Bayesian Belief Update

One of the leading models of how we learn is based on the groundbreaking work Karl Friston, the father of modern neuroscience. It was with the algorithms of his statistical parametric mapping that we began to unlock our ability to use fMRIs to see the mind in the act of thinking. This ability is completely transforming how we understand it – leading to a deeper appreciation of the Whole Mind.

At the core of Friston's work is something called Bayesian inference . This statistical method is focused on how one might best predict based on current knowledge. As a simple example, if someone was right handed and had a minivan, based on what we know about being right handed and minivans, which might be better to predict if that person had a family?

Here is how that statistical method is being applied to neuroscience:

>A central function of the nervous system is to use sensory information to infer the causal structure of the external world. According to Bayes' rule, the optimal way of using this information is to calculate the information's likelihood under various models of the environment, and to weight this likelihood by the strength of prior belief in each model to derive posterior beliefs. In recent years, the influential hypothesis has been advanced that Bayesian inference represents a unifying principle of neural computation (the Bayesian brain hypothesis. source

Friston developed a process theory called 'active inference' to explain his model for how humans learn, indeed how all living things learn. At the core of this theory was his concept of the Free Energy Principle.

In the learning process, a current action is made based on set of assumptions developed by past experience. When there is an unexpected experience, a 'surprise', order is disrupted and must be restored.

To reestablish order, small, continuous experiments are used, something Friston calls Epistemic Foraging, until there is a moment that a new understanding – a new order – is established, what he calls a 'Bayesian belief update'.

>This leads to Bayesian belief updates that are informed by beliefs about the future (prediction) and context learning that is informed by beliefs about the past (postdiction). source

This theory appears to closely align with Piaget's concept of Schematic Accommodation and we suspect that these updates may be experienced as Eureka Moments.

DOT FROM preview-next-diagram