This piece was inspired by Piotr Wozniak's jigsaw puzzle metaphor, and his writing about emergent conceptualisation in learning.
Suppose you're building a jigsaw puzzle, but you don't have the box lid to see what the result will be like.
You shuffle and search through random pieces and eventually find some that click. If they're blue, you might guess those pieces represent the sky. Then you find some other matching pieces that some metallic structure. Each time two pieces click you experience a small win, giving you energy to continue.
Over time, as more and more pieces interlock, you begin to get a hunch about the bigger picture that you're building. Eventually - perhaps with only a quarter of the puzzle complete - you realise: It's the Eiffel tower!
From then on, the nature of the job qualitatively changes. It's easier to sort the unmatched pieces into piles representing trees, people, arches, and so on. It can also become easier to guess which pieces will fit together, and what they might represent. This process, of working out the bigger picture by making small links, is what constitutes the development of deep knowledge.
Deep knowledge need not be as broad as understanding a whole field (e.g. "physics"). It is simply the insight that appears when small clues reveal a larger structure.
During waking hours, reams of detailed information enter the brain in an incessant stream. By putting many small pieces together, we understand our environment. This can be as simple as recognising a coffee cup from its colour, weight, size and so on. Thus, larger concepts emerge from the details.
Once we recognise multiple concepts, it becomes possible to work with them at a higher level of abstraction than pure sensation. It also becomes possible for larger concepts to emerge from smaller ones, leading to layers of understanding.
For example, once we subconsciously notice a coffee cup and the desk it sits on, combined with the time of day and other details, we effortlessly realise that we are standing in our home study. Starting from sensations and objects, we are able think in terms of places.
The process is effortless, but essential to understand if we want to help ourselves and others to learn effectively. It is the same process that happens as we learn algebra, or our native language, or as we interpret other people and where we fit into their lives.
In all these cases, patterns of sensory input get organised into basic concepts. Then those concepts can be related to each other to reveal even deeper concepts. Eventually, we may understand multiple fields and how they fit together. Artificial intelligence researchers try to replicate this natural human ability through so-called "deep learning" networks.
From the moment we are born, each of us has our own puzzle to solve, and that puzzle is the world around us. Because of our unique genes and circumstances, we each have a somewhat different puzzle; a different picture, with pieces cut in different ways.
Because deep knowledge must emerge from the parts, it can never be directly represented in sensory input, and so there is simply no way to transfer deep knowledge from one brain to another. When one person reaches an insight and tries to share it with another, their deep knowledge gets converted into sensory input for the receiver to interpret based on their existing knowledge.
To the student, the teacher and the school are just more puzzle pieces to make sense of! Because of this, school and teaching has no qualitatively superior status in the developmental process, compared to simply experiencing the world directly. In fact, it is generally better to experience anything directly than to discuss it from a distance, because the former provides many more details that can help the bigger picture emerge faster and more clearly.
In the end, everyone learns a different picture of the world, and the quality of their understanding depends on how well their knowledge is integrated to form a cohesive model.
Schools and universities use curriculums to dictate a specific order and linking of concepts for everyone to learn. This is incompatible with the processes through which deep learning actually occurs.
Instead, learners must be supported to form their own sets of concepts, and connect them in their own way. This process already happens within every learner's brain, but we obstruct their progress every time we attempt to standardise (see: Learning objectives stunt learning).
For learners to achieve high quality knowledge, they must decide on their own division and hierarchy of concepts. Only in this way can they build well-connected, high quality knowledge.
