Foundations & Frontiers

Foundations & Frontiers

Foundations & Frontiers

The Foundations & Frontiers Speaker Series brings leading cognitive scientists to U-M to present a special pair of presentations on the same day. The first presentation serves as an introduction to an important theoretical idea or method in the field -- the Foundations. The second presentation concerns the application of that idea or method to an innovative topic, thus exploring the Frontiers of the field in a way that highlights the significance of the theoretical idea.

Fall 2024

Angela Yu

October 7, 2024
East Hall 4448

Dr. Yu conducts research at the intersection of natural and artificial intelligence. Her group uses mathematically rigorous and algorithmically diverse tools to understand the nature of representation and computations that give rise to intelligent behavior, with particular regard to the challenges posed by inferential uncertainty and the opportunities afforded by volitional control. Using diverse machine learning and statistical tools, e.g. Bayesian statistical modeling, control theory, reinforcement learning, and information theory, theoretical frameworks and mathematical models are developed to explain disparate aspects of cognition importation for intelligence: perception, attention, decision-making, learning, cognitive control, active sensing, economic behavior, and social interactions. After completing her bachelor’s at MIT, PhD at UCL Gatsby Computational Neuroscience Unit, and postdoc training at Princeton University, Dr. Yu was faculty at UCSD Cognitive Science for 14 years before joining TU Darmstadt in Germany as an Alexander von Humboldt AI Professor in late 2022. With her interdisciplinary expertise and interest, Dr. Yu’s arrival strengthens and complements research at the Centre for Cognitive Science at TU Darmstadt and at hessian.AI, the Centre for Artificial Intelligence of Hesse based in Darmstadt. Her new group currently has positions open for postdocs, PhD students, master’s students, research interns, and visiting scholars.

Schedule

2:00-2:30 pm Foundations Presentation
2:30-2:40 pm Q & A
—10 minute break—
2:50-3:20 pm Frontiers Presentation
3:20-3:30 pm Q & A

"Computational Modeling of Human Face Processing"

Presentation Abstract

Face processing plays a central role in everyday human life. We investigate the nature of face representation and processing in the brain, by adapting and developing appropriate machine learning and computer vision methods. In the Foundations portion of the lecture, I will be giving a review of the technical methods utilised in our modeling work, including novel methods for combining image-computable machine vision algorithms with more classical psychological methods, such as multidimensional scaling, that take human similarity judgments into account.

In the Frontiers portion of the lecture, I will describe our work using a neurobiologically supported and psychologically informed face representation, to computational model and quantify human perception and judgment of faces, including a computational explanations of the provenance of certain racial and gender biases. Furthermore, we will examine some implications of these findings for the broader domains of psychiatry, social psychology, economics, and political science.

Previous Years Speakers

Winter 2024 Schedule

Sean Trott

February 19, 2024
4448 East Hall

Sean Trott is an Assistant Professor at the University of California, San Diego. He holds a joint appointment in Cognitive Science and Computational Social Science. His research focuses on how humans and machines understand language, and makes use of Large Language Models (LLMs) to test hypotheses about human cognition.

Presentation Video

"Using Large Language Models to Understand Human Cognition"

Presentation Abstract

Foundations: Many debates in Cognitive Science—such as whether certain cognitive capacities are innate, or acquired through specific experiential input—are entrenched and difficult to resolve. A new paradigm attempts to address these debates using Large Language Models (LLMs) to test competing theories of human cognition. In particular, because (most) LLMs are trained on linguistic input alone, they serve as useful baselines: measures of what kinds of behaviors and capacities could in principle emerge purely from exposure to statistical patterns in language. In this talk, I discuss the motivations for such an approach, and briefly survey several examples from the literature. Finally, I discuss the relevant trade-offs and considerations that might inform a researcher's decision about whether to use LLMs in their own research, including: the amount (and quality) of data an LLM has been trained on, issues of construct validity, and multimodal models.

Frontiers: Humans often reason about the mental states of others, even when those mental states diverge from their own. The ability to reason about false beliefs—part of the broader constellation abilities that make up "Theory of Mind"—is viewed by many as playing a crucial role in social cognition. Yet there is considerable debate about whether this ability comes from. Some theories emphasize the role of innate biological endowments, while others emphasize the role of experience. In this talk, I consider a hypothesis about a specific kind of experience: language. To test this "language exposure hypothesis", I use GPT-3, a Large Language Model (LLM) trained on linguistic input alone, and ask whether and to what extent such a system displays evidence consistent with Theory of Mind. The LLM displays above-chance performance on a number of tasks, but also falls short of human performance in multiple cases. I conclude by discussing the implications of these results for the language exposure hypothesis specifically, and for research on Theory of Mind more generally.

Fall 2022 Schedule

Max Kleiman-Weiner

Monday, October 31, 2022
4448 East Hall

Max Kleiman-Weiner is a co-founder of Common Sense Machines. He was previously a fellow at Harvard in the Data Science Institute and Center for Research on Computation and Society and completed a PhD in Computational Cognitive Science at MIT where he was a NSF and Hertz Foundation Fellow. His thesis won the Robert J. Glushko Prize for Outstanding Doctoral Dissertation in Cognitive Science. He has received best paper awards at COGSCI and RLDM for models of human cooperation and the William James Award at SPP for computational work on moral learning. Max was Co-Founder and Chief Scientist of Diffeo which was acquired by Salesforce in 2019. Previously, he was a Fulbright Fellow in Beijing, earned an MSc in Statistics as a Marshall Scholar at Oxford, and did his undergraduate work at Stanford as a Goldwater Scholar.

"Reverse Engineering Human Cooperation"

Presentation Abstract

Human cooperation is distinctly powerful. We collaborate with others to accomplish together what none of us could do on our own; we share the benefits of collaboration fairly and trust others to do the same. Even young children understand, learn from, and collaborate with others in ways that are unparalleled in other animal species and are still lacking in our most sophisticated artificial intelligences. What are the cognitive representations and processes that underlie these distinct abilities and what are their origins?

In the Foundations portion of the talk, I will review models of the emergence of cooperation that have been worked out over the past 50 years in game theory, psychology, evolutionary biology, and computer science. While highly influential, these models leave out some of the most important cognitive capacities that enable the unprecedented scale and scope of human cooperation. In the Frontiers portion, I will present a computational framework based on the integration of individually rational, hierarchical Bayesian models of learning, together with socially rational game-theoretic models of cooperation. In computational and behavioral experiments I will show how this framework can explain how the cognitive structures underlying cooperation might evolve, social knowledge can be learned, and how cooperative behavior is generalized in the moment across an infinitude of possible situations: inferring the intentions and reputations of others, distinguishing who is friend or foe, and learning a new moral value all from just a few observations of behavior.

Winter 2021 Schedule

Dr. Sydney Levine

Dr. Sydney Levine
Friday, March 5, 2021

Dr. Sydney Levine is a postdoctoral scholar in the Department of Brain and Cognitive Sciences at MIT and the Department of Psychology at Harvard University.

Schedule

3:00 pm - 3:30 pm Foundations Presentation
3:30 pm - 3:45 pm Q & A
—5 minute break—
3:50 pm - 4:40 pm Frontiers Presentation
4:40 pm - 5:05 pm Q & A

"Universalization: philosophical origins and cognitive applications"

Presentation Abstract

Some people think it is immoral not to vote, but why? Current theories of moral psychology — based largely on outcomes, rules, or affect — have trouble explaining this intuition. After all, one vote typically makes no difference in the overall outcome of an election. Moreover, there is not necessarily a rule or norm that mandates voting. And the thought of not voting isn’t particularly emotionally charged. Rather, I propose that moral judgments in cases like this arise out of the logic of universalization — essentially the process of asking “what if everyone did that?”

The logic of universalization is well-known to moral philosophers, appearing in the theories of Kant, RM Hare, George Singer, TM Scanlon and others. In the Foundations section of this talk, I will introduce the concept of universalization and explore how and why it has been used so effectively in moral philosophy. In the Frontiers section of the talk, I will demonstrate how I have used this philosophical concept as a starting place for a model of moral cognition. I will define a computational model of universalization and show how it predicts subject judgments with quantitative precision. In addition, developmental work suggests that universalization may even be used by children as young as 4. Finally, I will end by showing how universalization may be the key to a unified theory of moral cognition.

Presentation Video

Fall 2020 Schedule

Dr. Frederike Petzschner
October 30, 2020

Dr. Frederike Petzschner is a Carney Institute Fellow in the Center for Computational Brain Science and an Assistant Professor in the Department of Psychiatry and Human Behavior at Brown University.

Presentation Abstract

Dr. Petzschner gave two presentations on October 30, each followed by a Q&A session.

"A Brief History of Computation" (Foundations Presentation)
Our notion of what the capabilities and function of the brain and mind are has evolved fundamentally in the past century. As a result, we have moved from early Psychophysics to Behaviorism to the Cognitive Revolution, when theories of computation entered the forefront of modern Cognitive Science. This history and the fundamental questions posed at different times provide a great deal of insight into our modern thinking and paves the way where the field might take us in the future. In this lecture, I will try to provide a short guide through the history of computation and discuss what could be learned from it.

"Computational Approaches for Mental Health" (Frontiers Presentation)
The growing field of Computational Psychiatry provides a prime example of how theories of computation may provide not only insights into the function of healthy minds but also mental disorders. In this lecture, I will discuss three examples of where we apply computational methods to understand learning, perception or decision-making in Obsessive-Compulsive Disorder, Gambling Addiction and Disorders of Interoception.

Presentation Video

For more information about the Foundations & Frontiers Speaker Series, please contact:
[email protected].

Search: {{$root.lsaSearchQuery.q}}, Page {{$root.page}}

for

Search: {{$root.lsaSearchQuery.q}}, Page {{$root.page}}

Foundations & Frontiers

Fall 2024