Hamza's Website

This essay will be the first in a short essay series where I attempt to explain my understanding of the human psyche. This understanding is something I have cultivated over many years, and synthesis many seemingly disparate frameworks. Although the fundamental concepts and structure of the framework itself is simple, and can be seen in some of my prior essays (namely, the essay on the predictive psyche), I believe that it is difficult to grasp both its full importance and implications, and its deep grounding in historical research and modern literature. Starting off with everyday experiences, this series will slowly, sometimes laboriously, attempt to convey something that is perhaps beyond words. Little of the work and insights are my own.

Let us start off with a simple truth: there is importance and power in decomposing large, abstract theories into digestible chunks. If you have recently done mathematics, you will have a direct reference for this: concepts that seem completely incomprehensible at first over time become intuition and common sense. A recent example for me is the notion of Kolmogorov complexity. At first, the idea was completely beyond me: K(x) is the minimum length of an encoding of a program that could generate x when run on a universal Turing machine? I could barely say the word salad, let alone understand its implications and importance. But now, I see its elegance; it is simply the notion of compression wrapped up in the language of UTMs.

For anyone reading this who cannot recall the feeling I am describing, take Kolmogorov complexity as an opportunity to test this for yourself. Write the definition down, and look back at it over the course of the next few days or weeks. Whenever you get a chance, think about what the sentence is actually saying. Look up alternative formulations of what it is, examples of how to compute it, try going through some basic examples. Don't judge yourself or feel ashamed for how long it takes you to get it: explore with pure curiosity, and when necessary, compassion. Eventually, I guarantee that you'll wake up and realise that the concept has suddenly become tractable!

One interesting question that this raises: what is this process of decomposition and understanding? What does understanding something, reducing it to a wieldable, malleable, fixed yet pliable, mental object, mean? I used the word reducing there on purpose; I believe that reduction is primarily what is going on here.

Another simple truth: whatever we experience, we tend to view, and perhaps experience it, it in a way that makes sense to us. Kant put it pretty well: he posited that we view the empirical evidence of our senses using the presuppositions embedded within our tools of rationality. In this way, you can say that these presuppositions form mental frameworks. Prebuilt tracks for the ingestion of information.

Let us view this from another, modern perspective: Friston's free energy minimisation framework. Using sophisticated mathematics, it postulates that our brains are constantly minimising what he refers to as 'free energy'. Friston equates free energy to Energy - Entropy. Energy in this context is how surprising our observations and what we infer from them are, given some model of the external world. Entropy is how much we are hedging our beliefs. A very confident belief would have low entropy, and a less confident one would have higher entropy. Thus, by minimising free energy, we are simulatenously finding the least confident answer, that is the least surprising, taking into account what the implications of the observation are understood to be.

This framework assumes that we have something akin to a built in database in our mind of what to expect, and judge future experiences based off this expectation. Thus, our brain naturally attaches to experiences that matches this prior data, believing them to be of higher importance. However, alongside this, and perhaps more importantly, it will also conform our experiences to fit with this prior data, as such a conformation would be viewed as having a higher probability of occurring. Hopefully you can see the similarity between this and what Kant said! Kant said that we interpret the world through embedded presuppositions; Friston says that we contort our experiences to data accrued by prior experiences. They are essentially describing two different layers of the same phenomenen: Kant is describing fundamental presuppositions, forming interpretative frameworks that are fundamentally impossible to avoid. They define what reality is to us. On the other hand, Friston is describing the process of extending, and building more of these frameworks over time, perhaps on top of these more fundamental ones, based on data garnered through experience. Pay attention to the layering that this implies- we will touch on this in essay 2.

Friston's concept of free energy minimisation is, in many ways, a mathematical reformulation of something that has been understood in cognitive psychology for a while. The idea of mental frameworks, as just described, maps almost perfectly to the concept of a schema. Piaget first defined a schema in the 1920s as, roughly, 'a generalizable action pattern that the organism applies to objects'. Later, this was developed by Frederic Bartlett, another famous cognitive psychologist, to the field of memory and logic structures. He defined it as, roughly, 'an organised mental framework of knowledge and expectations that structures perception, memory, and interpretation of new information'. Both of these definitions and contributions, were, in their own ways, sidelined by the field for decades.

Bartlett arrived at this idea from his studies on memory and recall: he observed participants consistently and unconsciously fitting recalled memories to culturally familiar structures, even when there was no meaningful cultural connection.

Have you ever wondered how your brain is able to produce contiguous experience? Isn't it amazing that we can observe, process, and reflect, seemingly simultaneously? Think about it for a second: given the pure rate of information that we are almost perpetually ingesting, surely this could not be possible? In this way, the concept of schemas, and free energy minimsation, makes perfect sense. Brain's are not taking in all the information as if it were new: they ingest it relative, and according to, the mental frameworks that I had already composed, marking much as extraneous, unnecessary or redundant. Only through this contortion of reality could such an intractable feat of constant processing be made possible.

The fact that our brains can do this is a testament to how much redundancy there actually IS in the world. How much different experiences fundamentally resemble each other. How much different fields are doing the same thing. In formal disciplines, this looks like the emergence of elegant solutions, encapsulating chaotic, dynamic systems with simple equations and principles. Think of the eternal search for the theory of everything! Such a theory is not only being searched for: it is ASSUMED to be out there! It is assumed in almost every discipline that emergent properties that encapsulate much of a structure that initially seems random and chaotic, MUST exist.

To go one step deeper, and to get slightly philosophical, a more fundamental assumption that is even more rarely questioned is that such a search is a good thing. That gaining more explanatory power, connecting increasingly disparate topics, has purpose, has value. While I won't go into this much in this essay, there is much to discuss, and much that has been discussed, on this topic.

To really drill this point home, let us think through some of the consequences of this unavoidable, habitual use of schemas. First, consider the development of new relationships. You understand people that you meet for the first time by projecting relationships with past people onto them. Anecdotally, this process is often referred to by people as categorisation, or 'finding where they stand' relative to certain axes. Do you remember trying to gauge how 'cool' someone was in high school? Perhaps you can more recently recall trying to gauge how successful someone has been at a high school reunion. Maybe you're religious, and you catch yourself trying to work out whether someone is more orthopraxic, or perhaps even more orthodox, than you. One of the places where this is best understood in our society, and still so seldom corrected for, is in political discourse. Such axes are often intensely important for us, thus making our projection that much more stifling and complete. Recall people being surprised at the kindness of a Republican, or the industriousness of a Liberal.

There is much to be said about the precise employment of schemas at this more micro level. What decides how much unjudged, unprojected, 'reality' actually comes through. The phenomenology of this projection or contortion. This I will leave for another essay. For a preview, please take a look at my essay on emotion and rationality.

The direction I will continue in for now is beginning to answer the question of what is LOST in this process? When do such schemas fail, what does the failure look like, and are there concrete limits of the reliability of these schemas?

Let us now draw on existing literature to answer this question as well. First, let us turn to the book Black Swan, by Nicholas Nassim Taleb. Although the book covers a vast interconnected network of topics, the central tenet is the very human inveteracy we have been discussing: mistaking the map for the territory. This is meta to schemas, perhaps defining a schema for your relationship TO schemas (pay attention to this meta-schema concept. We will touch on this in essay 3). When discussing human cognition, it is the use of schemas, and then the action of forgetting that you are using it. Thinking the schema IS the actual thing that the schema is an attempt to model. In some sense, some degree of this forgetfulness is a prerequisite for using schemas in the first place. You cannot consciously put faith in it while believing that it is completely fallible. Thus, humans are constantly mistaking the map for the territory, and are doomed to do so to a certain degree forever.

But, this wouldn't be so bad if the map of the territory was an accurate one. Perhaps we can get incredibly accurate maps of the world, and not need to question them too much? In Black Swan, Taleb posits in the book time and time again that such a map would be fundamentally impossible to construct. That the real world, once you leave the domian of Platonic, ideal forms, is so abstract and chaotic that the underlying structure is simply beyond comprehension. ANY model is a conflation, elision, or violation in some primordial sense.

David Hume, a great philosopher and pleasantly jocund man, who was famously absolutely serene on his deathbed, made the similar audacious claim that we cannot know ANY of the causal structure of the world. We simply observe pairs of concomitant events, arranged in a temporal sequence, and at some fundamental level, assume causality. While he was a diehard empiricist, stating that all knowledge was derived from the senses, he was adamant about the limitations of this knowledge.

Perhaps the issue with modelling something abstract and chaotic is this very issue of causality. After all, what is inference using a model? It is making estimates and implications, using some set of limited observations, for future, or otherwise unobserved or hidden, behaviour. These implications can only ever repose on causality! Otherwise, how could you possibly draw conclusions about things not observed, from things that are observed?

The book 'Thinking in Systems', by Donella Meadows, a famous admonisher of the dangers of exponential growth (someone, perhaps, who will soon prove to be a Cassandra), explicates some of the reasons behind the deep-seated dangers in this habit of drawing maps out of territories. One reason is the presence and dependence of behaviour on feedback loops. In her book, she draws diagrams of 'stocks and flows' of resources, where stocks are accumulation points for resources, and flows are the causal structure (let's be careful here!) between these stocks of resources. The strength of a flow between stocks is some function over the size of the stocks it flows to and from.

Now, let us consider a simple example: the stock representing the population level, the number of humans on Earth. Now, let us add a simple and realistic flow: from the stock to itself. This flow can be defined by a simple linear function: the more people, the greater the rate of increase. This is biologically sound: more people, more babies. From this simple model, you would expect exponential growth of population: more people, more babies to make more people, even more people to make even more babies, and so on infinitely! Let us call this first map M1.

However, now let us add a second stock: natural resources. This general stock represents everything that humans rely on to sustain their existence: food, materials, land, oxygen, etc.. Let us also add a simple, linear flow from the humans to the resources: the more humans, the less resources. Moreover, let us add a second, negative flow from resources to humans: fewer resources, fewer humans. However, this flow is not linear: it has a strength of zero until there is a sufficiently constrained number of resources, at which point it dramatically curves upwards. Let us call this newer, updated map, M2.

Let us now play out the simulation that this diagram represents, starting at a very simple baseline of a small number of humans and a large number of resources, and play it out over time (temporal inference). At first, the number of humans rises. Slow at first, then dramatically upwards, exponentially growing. But, at some, unexpected, perhaps arbitrary point, the growth stops: the resources become sufficiently depleted, and the level of humans they are capable of sustaining drops dramatically.

To someone observing the behaviour from M1, expecting simple exponential growth, they would be bewildered at this outcome. What happened wasn't simply an unexpected simple decrease in population, or a deceleration in the increase; the behaviour changed on a much more fundamental level. Donella's book calls this a change in the dominant feedback loop. In this context, a feedback loop just refers to a cycle between stocks.

M2 has two feedback loops: one positive, from the population stock to itself, another negative, from the population stock to the resource stock, and finally back to the population stock. In a system at any one time, it is very likely that there is a small number of dominant feedback loops, or feedback loops that control the primary behaviour of the system (by behaviour, I just mean flow of resources). At first, the dominant feedback loop was the first flow, from the population back to itself. Thus, to someone doing inference over time using M1, the observed behaviour made perfect sense, and could be justified. However, at some, completely arbitrary point from the perspective of M1, the dominant feedback loop changed: the behaviour of the system was suddenly and dramatically different. And, of course, at some future time, M2 will also fail to explain the observed behaviour, for it too is a map, and not the whole territory.

The importance of these feedback loops is fundamentally one of nonlinearity. A feedback loop that is linear can be accounted for, is highly predictable, and for feedback loops or resources with small effect sizes, mostly negligible. However, nonlinear feedback loops are the opposite: future behaviour can rarely be accurately inferred from present behaviour. Stocks that initially seem negligible, or small, can suddenly be of utmost importance.

This concept of nonlinearity is fundamental to the ideas presented in Black Swan. Taleb rips into the fundamental bias of both our cognitive and statistical tools towards linearity, such as the cognitive bias of scope insensitivity (how hard it is for us to fathom the difference between 1,000 and 1,000,000), and the statistical Gaussian distribution (built on the assumption that proportional aggregation is additive). From this lens, assuming linearity on a societal scale leaves us especially vulnerable to changes in the dominance of feedback loops, making the maps that we draw, as Taleb put it, especially 'fragile', meaning hurt by change.

Now, let us finally loop this back round to the mind: our mental frameworks ARE maps of the causal structure of reality. Friston's free energy minimisation framework posits that we are constantly interpreting our experiences through the lens of these frameworks. Taleb warns us of the linearity assumptions baked into our cognition (which for various reasons, have been very good until the modern age). Meadows warns us of the dangers of nonlinearity when creating frameworks. Are you beginning to see the problem?

This marks the end of essay 1, but there is much more to discuss, and much more to connect!