Prediction plays a fundamental role in cognition. Accurate prediction allows humans and other animals to act in ways that anticipate future states of the environment, potentially reducing the threat posed by dangers and increasing the benefits of positive events. Prediction also permits cognitive systems to forecast what will happen beyond the next instant, or forecast what would happen were the world different. Cognitive systems can plan, strategize, and learn by using prediction and forecasting to adapt to a changing environment.
The theory of predictive coding suggests that higher cortical areas integrate environmental information from perception and sensation and contextual information from memory to generate hypotheses about the state of the world. Subsequent sensory feedback is integrated into this information to help detect differences between the original hypothesis and the actual state. These differences, called prediction errors, can be utilized to update hypotheses, generating new predictions about how the environment is changing. In the past two decades, theory and research have made central the importance of predictive coding in the computational foundations of both human perception and cognition and machine learning. In this seminar, the merits and pitfalls of this approach to understanding the brain and cognition will be explored.
Speakers:
David Danks, Professor of Philosophy and Psychology, Carnegie Mellon University
Karl Friston, Wellcome Principal Research Fellow and Scientific Director, Wellcome Trust Centre for Neuroimaging; Professor of Neurology, University College London
Carol Krumhansl, Professor of Psychology, Cornell University
Moderator:
Christopher Peacocke, Johnsonian Professor of Philosophy, Columbia University
This event is free and open to the public. Reception to follow.
This event is part of the Seminars in Society and Neuroscience series.
