SocialAI Research Group
Who we are.
The SocialAI Research Group is an interdisciplinary team whose goal is to develop computational models that integrate current approaches from neuroscience, psychology, sociology, and cultural anthropology with the framework of artificial intelligence to understand the emergence, maintenance, and change, in social cognition and societal structures.
Our current research uses multiagent deep neural networks to examine how different known patterns of current and historical human behavior and social biases can emerge from even a surprisingly limited set of simple psychological and computational principles embedded in dynamic social environments. We can then use these models to develop novel predictions about human behavior that can be tested using experimental psychology and neuroscience methods; as well as calibrate the models to real-world events.
Understanding and formalizing the emergence of complex human behaviors and social structures has implications both for theoretical research and for application to AI design and interventions. Theoretically, the work of SocialAI provides novel empirical insights into where human behavior comes from, and the simple primitives necessary to give rise to even complex thoughts, actions, and structures. From an applied perspective, the research of SocialAI will help inform AI designs that allow for the emergence of more fair, safe, and responsible human behaviors and structures.
SocialAI projects provide an answer to a more formal, computational depiction of emergent human group behavior. Although these experiences are a ubiquitous part of social life, the ways that people achieve these outcomes is computationally taxing, as successful cooperation requires that group members accurately predict each other’s behaviors, and people are themselves complex entities with their own potentially hidden motivations, goals, and desires. Further, both individuals and the groups to which they belong respond to cues from their environment—the need to procure food and shelter, to escape predators, to accomplish communal tasks—which are constantly shifting. By focusing on developing a core set of processes that interact with the social and ecological environments, we seek to not only understand cooperation and success, but also the dark forces of social life such as bias, stereotyping, and war.
Group formation and affiliation are pervasive aspects of human social life. Groups confer numerous advantages to individual members, including access to resources, cooperation, and a sense of belonging. Although these experiences are a ubiquitous part of social life, the ways that people achieve these outcomes is apparently computationally intractable, as successful cooperation requires that group members accurately predict each other’s behaviors, and people are themselves complex entities with their own potentially hidden motivations, goals, and desires. By focusing on developing a core set of processes that interact with the social environment, we seek to not only understand cooperation and success, but also the dark forces of social life such as bias, stereotyping, and war.
Development of Cultural Views
On its own, the formation of cooperative and often complex groups is a major way in which humans construct more predictable environments for ourselves, bringing the inherent precarity of the natural world and other people within that world more and more into our control. To make social living computationally tractable, political systems and shared worldviews may have arisen to restrict the possibility space of group living through a shared understanding of duties, rights, and responsibilities. These systems not only dictate our expectations for social life and human behavior, but also allow the group to work from a shared set of assumptions about what beliefs and behaviors are acceptable and valued. In predictive processing terms, ideologies and worldviews comprise hierarchical generative models specifying a set of predictions about the world and other people that facilitates group cooperation when shared with other members of the group. Although any particular set of rules and regulations may be arbitrary, and different groups of people can choose to set these rules in different ways, having a shared system that all members of a group adhere to reduces the variance in group members’ thoughts and behaviors, rendering the social world more predictable.
Pedagogical Sharing of Social and Cultural Norms and Stories for Learning
[TEXT GOES HERE]
Understanding Functions, Biases, and Change in Shared Cultural Reality
Consequences of Social Structure for Biased Perception, Cognition, and Decision-making
A core aspect of social perception is thinking about people in terms of their social group memberships. Perceiving an individual in terms of their social category membership allows perceivers to apply pre-existing information and attitudes about that social category to the individual. By providing information about otherwise novel individuals, social categorization reduces uncertainty about the social world and helps perceivers anticipate the characteristics of others and calibrate their own behavior. Although social categorization is necessary to make informed inferences about novel people, its reliance on cognitive shortcuts such as heuristics, stereotypes, and prejudices may lead to systematic inaccuracies in the processing of any given person. In this line of work, we examine how stereotypes shape the predictions that deep neural networks make about people when social categories can be used to make inferences about people. For example, the very act of having structured group labels may bias early perceptual hidden layers to exaggerate features that define the labeled groups, and may “fill in” missing perceptual details (e.g., a gun instead of a tool). By varying the nature of the learning environments, such as by limiting exposure to certain groups, or by establishing cultural myths about social groups (see project 1) we can understand how rational cognition can give rise to inaccurate and immoral beliefs in biased contexts.
CURRENT TEAM MEMBERS.
Wil is a professor of psychology at The University of Toronto. His research explores the computational cognitive processes of social cognition and intergroup relations.
Ethan is a professional machine learning scientist based in Toronto, Canada, where he has been affiliated with the Vector Institute for artificial intelligence since early 2018. During his PhD studies, Ethan focused on evolutionary algorithms in deep learning while also pursuing research interests in computational neuroscience.
Daniel Hoyer is a computational historian and social scientist with the Evolution Institute’s Seshat: Global History Databank project. His research seeks to understand how societal dynamics determine well-being outcomes.
Tessa is a postdoctoral research fellow at Harvard University and the University of Toronto. Her research examines how and why social change happens over decades or more.
Rebekah is a cognitive scientist and PhD student in the Department of Psychology at the University of Toronto. Her research focuses on developing computational models investigating the role of teaching and social learning in forming beliefs and knowledge.
Shon is a computer science student at the University of Toronto. His research projects include generative music modeling, analysis of climate change sentiment on social media, and RL obstacle avoidance models for autonomous driving.
Valerie is a Computer Science student at the University of Toronto who is interested in learning machine learning and its applications. She currently provides support for the generation of face datasets.
Yikai is a psychology student at the University of Toronto, who is interested in using machine learning to study social cognition. His research projects cover topics in intergroup relations and memory functioning.
Ruth is a recent graduate from the University of Toronto where she studied cognitive neuroscience. Her work involves modelling the cognitive mechanisms behind immediate impression formation using neural network, and behavioural techniques.