The end of a year and the start of a new year brings many predictions about things to come. Predictive coding has nothing to do with “coding” computers or predicting trends and everything to do with our personal neuroscience.
The classical view of perception states that we experience the world by receiving input from our environment, processing it at the higher levels of our brain, and then responding accordingly.
A newer alternative theory proposes to add to those three steps that our higher faculties often “predict” the input from our environment. That means we have a perception of some things before we experience it. This is called predictive coding or predictive processing.
I read an article by Sara Briggs and then followed up with another titled “To Make Sense of the Present, Brains May Predict the Future.” In those readings, I encountered this theory (still controversial) that suggests that perception, motor control, memory, and other brain functions all depend on comparisons between ongoing actual experiences and the brain’s modeled expectations.
The next day I noticed a connection when my son’s visiting dog seemed to do some predictive processing. Pepper reacts to her doorbell at home by barking and sprinting to the front window. We were watching the movie Love Actually and in one scene Hugh Grant’s character rang a series of doorbells looking for a woman’s home. Even though these were different doorbell sounds from the sound in Pepper’s home, she reacted to each ring in the same way that she does at home. Her actual experience in my home and her brain’s modeled expectation created a match.
One way scientists look for evidence to support this theory is to look at cases where the brain predicts too much or too little. For example, individuals with autism would presumably have a weak predictive filter. That would mean that they have a harder time categorizing items based on past experiences. They would have an extreme sensitivity to input from the environment and the many “new” experiences could be overwhelming.
A person with schizophrenia would be at the other extreme with an overly strong predictive filter. Their brain would be so certain about what it’s looking at, it will cancel out new information and have false perceptions, possibly even hallucinations.
What is considered “normal” is somewhere in the middle of this spectrum.
Of course, we can change that by changing our brain chemistry. That is why some research uses psychedelic substances. Some neuroscientists might say that our “normal” perception is a “controlled hallucination.” Substances like psilocybin and LSD remove the predictive filter and so when we under that influence someone sees something common to daily life, such as a tree, there is no prediction and it alternative perceptions emerge. The branches moving in the wind are arms and the leaves are flames. The drugs don’t add to perception but by removing the filter they allow other possibilities.
How does this predictive coding affect learning new things?
To learn new things we need to be open to new perceptions which means the filter must be reduced to some extent. But in order to retain the new information and use it in the future, we need a predictive model of that information, which requires that filter to be operating normally. When the two are balanced, learning and memory are optimized.
In a more simplified explanation, being open-minded should lead to greater learning. We don’t put information in a box and move on.
Some of this theorizing isn’t new at all. Back in the 1860s, the framework known as the “Bayesian brain” was introduced and Helmholtz’s concept of unconscious inference emerged. It proposes that the brain makes probabilistic inferences about the world based on an internal model, – it calculates a “best guess” interpretation of what it’s perceiving. The name comes from Bayesian statistics which quantifies the probability of an event based on relevant information gleaned from prior experiences.
These “controlled hallucinations” based on predictions don’t wait for all sensory information to drive cognition. We are constantly constructing hypotheses about the world. We use these to explain new experiences. The brain is constantly generating and updating a mental model of sensory input.