Plenary Speakers

Abstract: Analysis and experimental data from behavioral economics have shown biologically natural limits on the depth of strategic thinking. A recent innovation in this analysis is the idea that mental attention or other cognitive resource limits can endogenisely determine strategic depth. This talk reviews theories and a variety of evidence with an eye toward informing studies of multi-agent control. 

Bio: Colin F. Camerer is a behavioral economist. With the goal of improving the economic analysis of decisions, games, and markets, he uses methods from psychology and neuroscience, including eye-tracking, lesion patients, EEG, fMRI, wearable sensors, machine learning, and animal behavior. His recent research has focused on visual salience and habits. Camerer received his BA in quantitative studies from Johns Hopkins University, and he holds an MBA in finance and a PhD in decision theory both from the University of Chicago Graduate School of Business. Before coming to Caltech in 1994, he was a faculty member at the University of Chicago GSB (1991–1994), the Wharton School of the University of Pennsylvania (1983–1991), and the Kellogg Graduate School of Business at Northwestern University (1981–1983). He has held visiting professorships at Oxford University (2014–15) and Caltech (1987), and, since 2016, he has been a visiting fellow at the Wharton Neuroscience Initiative. He is a member of the American Academy of Arts and Sciences, a fellow of the Econometric Society, and a fellow of the Society for the Advancement of Economic Theory. Camerer was president of the Society for Neuroeconomics from 2005 to 2006 and president of the Economic Science Association from 2001 to 2003, and he has served on many editorial boards. Since 2007, he has been a chair of the Russell Sage Foundation Behavioral Economics Roundtable. He has one patent accepted on "Active Learning Decision Engines." Camerer was named a MacArthur Fellow in 2013, and he received an honorary doctorate from the Stockholm School of Economics in 2019.

Abstract: As autonomous systems become teammates in increasingly complex settings, understanding how people develop and update trust is essential to the modeling and design of cyber-physical human systems. How does trust evolve through interaction, and how can computational models capture this process as teams expand beyond one human and one autonomous agent? 

In this talk, I will bring together empirical and computational approaches to examine trust dynamics in dyadic and multi-agent human–autonomy teams. I will first present experimental findings identifying the properties of trust dynamics and discuss their implications for modeling how trust develops over time. I will then introduce a computational model grounded in these properties. Building on this foundation, I will present our recent progress in modeling trust in multi-human, multi-robot teams and discuss the challenges that arise when moving beyond dyadic interaction. I will conclude with implications for the design of autonomous teammates and open questions in understanding and modeling trust in increasingly complex human–autonomy systems.

Bio: Jessie Yang is an Associate Professor in the Department of Industrial and Operations Engineering at the University of Michigan. She also holds a courtesy appointment at the School of Information and an affiliated position at the Robotics Institute. Prior to joining U-M, she was a postdoc at the Interactive Robotics Group of MIT and the SUTD-MIT International Design Center. She obtained her Ph.D. in Mechanical & Aerospace Engineering (Human Factors) from Nanyang Technological University, working with Taezoon Park and Christopher D. Wickens. Her research interests include human-autonomy/robot interaction, human factors in high-risk industries, and user experience design.

Abstract: We as humans have a wide range of values. Values such as safety, autonomy, community, comfort, fairness, quality of life, privacy, and so many others. These human values are then emergent behaviors that technology solutions often struggle to capture because there is not a direct way to measure them. Instead proxies are required in order to quantify these qualitative and usually fuzzy human values. Value sensitive design, systems engineering and assurance cases are some tools that engineering teams can leverage in order to prove emergent behavior and ensure desired human values are present in the technology they build.

Bio: Sarah Thornton is the inaugural Head of Product Integrity for AD/ADAS at Woven by Toyota, where her team leads software safety and quality, cybersecurity, operational safety, and assurance cases. She received her Ph.D. in mechanical engineering at Stanford University in 2018. Her thesis was on designing autonomous vehicle motion planning algorithms with ethical considerations. She obtained her Master's in mechanical engineering from MIT in 2013 and her Bachelor's in mechanical engineering from UC Berkeley in 2011. She previously worked at Nuro as a Systems Engineering TLM for Contingency Management, where her team was responsible for fault management, backup systems, and fail-operational Autonomy. Prior to Nuro, she worked as a Senior Robotics Engineer at Built Robotics, where she implemented an array of projects ranging from architecting parts of their safety system to developing high-level planning algorithms.