Depression and the Dark Room Problem

Trigger warning: depression, schizophrenia, mental illness

Predictive processing is an exciting new paradigm in computational neuroscience. Its essential claim is that the brain processes information by forming predictions about the world. Depending on who you ask, it's either going to solve everything, or turn out to be relatively uninteresting. I'll maybe discuss it in more detail in a future post, but today I want to focus on just one aspect of the theory.

A central principle driving predictive processing is error minimisation. Each prediction that the brain makes is compared with incoming sensory data, and this generates an "error signal" that reflects any mismatch between the prediction and the data. The brain is then driven to either make a more accurate prediction or modify its environment so as to conform with the inaccurate prediction, in order to minimise this error.

This leads to the so-called "dark room problem". If all we are driven to do is minimise prediction error, then why don't we just lie absolutely still in a dark room, thus enabling the formulation of a stable, accurate prediction? There are several ways of responding to this problem, but all share a general assumption that it is a problem, and that we aren't ever driven towards dark rooms.

Now, most of the time this is going to be correct, but on first hearing about the dark room problem my reaction was that actually I sometimes do just want to lie in a dark room. I suffer from periodic bouts of depression, and during these depressive episodes a dark room is pretty much all I can cope with. So perhaps whatever mechanism drives us away from dark rooms in everyday life is switched off during depression?

The Dark Cave Problem

This reminds me of an evolutionary theory of depression that I've heard of, which says that back when we were hunter-gatherers it made sense to occasionally withdraw from the world, as a survival mechanism in case of bad weather or other dangerous circumstances. In cases of depression this mechanism is simply over-sensitive or, in the worst cases, always switched on. I'm not sure how much I'm convinced by this theory, but lets assume that there is at least a shred of truth in it.

It also fits well with predictive processing and the dark room problem. Predictive processing has already been applied to the positive symptoms of schizophrenia and other delusions (in the form of "false" error signals), and similarly I think we could say that in some cases the dark room problem simply isn't a problem. Depression might be the result of a mechanism that shuts off whatever it is that drives us out into the world, with the result that we are content to minimise error by lying in a dark room.

On the other hand, depression and other mental illnesses are extremely complex, and I remain suspicious of any theory that tries to tell one simple story about them. Better perhaps to treat the dark room as just one of many contributing factors, or even just a useful metaphor.