Colloquium
Tom Griffiths
Computational Cognitive Science Lab
UC Berkeley
Friday April 4, 2008
3147 MEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Host: Hal Daume III
Title: Analyzing Cultural Evolution by Iterated Learning
Abstract
Much of the knowledge that human beings have about their world comes not from direct experience, but from interacting with others. This raises an interesting question: what are the consequences of learners learning from other learners? I will present both theoretical and empirical results on the implications of such a process of "iterated learning" for the information being transmitted between learners. Specifically, I will show that iterated learning with Bayesian learners, each observing data generated by the previous learner and then selecting hypotheses in accordance with Bayes' rule, results in convergence to a distribution over hypotheses determined by the prior of the learners. This result has implications both for methods for developing methods for investigating the inductive biases of human learners, and for understanding how those biases influence the evolution of concepts and languages. I will describe the results of a series of experiments bearing out the basic predictions of analysis for simple cognitive abstractions such as functions and categories as well as more complex linguistic objects such as frequency distributions and systems of color terms.
Bio
Tom Griffiths is an Assistant Professor of Psychology and Cognitive Science at UC Berkeley, with courtesy appointments in Computer Science and Neuroscience. His research explores connections between human and machine learning, using ideas from statistics and artificial intelligence to try to understand how people solve the challenging computational problems they encounter in everyday life. He received his PhD in Psychology from Stanford University in 2005, and taught in the Department of Cognitive and Linguistic Sciences at Brown University before moving to Berkeley. His work and that of his students has received awards from the Neural Information Processing Systems conference and the Annual Conference of the Cognitive Science Society, and in 2006 IEEE Intelligent Systems magazine named him one of "Ten to watch in AI."
Return to 2008 Events Calendar
Much of the knowledge that human beings have about their world comes not from direct experience, but from interacting with others. This raises an interesting question: what are the consequences of learners learning from other learners? I will present both theoretical and empirical results on the implications of such a process of "iterated learning" for the information being transmitted between learners. Specifically, I will show that iterated learning with Bayesian learners, each observing data generated by the previous learner and then selecting hypotheses in accordance with Bayes' rule, results in convergence to a distribution over hypotheses determined by the prior of the learners. This result has implications both for methods for developing methods for investigating the inductive biases of human learners, and for understanding how those biases influence the evolution of concepts and languages. I will describe the results of a series of experiments bearing out the basic predictions of analysis for simple cognitive abstractions such as functions and categories as well as more complex linguistic objects such as frequency distributions and systems of color terms.
Bio
Tom Griffiths is an Assistant Professor of Psychology and Cognitive Science at UC Berkeley, with courtesy appointments in Computer Science and Neuroscience. His research explores connections between human and machine learning, using ideas from statistics and artificial intelligence to try to understand how people solve the challenging computational problems they encounter in everyday life. He received his PhD in Psychology from Stanford University in 2005, and taught in the Department of Cognitive and Linguistic Sciences at Brown University before moving to Berkeley. His work and that of his students has received awards from the Neural Information Processing Systems conference and the Annual Conference of the Cognitive Science Society, and in 2006 IEEE Intelligent Systems magazine named him one of "Ten to watch in AI."
Return to 2008 Events Calendar