Prediction and predictive codes are now ubiquitous computational viewpoints from which we may better understand neural circuit organization and signal transmission in the brain.

In this talk I will present a predictive view of changing brains, over lifespans, based on the Free Energy Principle, a theory of hierarchical empirical Bayesian inference in the brain (Friston 2013). This particular formulation of the Bayesian brain produces predictive coding schemes that have been used to inform the principles of perception, action and decision-making, accounting for how sensory information combines with our own prior beliefs about the world to shape brain activity and behavior. There are many ways that a brain could perform Bayesian inference and the hypothesized scheme under the Free Energy Principle in the perceptual domain posits a variational algorithm where posterior density estimation is recast as an optimization problem. In this guise the scheme becomes a predictive coding algorithm, with hierarchical structure and attribution of optimization dynamics to particular components of neuronal circuits.

In this talk I will present evidence from neuroimaging studies of brain circuits (using dynamic causal models) that age-related connectivity changes are commensurate with long-term Free-Energy minimization. I will present work from sensory learning, memory and decision making paradigms that show that the neurobiological implementations of prior beliefs grow stronger in older brains. I will explore how this relates to faster timescales of prediction in terms of electrophysiological correlates and in decision making paradigms.