I’ve been thinking quite a bit lately about knowledge and its organization, distribution, and absorption, particularly in the context of education in the digital age. As I explore the multiverse of online resources for learning, I notice that although each has its own peculiarities, they also more often than not share a number of striking similarities (intentionally or not) in structure, pedagogical framework, and philosophy.
THE DIPOLE OF THE CURRENT MODEL
I observe, generally, a certain tension between what we could call “customization/flexibility” and “organization/guidance”, both of which have great uses but neither of which I think fully captures the potential to optimize learning.
On the one hand, sites geared towards “customization/flexibility” are user-reconfigurable and user-explorable to a high degree, allowing for a wide number of individuals with divergent interests and needs to find what they’re looking for. These usually contain a large amount of information, and tout the idea of infinite flexibility—users should be able to find things on demand as needed, with instant accessibility and low switching costs, and not be limited by predetermined constraints. In effect they lean pedagogically agnostic, rarely pretending to know what a user needs but instead preferring to let them forge their own path through the knowledge forest. Examples: Wikipedia, Google.
On the other hand, tools that take “organization/guidance” as their core principles take the beneficent position of structuring content and leading the user on (if all goes according to plan) a best-fit route through a given subject. This can be as specific as a design tutorial or as general as an introductory course on computer science, Spanish, or philosophy. There is a guiding idea of structure and coherence, of things fitting together well, with a clear, preset linear progression for learning and research. The main idea is that of preselecting, screening, filtering and approving content, and carries an implicit authority that requires trust on the part of the user. Examples: The prototypical MOOC or tutorial.
ON UNIFICATION
I think it’s imperative that everyone be able to find information as they need and best see fit for their given purposes, and be restricted by arbitrary or artificial limitations as little as possible. But I also recognize that structure is usually a good thing and can enable people to ensure that their path makes sense, that they will gain clear utility from their course of study, and that they will be able to take advantage of expertise laid out by those who have learned before.
What I would like to propose, in rough but (I hope) clear terms, is:
A modular system for growing, distributing, and finding one’s way through the building blocks of education, from the most elementary atomic particles (which we can define as single discrete ideas—however that may translate) to the corpora of entire academic fields. I’ll talk about this largely in the hypothetical sense, and I’m aware of the massive undertaking it would be to translate this idea into applicable software form, but I also want to present some less abstract ideas of what this might look like in simplified, alpha form.
A BEAUTIFUL STRUCTURE
Depth vs. breadth of learning is a problem I struggle with and something very relevant at present with exponentially increasing information overload. The root of this problem makes me think of the very interesting and beautiful structure that underlies our organization of knowledge:
We can think of the totality of the knowledge space as something fractally defined, in that within each topic you can split and focus more deeply/specifically on subtopics to an almost infinitely recursive level (until reaching the limits of human knowledge—at which point we can further push against the boundaries of an obscure niche and create deeper layers of the fractal).
This fractal metaphor isn’t literal, as repetition doesn’t occur with identical components and isn’t neverending, but I think it’s an interesting way of thinking about it structurally, as sort of an irregular but somewhat predictable/plottable branching tree hierarchy.
This system I’m interested in—that I think we’ll need more and more as information and knowledge continues to proliferate ever more rapidly—will be a tool to help wrangle, define, shape and explore this fractal. I see it as almost quantum mechanical, in that you not only can never fully explore or absorb the content of the fractal, but when you zoom in far enough you actually can’t help but influence the shape itself. Each person will experience this fractal differently, often to an incredible degree. At the highest levels much experience will overlap (e.g. basic curricula of elementary school and high school) but the deeper you explore it, the more it starts to diverge, as the possibility space expands rapidly beyond the ability of any individual to take in. Another benefit of such a tool, then, might be to help teams of people—be it corporations or academic research groups—assemble complementary skill sets into a collective body with the knowledge-potential most likely to fruitfully tackle a given problem.
THE HYPOTHETICAL SYSTEM
In creating this modular system, I think techniques for optimization and probabilistic modeling tools will be quite helpful. Although the specifics of that these techniques might look like is beyond my present understanding, I envision something along the following lines:
Experts in a given domain or subject (well, an army of them—but we’ll start with an algorithmically-assisted handful) would map various collections of databases or repositories of as much available knowledge as is currently possible, taking a curatorial role and differentiating between the valuable and the useless or redundant. Though all this information would have to be searchable and accessible, and exponentially more filtered and structured than, say, the entirety of Google’s cache of the Internet, it certainly could be sourced from a large variety of places, from existing course content and written material to audio lectures and field recordings of experts, and more. A severely incomplete approximation at first, which would gradually refine as a function of human effort * technological advances.
Most importantly there would need to be a process of weighing content so as to strongly privilege that which is tailored to be useful as a means of instruction or tutelage (rather than unfiltered raw data or information-overabundance; we want this to be human-manageable!)—but at the same time it would be important to collect content at many possible levels so that within the knowledge corpus of “mathematics” both grade school and PhD students would be able to benefit from the resources therein.
A further step, given this hypothetical library of content, would be creating a system to sort, tag, label, prioritize, and otherwise give coherent structure, hierarchy, and categorization to the whole lot of it. Again, this would be tremendously complicated, but I see it as a necessary step that could be iterated towards over time.
Then, we would in theory (or perhaps not; for all I know this problem may be NP-complete…but let’s pretend) be able to plot out the most likely statistically useful paths for given learning objectives or outcomes (as enumerated and decided by some combination of teacher and student). Diversions and detours from these primary paths would correlate to variation in needs, constraints, learning styles etc. Thus with a set of inputs provided by a user—e.g. time, scope, goals, subjects—it could be predicted what methods and modules would optimize the learning process for a given person and provide best-fit path for unique situation.
TOWARDS A PRACTICAL TEST
I fully acknowledge that the process of creating such a system is likely to require advanced machine learning algorithms, complex techniques for data analysis and statistical modeling, a vast amount of human labor and more, to work at a high level and on a large scale. However, I think it possible to create a serviceable MVP of sorts, perhaps on the concierge model, plotting out a constrained version of these branching learning paths and developing an initial set of modules within a very limited scope—within a single sub-domain, even.
To start, we could look at pre-existing content, linking to it where useful and possible and perhaps conscripting original content producers to fill in the most obvious gaps. The plethora of existing online courses would be a fantastic place to start—it wouldn’t be too much trouble to choose a specific subject and try picking apart and organizing existing content into sets of smaller modules to get a better idea of what this sort of structure might look like.
For example, to attempt a rough modular model for learning basic CS/programming: we could have modules of various detail and difficulty level focusing on web design, web development, basic scripting, databases, theoretical concepts, mobile, GUIs, UX, systems architecture, etc. The recommended path through a subset of these disciplines would differ based on whether a student desired a high-level overview of the field, a specific skill set to prepare for a career switch, or a basic understanding of complementary skills for a related field.
To start, we could source a few courses in each of these broad topics and explore how the pieces fit together, consider to what various degrees they are or aren’t divisible, experiment with plotting out a variety of paths through different varieties of “modules” and more.
There are existing technological precedents for this sort of thing. In search algorithms like Google’s, of course, but more fundamentally in the logic of hyperlink, commonly considered one of the most centrally important structural elements of the Internet. In the hyperlink we have contextually determined flexibility; it allows both linearity and near-infinite combinatory path-construction on the part of the reader/learner. The visual structure of the Internet as a whole—and I’ve not seen anything that purports to map its entirety, but I have seen many partial visualizations—I suspect looks quite similar to how I envision this knowledge hierarchy. One main difference, of course, is that the Internet has taken shape in a process of chaotic self-assembly across a multitude of interests (technological and educational, but also for purposes ordained by government, industry and commerce, entertainment, creativity, communication and more). The total field of information or knowledge in an education-oriented sense is no doubt chaotic as well, but I like to imagine it with at least a mote more structural elegance.
IN CONCLUSION
Taken to the furthest levels of complexity, I suppose what I’m envisioning would require nothing short of a complete revolution in data analysis, machine learning, and a host of other computational problems. It may have intractable structural difficulties and continue to be a pipe dream for decades or centuries to come. I’m fine with that, so long as we continue to evolve in the right direction. There are thousands of people much smarter than me working on these hard problems, so I’m optimistic as far as that’s concerned.
But in simplest form, what I’m proposing is a re-imagining of curriculum design—something I don’t know much about but that seems potentially a very interesting discipline, especially if it brings together a wide range of experts, polymaths, thinkers, educators and technologists to define problems and explore the possibility of solutions. This is something we can and should be pursuing right now, and in basic form requires only creative thinking and resourceful experimentation to test further.