The Meta-management Theory of Consciousness uses the computational metaphor of cognition to provide an explanation for access consciousness, and by doing so explains some aspects of the phenomenology of consciousness. For example, it provides explanations for:

1) intentionality of consciousness - why consciousness "looks through" to first-order perceptions etc.
2) causality - that consciousness is "post-causal" - having no causal power over the event to which we are conscious, but having direct causal effect on subsequent events.
3) limited access - why we only have conscious experience associated with certain aspects of brain processing.

https://malcolmlett.medium.com/the-meta-management-theory-of-consciousness-17b1efdf4755

I know that I am going to get roasted by this wonderful and thorough community, but I wanted to share my theory both for your own enjoyment and so that I can learn from the discussions. Some of you might remember me from a long time ago. Those discussions were extremely useful, and have helped me entirely revise both my theory and how I describe it. I really hope that you'll enjoy this. Roast away.

I'll try to summarise here as best I can. The details are in the blog post - it's only a 40 minute read.
I'll explain it here in two parts. The first part is the computational argument, which depends on knowledge of predictive processing paradigms such as that of Predictive Coding or the Free-Energy Principle. I know that it's easy to get lost in the details, so I'll try to summarise here just the key takeaways.

The second part will focus on how I think my theory makes headway into the trickier philosophical debates. But bare in mind that I'm a little clumsy when it comes to the exact philosophical terminology.

Computational background:

• The theory starts with a computational discussion of a particular problem associated with cognition: namely deliberation. Here I am using cognition to refer generally to any kind of brain function, and deliberation to refer to the brain performing multiple cycles of processing over an arbitrary period of time before choosing to react to whatever stimuli triggered the deliberation. That is in contrast to reaction where the brain immediately produces a body response to the stimuli.
• I argue for why deliberation is evolutionarily necessary, and why it is associated with a considerable step up in architectural complexity. Specifically that it is "ungrounded" from environmental feedback and can become unstable. To resolve that it requires meta-cognitive processes to "keep it in check". I label the specific meta-cognitive process in question as meta-management.
• Meta-management is best described in contrast to 1st-order control. Under 1st-order control, a 1st-order process has the task of controlling the body to respond appropriately to the environment and to homeostatic needs. A meta-management process is a 2nd-order process tasked with the control of the 1st-order process, particularly from the point of view of the homeostatic needs of that 1st-order process. For example, the 1st-order deliberative process can become ungrounded as mentioned before, and it is the task of the 2rd-order process to monitor and correct for that. In the blog post I explain my concretely what I mean by all this.
• I propose a functional structure whereby the 1st-order order process can meta-manage itself without the need for an explicit separate 2nd-order system. I explain how this is possible via predictive processes (think bayesian inference), in conjunction with a special kind of feedback loop - the meta-management feedback loop.
• The Meta-management Feedback Loop captures certain aspects of the current state of cognitive processing and makes them available as a cognitive sense, which is perceived, processed, modelled, predicted and responded to like any other exteroceptive or interoceptive sense. I provide a computational and evolutionary argument for why such a feedback loop captures only a minimal amount of information about cognitive processes.
• While it seems at first counter-intuitive that such a system could work, I argue how it is possible that a 1st-order process can use this feedback loop for effective meta-management of itself.
• Looking at just cognition alone, this provides a general framework for understanding deliberation, and by extension for human's ability for abstract thought (I explain this last point briefly in the blog post, and in more detail in another if anyone is interested to know more).
  
  So that all leads to this functional structure which I argue is a plausible mechanism for how human brains perform deliberation and how they keep their deliberative processes "grounded" while doing so:

I have no strong empirical evidence that this is the case for humans, nor even that my suggested architecture can actually do what I claim. So it's a thought experiment at the moment. I plan to do simulation experiments this year to build up some empirical evidence.

Implications for understanding Consciousness:
Now I shall try to argue that the above explains various aspects of access consciousness, the phenomenology of consciousness (ie: specific identified characteristics), and that it might even explain the existence of phenomenological consciousness itself.

Conscious Contents. The Meta-management Theory of Consciousness (MMT from hereon) suggests that the contents of consciousness is the result of whatever is fed back as a sensory input via the Meta-management feedback loop (MMFL). The predictive theory of perception suggests that cognitive processes in relation to exteroceptive and interoceptive sensory signals do not take those signals literally and use them in their raw form throughout. Rather, the brain first uses those sensory signals to predict the structure of the world that caused those sensory signals. This creates a "mental model" if you will, of what is "out there". The cognitive processes then operate against that model. I propose that the same goes for the sensory signal received via the MMFL. This creates a cognitive sense that is processed in the same predictive fashion as for other senses. To be clear, I mean that the "cognitive sense" provides a sensory signal about the immediately prior cognitive activity, in the same way that a "visual sense" provides a sensory signal about optical activity. And that then this cognitive sense is subject to the same predictive processes that create a "mental model" of the structure of whatever caused that cognitive sensory signal.

Here I am talking about access consciousness. Why do we have access consciousness at all? I propose that it is because of the MMFL and its associated cognitive sense. There's nothing new in my proposal that we have access consciousness because the brain has access to observe some of its own processes. What's novel is that I have grounded that idea in a concrete explanation of why it might have evolved, in the specific computational needs that it meets, and in the specific computational mechanisms that lead to that outcome.

Intentionality. MMT provides the background for Higher-Order Thought (HOT) theory. It describes that the cognitive sense is a HOT-sense - a specific second-order representation of the first-order process. I won't go on at length about this but I'll draw out some specific nuances (again, there's more details in the blog post):

• The intentionality of access consciousness is the behavior of the first-order cognitive processing system. It exists to support meta-management, which needs to monitor that behavior. This means that the intentionality is not of our first-order perceptions, per se, nor even of our thought, per se. Rather, it is the static and dynamic state of the first-order process while it operates against those first-order perceptions/thoughts/etc. In a very practical way, that static and dynamic state will inevitable capture aspects of first-order perceptions and thoughts, but it will always be by proxy. The cognitive sense about the immediately prior cognitive activity represents the fact that the brain was thinking about an apple that it sees, rather than directly representing the apple seen. Or perhaps the cognitive sense represents that the brain was at that moment attending to the visual perception of the apple, and thus by proxy the cognitive sense captures many aspects of that visual perception, but not the actual visual perception itself.
• In contrast to some debates about HOTs versus HOPs (higher-order perceptions), MMT suggests that the cognitive sense is a higher-order representation of cognitive activity in general. Thus it is neither accurate to refer to it as a higher-order thought nor as a higher-order perception. Better terms could be Higher-order State or Higher-order Behavior.

Causality. One of my favorite learnings from this exercise is that consciousness is what I call post-causal. Events unfold as follows: 1) the brain does something, 2) the MMFL captures that something and makes a summary of it available as a cognitive sense, 3) the brain gets to consider the cognitive sense that is now being received. Anything that the brain considers "now" in relation to the cognitive sense is about something happened earlier. Thus our conscious experience of cognitive activity is always one step behind. Consciousness has no causal power over its intentional object. However, conscious access very much has causal power over everything that follows.

Phenomenological Consciousness. I doubt that I have the philosophical savvy to make any real headway in arguing that MMT explains the existence of phenomenological consciousness. But I do believe that it does and I'll try to state the beginnings of an argument.

• 1) All exteroceptive and interoceptive senses undergo transformation from raw signal to a prediction of the world structure that caused them. In some circles this is known as the inferred latent state. I believe that the brain doesn't just infer the latent state, but also associates other meaning such as an inference of how that particular latent state relates to the needs of the individual. So, for example, the perception of a lion without any protection in between is associated with knowledge of danger. Part of that associated meaning is identification of self vs other (I know that the lion is other, and that my body is part of my self, and that it is my body that is in danger). This depends on a mental model of what a "self" is.
• 2) In the same way, our cognitive sense can be further processed and meaning associated to it. In particular, identification of the relationship between the cognitive sense and the "self". Thus any further cognitive processing of this cognitive sense has immediate access to the full enriched concept of "self" and its relationship to to whatever the cognitive processes were doing immediately beforehand. This would seem to me to be the underlying mechanism behind what Nagal called "what it feels like" to be conscious.