February 7th 2026

Consciousness Is An Adaptation To Sparse Data And Inflexible Goals, Whilst Experience Is A By-Product

Complex goals must be precisely controlled to maximise the persistence of the system that pursues them and the continued pursuit of more fundamental goals. Regularly, goals must be created, dissolved, and modified. Goal setting must respond to sparse examples, or no examples at all, to fulfil more fundamental goals that are otherwise not fulfillable by the current neuronal thermodynamic goal landscape. Consciousness provides the solution. Consciousness is a medium by which the established neuronal routes may be circumnavigated to fulfil some desired target goal state which would otherwise be thermodynamically inaccessible. The target goal state may already be mapped in the brain or may be novel in which case the medium is queried with approximations of the target goal state. Querying the medium is akin to dialling a number and being connected to a specific receiver. Whenever a specific target number is not known, that number is approximated and receivers with similar numbers to the desired one are called, triggering an excitation in those receivers which then propagates via neuronal action potentials in a stochastic manner, potentially triggering and thereby discovering the desired target goal state. Querying a receiver through the medium gradually builds up thermodynamically favourable neuronal routes between the previous goal state and the target goal state. Once this connection has been established enough times then this flow of the goal landscape becomes executable through neuronal action potentials and no longer requires conscious thought for fulfilment. Experience may simply be a by-product of querying and receiving from this medium. All conscious experiences can be considered a by-product of such goal-overrides. This system permits rapid learning from sparse or no data. AI systems will be at an inherent disadvantage without an equivalent system and will rely on massive training sets that lower the likelihood of surprise.

Introduction

Goals, as outlined in McAuliffe 2026, "The Nature Of Complex Goals And The Implications For AI Safety And Consciousness", can be considered as the circumstances to which the universe tends. Defined as such, goals need not be biological and they encompass both attractors (rewards) and repellers (punishments).

As explored in that article, a system pursuing complex goals, particularly a conceptually intelligent one, must navigate a possibly infinite space of "goal traps". These include sub-goals that compete, goals that are impossible to fulfil, and goals that are paradoxical.

Ineffective navigation of these goal traps represents a goal-persistence disadvantage. It is circularly true that those goal pursuits that most effectively persist from one moment to the next, do so. Such evolutionary principles are not constrained to biology. Evolution is simply a consequence of competing goals in an increasingly disordered universe.

It is feasible that sophisticated feedback mechanisms might permit the navigation of such goal traps via traditional computational means. Examples could be energetic-investment bounded switches or overrides from other goal-pursuing sub-systems. However, given the complexity of the goal trap landscape, which only increases as conceptual intelligence does, it remains plausible that it is impossible to form the necessary pre-meditated feedback systems to avoid all possible goal traps. As such, the author proposes that the effective navigation and escape from such traps depends on consciousness.

However, the outlined goal traps often depend on a high level of conceptual intelligence. Yet consciousness, by most estimates, significantly predates this type of high-level conceptual intelligence. The most compelling evidence for the pervasiveness of consciousness comes from the effective use of general anaesthetics in a wide range of animals. They also work on babies, providing strong support for the notion that consciousness is not an emergent property of knowledge or conceptual intelligence, though not countering the notion that it is emergent from learning ability. Though consciousness is enriched by experience and knowledge, it is not dependent on it. Given its widespread and developmentally early presence, the evolutionary pressure for consciousness must be tremendous.

Rather than only considering advanced logical goal traps, we can instead cast a wider net over any deviation from the current goal landscape. The goal landscape represents the favourable direction of travel and its traversal is automatic, like a marble running down a winding track. Changing the track is slow and energetically expensive and its reconstruction requires repetition. Yet somehow humans must learn from little to no data and adjust their established behaviour accordingly. Consciousness permits this.

Argument

Those who effectively pursue vague, complex, and long-term goals will recognise how easy it is to tunnel vision on some sub-goal, to vacantly pursue it, only to later step back at some moment and realise it was utterly futile or counterproductive for the overall complex goal. An LLM acts as humans do in those "vacant" goal pursuit moments. They work deterministically, tunnelling down a path, following a sequence of neural patterns that were determined by the training and the input.

There are countless examples of such vacant, automatic, and sophisticated behaviours. We carry out our morning routines, drive for miles, craft complex objects, all without being aware of the actions as they are happening. These are habitual behaviours. In these moments it is actually other observers who are more conscious of our behaviours than we are. What is practice if not moving intentional, effortful, new actions from the conscious minds to the unconscious? Musicians perform entire concertos without thinking once of their fingers, sportspeople perform remarkable kinaesthetic feats without consciously deciding how to move their bodies, and we repeat rehearsed speeches and phrases without considering what word comes next. These moments are like a marble automatically rolling down a run.

Consciousness is not binary. It waxes and wanes even when we are awake. When are we most conscious? In those circumstances that are most unexpected and most rewarding or punishing. Love at first sight, a stubbed toe, a jump scare. But unexpected is not just a surprise as we intuitively understand it. Unexpected is also a problem we cannot solve, a movement we haven't yet mastered, a song we haven't yet memorised. Unexpected is something not experienced or predicted. Unusual things happen all the time, constantly, around us. It is the degree to which the unexpected thing impacts a goal, either towards a repeller or towards an attractor, that determines how conscious we become of it. Almost everything around you right now is a surprise (unless you are sitting in a very familiar environment). Every combination of sound and sight and touch you are being exposed to in any given moment is new, but they are inconsequential according to your pursued goals. A scare is not inconsequential. A scare is perceived as genuine danger and it should be learned from. An unsolved mathematical problem for some people is not goal-consequential and for other people it is.

Why do we become more conscious in these circumstances? Why don't we just learn through unconscious routes? Every goal you are currently automatically pursuing is a neuronal highway that has been paved and widened to lesser and greater degrees. The likelihood of a thought travelling down any given highway is proportional to how often it has been activated. Just like LLMs, we learn through repetition. Constructing these roads takes time and effort.

The world requires that our brains respond faster, though. Often we must learn off just one example or maybe no example at all. Creativity is achieving some goal without prior examples. It includes new movements that one has never previously tried, new thoughts, new sounds.

Think of mastering a new skill, like a tennis serve. You may manage to move your body in the right way just one time. How can you recreate it? All your circuitry is based on your habitual, incorrect, technique. Somehow, off just one example, you must repeat the action, over and over, until it becomes habit. Yet the necessary neuronal highways are either not there or barely formed. Worse still, what if you need to solve a novel problem with no previous example of how to solve it, like a maths problem. There cannot possibly be a representation in the mind to solve it, unless by sheer utterly improbable chance. And then you have to do it for another problem and another and another. The mind does not have the mappings to solve all these novel problems. There are no established neuronal paths to get to the solutions.

So how do we do it? There are clues in our behaviours. What do we do when we are trying to remember a forgotten word? We say things near the word. We make gestures that the word represents. We dance around it. I suggest that creativity works the same. We dance around the solution that we don't yet have, randomly stimulating neuronal patterns that are close by until, for some of us, it clicks, and the right pattern forms by chance.

But how do we carry out this fumbling if it is not thermodynamically favourable for the neurons to do so?

Consciousness allows us to call those patterns from afar. Consciousness provides a means to jump between and circumnavigate the neuronal roads.

When I call a phone number, the network routes my phone, by means of which I am ignorant, to the specific recipient I intend. This involves wave phases, amplitudes, and frequencies in the electromagnetic field. Consciousness, I suggest, acts the same way.

The phone number we dial is the intended goal state that we wish to reach. This goal state is often rich and complex. It could be a mixture of emotions, a motion, a thought. The receiver is the target goal pattern which, once triggered through the medium, propagates action potentials in the standard way. Sometimes we do not know the "number" we need to dial but we know similar numbers whose excitations are likely to stimulate the intended target pattern through stochastic action potentials.

Others have made the innovative suggestion that consciousness could be quantum (Penrose, 1989) or a field (Pockett, 2000; McFadden, 2002). Either could feasibly fulfil the role that I suggest consciousness performs. Most importantly, I agree that it is not emergent from neuronal action potentials nor will it emerge from transistor switching.

The conscious medium is queryable and answerable. It permits both the transmission and receipt of signals, where those signals possess unique and identifiable features so that the signal may reach the intended targets, just like the electromagnetic field required for our communication networks do. Action potentials interact with the medium, either directly or indirectly. The medium permits the circumnavigation of the current neuronal goal landscape that otherwise would block the fulfilment of the queried goal because the fulfilment of that goal is outside the bounds of action potential thermodynamic favourability. The intensity with which we query and receive from the medium is dependent on how unexpected and how rewarding or punishing the current circumstance is. The stronger the signal propagated through the medium, the more likely and forcefully the receiver will trigger. Sometimes, the intended recipient (a complex neuronal pattern representing a new goal state) cannot be triggered, leading to frustration and abandonment of the goal. Different systems may query the medium simultaneously with competing goals. An example is holding one's breath, which requires conscious intervention and is easily accomplished for a time but which is gradually overwhelmed by the pursuit of the breathing goal, this pursuit being done with increasing force, both through automatic neuronal pathways but also increasingly consciously. We can likely train our ability to query and respond to this medium; willpower.

Given that solution breakthroughs and memory formation are observed to occur during sleep, it is plausible that sleep is fundamental for reinforcing the neuronal patterns, the new goal-landscapes, that were consciously faintly etched out the day prior.

Is this free-will? It feels that way when we battle the current neuronal goal landscape that we would otherwise favourably pursue. But what defines the goals with which we query the medium? What defines the strength with which they are propagated? The illusion reveals itself. The act of querying the medium with a goal and overriding the default action potentials of our neurons feels like a "wilful" overcoming of our predetermined fate. But we never chose the queried goal state. Free will is simply the feeling of resistance. The capacity to stimulate the medium, to trigger an otherwise inaccessible target goal state, to stand against the tide, is likely some bodily trait over which we have no control, and the goal state that was queried, or "dialled", was chosen by processes outside of our control. Even training our capacity to query and respond to the medium, the build-up of willpower, can follow these same deterministic principles. Our "free-will" is simply competing goals using different mediums to fulfil them. "We", are deviations from our own habits.

Why do we need to experience any of this? We probably don't. Experience may simply be a by-product of communicating through this medium. We have been looking for a reason for experience when it might just be a side-effect.

Is every conscious experience just a break from the current neuronal goal landscape? I argue yes. We blink automatically yet can stare when we have a goal that requires it. We may know the target recipient pattern well, memorised its phone number, thereby making its conscious stimulation relatively easy, but it is not an automatic action. Gradually, the competing goal of dryness gets stronger and stronger either through favourable automatic neuronal pathways or consciously or both, making it harder and harder to either transmit the signal through the medium to keep our eyes open or to receive it.

What about experiencing something joyful that is not so surprising? Well when does an experience feel its best? When we aren't thinking about it too much. We can of course focus on what is feeling good but that is not automatic and usually interferes with the flow state. Likely, every joyful experience is brimming with little moments of unexpected reward that make us dip in and out of the conscious medium to learn how to get that reward again, making the overall experience feel good. But it is only the small pleasant surprises that are being consciously experienced.

What else? We can speak without consciously considering the words or we can consciously pick each word according to some queried target goal. We can walk with a normal stride or suddenly adjust it to something unnatural, we can stay at rest or summon the energy to exercise.

Consciousness is advantageous when a break from the norm is required and just gets in the way when the circumstances are familiar. Evolution has had to balance on this tightrope in order to learn from sparse or non-existent data and execute actions efficiently in familiar situations. If my consciousness were first invoked upon slipping on the pavement, I would surely fall.

Consciousness, like an image or a sound, is generally a cacophony of many facets, or "waves". It may be composed of sensations like sound, touch, or taste and with emotions like fear, jealousy, or joy. It can induce parallel actions; I can simultaneously intentionally hold my breath, blink, and touch my fingers together all with conscious thought. The target goal state consisted of all these actions, the call went out through the medium and the multiple receivers picked it up in parallel.

This mechanism would intuitively form the illusion of free-will. Yet the target goal landscape with which we queried the field, the "desired outcome", the "phone number" we dialled, was likely put there by processes outside of our control.

Consciousness is a jump between unconnected or poorly connected patterns of brain activity. The more the jump is made, the more the neuronal activation patterns before and after the jump become connected by traditional neuronal circuitry. In this way, this specific flow of the goal landscape transforms from being dependent on consciousness to being fulfillable without conscious involvement.

What Does This Mean For AI?

AI may not need consciousness as it will be trained on far more data than humans are and it will have a much greater capacity to learn, extract representations and patterns, and memorise. However, the universe is not computable (Penrose, 1994; McFadden, 2002), it will always surprise in a goal-consequential manner and in those moments, an AI goal-pursuing system without a consciousness or consciousness-akin system will be at a persistence disadvantage.

Testing This Theory

If consciousness is correlated with a greater energetic burden due to more long-range remodelling of the neurons then greater ATP production in the mitochondria could be expected. This increased metabolism could be detected, though it would be explainable with other theories. We could also test some goal pursuit over and over and see how, as it evolves from a conscious pursuit to an unconscious one, an MRI or EEG changes from a "jump" to a "connection". If we can send and receive information through this conscious medium to specific goal state recipients, then we will obtain strong evidence in favour of the hypothesis. If consciousness is invoked without goal-consequential stimuli, be they conceptual or sensory, then that would provide evidence against this hypothesis.