Our capacity to construct conceptual and mathematical models is central to scientific explanations of the world around us. Neuroscience is unique because it entails models of this model making procedure itself. There is something quite remarkable about the fact that our inferences about the world, both perceptual and scientific, can be applied to the very process of making those inferences: Many people now regard the brain as an inference machine that conforms to the same principles that govern the interrogation of scientific data.
Karl Friston is rated as the most influential living neuroscientist due to his seminal work with brain imaging, the development of the Bayesian Brain theory and application of the variational free energy principle to our understanding of existing entities.
The several brian imaging techniques that Karl pioneered still dominate the field.
Many neuroscientists herald the Bayesian Brain model as providing the most promising attempt at a unified theory of brain functions. It models numerous brain processes as performing a selection from competing internal models of the outside world developed by the brain based on the sensory data available in support of each model. As more data becomes available, the probabilities of certainty for each model are updated by a process analogous to Bayesian Probability. Through a Darwinian process, selecting from the competing models the one best supported by the evidence, a basis for action is chosen.
Let's say we are in a fairly stressful situation, such as driving in the city, and we see something out of the corner of our eye that seems to be moving towards us. Because we cannot see the object very well, we may be unable to assign accurate probabilities to the various imaginable possibilities. Our brain develops numerous competing models of what the object and its significance to us might be, but as it has insufficient data to favour one model over the others, it must assign nearly equal probability to a number of the competing possibilities. We are in a state of uncertainty.
In response to this situation, unconsciously, we turn our head or eyeballs and bring the object into focus. We gather an accurate and relevant data set concerning the object, and our brain updates the probabilities assigned to its competing models based on this data. Usually, we can award one model a value approaching certainty and that model is selected to inform our actions. Our brains, through a rational Darwinian process, have made us better informed and optimized our response.
Since about 2005, Karl has turned his attention to the development of the variational free-energy principle (FEP), which generalizes the mechanisms of the Bayesian Brain to other domains of reality, including physics, biology and culture. The FEP states that existing entities contain models executing strategies that achieve their existence. In essence, the FEP says that all entities or systems strive to minimize the error or discrepancy between their model predictions and the existence they achieve.
Existing entities accomplish this in two fundamental ways, either by making the models more powerfully predictive or by taking actions that cause their model predictions to occur. The first way includes Darwinian processes, which, for example, natural selection, cause genetic models to predict new forms of phenotypes. The second is achieved by what Karl calls Active Inference, where the model instructions cause the predicted existence to occur, as with developmental embryology where genetic instructions bring the developing phenotype into existence.
Both of these general methods of minimizing free energy involve inference, or the relationship between models and evidence, and are analogous to Darwinian processes.