Why we only experience ‘Controlled Hallucinations’.
How the brain works according to Anil Seth and where it reinforces the work of Jeff Hawkins.
One of my first posts was about how “our brains are the measure of all things”, which specifically focused on the work of Jeff Hawkins. In this post I already mentioned another great book on the same topic, but viewed from a different angel, that of Anil Seth, Being you: a new science of consciousness. Let’s see where they differ and where these two theories reinforce each other.
The theory of Anil Seth on ‘controlled hallucinations’ is very similar to the theory of a Thousand Brains, but Seth uses a different terminology and arrives at his conclusions via a different route. What does Seth mean when talking about controlled hallucinations? Controlled hallucinations are internally generated perceptions, like a seeing or hearing that isn’t actually there. We never experience sensory signals themselves, we only experience interpretations of them. The essential ingredients Seth explains, are the following: First, the brain is constantly making predictions about the causes of its sensory signals, predictions that cascade in a top-down direction through the brain’s perceptual hierarchies. Besides the statement that the brain primarily makes predictions, just as Hawkins claims, these hierarchies can be compared with the different models and reference frames that filter the sensory signal and direct you to what your model estimates it is likely you are observing. Second, Seth explains: “sensory signals – which stream into the brain from the bottom up, outside in – keep these perceptual predictions tied in useful ways to their causes: which is whatever you are observing. These signals serve as prediction errors registering the difference between what the brain expects and what it gets at every level of processing. By adjusting top-down predictions so as to suppress bottom-up prediction errors, the brain’s perceptual best guesses maintain their grip on their causes in the world. In this view, perception happens through a continual process of prediction error minimisation.”[1] The third and most important ingredient Seth provides, is that the subjective experience of seeing an object, is determined by the content of the top-down prediction and not by the bottom-up sensory signals.[2]
Seth also has an explanation for how our brains decide what model our brains ought to invoke, which Hawkins explains by way of a ‘voting’ mechanism. The brain continuously tries to minimise prediction errors, or in other words, the brain wants to make sure it presents the most accurate model to you. First, our brains have what Seth calls generative models that determine the repertoire of perceivable things. Second are the perceptual hierarchies which determine the context for which my brain needs to make predictions. Seth: “If I know I’m in the zoo, perceptual regions of my brain will be more prepared for gorilla sightings than when “I’m wandering down the street”. Third, is the role of the estimated precision of sensory signals. This is very similar to paying attention, and this can also explain why we sometimes don’t see things, even when they are in plain view.[3] Take for example the experiment with the video where different people are passing a ball and you are given the task to count how many times the ball has been passed between people. Many people completely miss the gorilla that passes the scene. People will only see the gorilla the second time they are shown the video.[4] Although the mechanism is explained differently, I think the overall mechanism is the same as Hawkins voting mechanism. Our brains needs to choose between different models (or hallucinations). What ultimately determines the voting or prediction error minimisation is a combination of the repertoire of existing models and reference frames in the brain, which in turn determines what to predict and look for, which in turn determines what sensory signals enter your brain.
The conclusion we can draw from this theory is that there is no exact representation of reality in our brains. There are only reference frames that are continuously updated that help create models of the world. Some of these models might perhaps be very close to reality, but we can never be certain of it, and therefore never achieve it. What we do know is that we humans agree on many of these more objective models, and thus assume that this is an approximation of reality. Anil Seth and Jeff Hawkins’ theories not only sheds light on how we know, but it also explains why we are limited on what we can know.
[1] P82, 83 Anil Seth, Being you: a new science of consciousness
[2] P83, 84, Anil Seth, Being you: a new science of consciousness
[3] P107, Anil Seth, Being you: a new science of consciousness
[4] You can easily find the experiment online.
Bibliography:
- Anil Seth. Being you: A New Science of Consciousness (London: Faber & Faber Ltd, 2021)
- Jeff Hawkins. A Thousand Brains: A New Theory on Intelligence (New York: Basic Books, 2021)