The Sausage Factory of the Mind: Hypnosis, Illusionism, and the Consciousness Myth
Looking at hypnosis and hypnotic suggestion is trying to make sense of a perceptual system that can replace "outside" perception with "inside" perception. There are several theories on how hypnosis can do this. Regardless of how the theories work, they have a common problem – they have to draw a line from the hypnotic suggestion all the way down to the brain's interpretation and the cognitive neuroscience that makes up the brain, all the way down to the suggested perception and its representation in the executive centers of the brain in the prefrontal cortex.
Cold Control Theory (and the follow up paper Phenomenological Control as Cold Control) is probably the best top level description of what's happening, as it explains involuntary action and changes in cognitive processing much better than Lynn's Integrative Sociocognitive Theory. However, I have struggled with Cold Control Theory, because it connects hypnosis with an model of consciousness called Higher Order Thought (HOT).
HOT is an abstract model of consciousness: it doesn't have a connection to the physical reality of the brain, whereas Integrative Theory maps down to predictive processing and active inference in the brain. I didn't trust a theory that appeared to rely on a free floating philosophical concept tied to something as fuzzy as consciousness.
But! There is a way out. It turns out that HOT is detailed enough that it can be mapped to an implementation, i.e. it can be connected to one of the neuronal models of consciousness like Global Cognitive Workspace Model. I really like a lot of the work that's happening in these neuron models -- in particular, there's some great research discussing how these models explain p-consciousness and access consciousness and seeing how these models could be implemented in practice using recurrent neural networks. Likewise, these models can be mapped to biologically plausible recurrent neural networks, providing an analogue of the brain that reflects cognitive neuroscience.
This stepwise approach from top down Phenomenological Control to the substrate of predictive processing is much more cohesive than the Integrative Model, which dives straight from response set to the low level implementation of predictive processing and active inference.
Here's a rough diagram showing the hierarchy from high level concept to low level processes.
Even with an end-to-end model of consciousness, I still felt uncomfortable. I don't think this is a problem with Cold Control, but more about the general concept of consciousness. The way we describe consciousness makes me itch.
Cognitive neuroscience often uses vision to demonstrate the disconnect between the product and the engine. Regardless of how the human brain perceives vision as a seamless whole, the actual engine that produces that perception is utterly unlike our perceived experience. To compare our experience of vision with the actual mechanisms that create vision is to confuse a sausage with a sausage factory.
In the same way, cognitive consciousness – our perception of our own thoughts and cognition – is not a privileged representation of how cognition is constructed. Dennett posited the desktop metaphor, in which a simple graphical user interface can look simple while being a complex interaction of billions of concurrent instructions in a CPU interacting with a graphic processor unit. This philosophical stance is Illusionism. Keith Frankish does a great job of explaining illusionism as a theory of consciousness and I broadly agree with the argument. I would say that rather than saying "I am conscious" it's more accurate to say "I perceive consciousness" – and even that perception is necessarily the end result of many unconscious perceptual engineering layers.
In short, when we talk about neuronal models of consciousness or the cognitive neurosicence behind it, we are not really talking about the the end product -- the perception of consciousness -- but about the possible workings of the consciousness factory.
More than that, I would have to argue that the idea of consciousness is a cultural hallucination. In hypnosis, the idea of trance as a "hypnotic state" is so ingrained since Trilby that it has been called the mother of all myths by some academics. Despite the certainty that a hypnotic state exists, there is no set definition of what this hypnotic state is, or how it can be identified – believers have their own definitions and their own certainty despite the lack of commonality. In the same way, consciousness (and ancillary concepts such as stream of consciousness) is a label that gets slapped on anything that looks potentially relevant to the subject at hand.
As an example, selfawarepatterns reviews this debate on consciousness and concludes "the differences between these theories largely amount to varying definitions of consciousness." There are papers that point this out and suggest that all of them could be valid.
My favorite explanation is from Attention Schema Theory, which is a cognitive schema of the brain's own attention at a high level. Because this is an abstracted representation of the brain itself, the brain conflates the model with the reality. The author, Graziano, has already implemented a neural network with an attention schema and demonstrated that reflection can dramatically improve model performance.
Graziano also has a number of things to say about bridging the gap between the "physically incoherent consciousness we think we have and the complex, rich, but mechanistic consciousness we may actually have." Graziano defines two different kinds of consciousness:
i-consciousness (informational): How information is selected, enhanced, and processed in the brain - explained by Global Workspace Theory
m-consciousness (mysterious): The experiential essence people claim to have - the "hard problem"
And describes different theories as different perspectives on the problem.
Global Workspace (GW): Explains i-consciousness - attentionally enhanced information reaching brain-wide networks
Higher-Order Thought (HOT): We contain meta-information about how we process information
AST: Provides the social-cognitive model explaining why we claim to have subjective experience
Illusionism: Consciousness as the brain thinking it has something it doesn't actually have.
Cold Control Theory and HOT are perfectly valid high level theories, and there's no argument that consciousness is a lot easier to say than "metacognition" or "reflective executive processing" but I do quietly hope that the terminology becomes more concrete and we move away from the idea of consciousness altogether.