Abstract: For AI systems to integrate with people in real-world settings, they will need to respond properly to social events with potential moral implications: such as acts of physical harm, bullying, cheating or acts of goodwill. This requires an inner sense of morality as well as mind-reading capabilities that approach human level. To date, however, no artificial system is near these capabilities. Our larger research project aims to capture the full range of human social intelligence, and to this end, our current goal is to model when, how, and why people choose to communicate social information to others—what has been called gossip—and to model this process in an artificial social agent. In this talk I present our current progress. I will describe (a) our social intelligence and communication framework, (b) our test paradigm, (c) current computational model developments, and (d) an empirical experiment we conducted to obtain novel data to test model predictions. I will then present key empirical findings together with the best fits from our model compared with two others. Our model outperformed the others by best fitting both the general patterns and specific results of the empirical data (i.e., percent choosing to communicate information about a target individual to others). Thus, our model successfully captures key aspects of human moral decision-making, helping to move toward human-like social understanding and communication in artificial systems.

Biography: Jerald Kralik is a Visiting Professor in the Department of Bio and Brain Engineering at Korea Advanced Institute of Science and Technology (KAIST). He received his B.S. degree in Zoology from the Honors College at Michigan State University, studied Psychology and Artificial Intelligence at Georgia Tech, then received A.M. and Ph.D. degrees in Psychology from Harvard University. He completed a post-doctoral position in Cognitive Neuroscience at the Duke University Medical Center, where he co-authored a paper in Nature on brain-machine interfaces that received world-wide attention, in which neural recordings from monkeys controlled a robotic arm. He then completed another post-doctoral position at the National Institute of Mental Health (NIMH), where he won an NIMH research award while conducting leading-edge investigations of the neurophysiology of the prefrontal cortex. He then was an Assistant Professor in the Department of Psychological and Brain Sciences at Dartmouth College before coming to KAIST. Prof. Kralik has received research funding from the United States National Institutes of Health and Department of Defense, and has published widely in high-profile journals in Neuroscience, Psychology, and Artificial Intelligence, including Nature, PNAS, PLOS Biology, Journal of Neuroscience, Psychological Science, Cognition, and Robotics and Autonomous Systems. His research interests include evolutionary, cognitive, social and computational neuroscience, as well as cognitive architectures, artificial intelligence and brain engineering.