP) certified by the Board of Certification in Professional Ergonomics (BCPE). Maya earned her M.S. in Industrial and System Engineering from Rochester Institute of Technology (RIT) in 2018, where her master thesis research area focused on ergonomic occupational safety, and she earned her B.S. in Electrical Engineering from Mississippi State University (MSU) in 2015, with a focus in high power electronics and distribution. Her research area is within Human-AI Teaming (HAT) where she investigates various aspects of human cognition, behavior, and perception to provide implications towards improving the designs of diverse complex autonomous systems