Bayesian Baby

the hallucination we call consciousness, or how we start the long road to reality

Lewis has got me thinking about how we make sense of the world. Here he is, just barely a month old. He can’t grasp the idea of who he is, that he is an individual, separate from the sensations of which he is partaking. For that matter, even grasping with his hand is very much in an early stage of development.

Week by week his gaze seems to be taking on an intentionality. Looking here, then slowly looking there. Settling on a face. Many of his facial expressions are inscrutable, but one has to assume that his 12-14 ounce brain is assimilating data at an astounding rate. No wonder he sleeps so much.

I recently watched a TED talk by Anil Seth (here) about consciousness. A Professor of Cognitive and Computational Neuroscience at University of Sussex (U.K.), he describes his theory of consciousness as a controlled hallucination.

Imagine being a brain. You’re locked inside a bony skull, trying to figure what’s out there in the world. There’s no light inside the skull. There’s no sound either. All you’ve got to go on is streams of electrical impulses which are only indirectly related to things in the world, whatever they may be.

So perception -- figuring out what’s there -- has to be a process of informed guesswork in which the brain combines these sensory signals with its prior expectations or beliefs about the way the world is to form its best guess of what caused those signals. The brain doesn’t hear sound or see light. What we perceive is its best guess of what’s out there in the world.

The theory reflects Bayesian logic¹ – that we make assumptions about the world and constantly update our prior assumptions, revising what we know as we gather new clues.

An actual volume in Lewis’ library, a gift from Lauren Gilbert. It’s an insightful and funny look at how to predict the likelihood of randomly getting a bite of cookie that has a piece of candy in it.

It is not hard to imagine Lewis engaged in this process of successive approximation during his waking hours. Each movement of his arms tests some elementary idea of what is “in reach” based on what he touches, versus when his slow swing catches nothing but air. When he and I looked out of the front window today to view the approaching rainstorm, he seemed to scan the salient features with a pre-verbal interior monologue along the lines of “Big square frame, used to have lots of light inside. Note to self: it looks really different this time.”

Hundreds of trial arm movements each day, hundreds of impressions all contributing to some catchment of assumptions about the world around.

Dr. Seth goes on to point out the distinction between perception — how we take in information about the world around us — and interoception, which is the term for how we sense what is going on inside our physical body.

Experiences of the body from the inside are very different from experiences of the world around us. When I look around me, the world seems full of objects -- tables, chairs, people. Even my own body in the world; I can perceive it as an object from the outside.

But my experiences of the body from within, they’re not like that at all. I don’t perceive my kidneys here, my liver here, my spleen ...I don’t know where my spleen is, but it’s somewhere. I don’t perceive my insides as objects.

In fact, I don’t experience them much at all unless they go wrong. And this is important, I think. Perception of the internal state of the body isn’t about figuring out what’s there; it’s about control and regulation — keeping the physiological variables within the tight bounds that are compatible with survival.

When the brain uses predictions to figure out what’s there, we perceive objects as the causes of sensations. When the brain uses predictions to control and regulate things, we experience how well or how badly that control is going.

What are the percentages of perception and interoception in Lewis’ waking consciousness? I am sure someone has a benchmark for the first months of life. I’m guessing there is more than we might think in the direction of “keeping the physiological variables within the tight bounds that are compatible with survival.”

For Lewis this is expressed in some pre-verbal version of “All you who have ears, listen!” followed by one or more of a few specific items pertinent to control and regulation: hunger, dampness, sleepiness. The list is not long, but the signal is clear — some physiological variable requires outside intervention, preferably sooner than later.

Meanwhile, here am I — deep inside my controlled hallucination, and have been for decades. Fortunately, as Dr. Seth observed, “We’re all hallucinating all the time, including right now. It’s just that when we agree about our hallucinations, we call that reality.” And so far, it has worked pretty well, as evidenced each time I use a pedestrian cross-walk and cars stop for me.

Image from Anil Seth’s Being You book cover

1. For more information see Seth, A.K. (2015). The cybernetic Bayesian brain: from interoceptive inference to sensorimotor contingencies. In Open MIND, eds. T. Metzinger & J. Windt.