How do we learn? Just read this new pre-print from the great Slava...

How do we learn? Just read this new pre-print from the great Slava Kalyuga which provides a slightly updated model of how we acquire knowledge through the conduit of an "explicit intention to learn". 🧵⬇️

As you'd expect from Kalyuga, he argues that instructional strategies should incorporate CLT (cognitive load theory) to support biologically secondary learning (stuff kids learn in school), emphasizing worked examples, guided discovery, gradual reduction of scaffolds etc. but what's interesting is the idea that the "explicit intention to learn" is the driver behind humanity's cultural evolution and that intelligence is an emergent property of this.

He outlines 6 key principles of human cognitive architecture which are oldies and goldies (except for the last one):

1. Information Store Principle: Emphasizes long-term memory as a repository of structured knowledge patterns, distinguishing between biologically primary (intuitive) and secondary (explicit, learned) information.
Primary information is innate or rapidly acquired (e.g., language), while secondary requires effortful, intentional learning (e.g., reading, scientific knowledge).

2. Borrowing and Reorganizing Principle: Highlights that knowledge is largely acquired from external sources, such as social interactions or written materials, and reorganized to fit existing cognitive structures.
This principle underpins knowledge transfer and cultural accumulation.

3. Randomness as Genesis Principle: Explains the role of random trial-and-error or problem-solving heuristics in acquiring new information. So in humans, it interacts with intentional learning to refine random processes into structured problem-solving strategies.

4. Narrow Limits of Change Principle: Suggests that working memory constraints limit the extent of modifications to knowledge, protecting existing structures from disruption. This constraint ensures stability while allowing incremental learning.

5. Environmental Organizing and Linking Principle*: Discusses how familiar environments relax cognitive constraints, enabling more efficient information processing through chunking and automation.
*This is a lesser-known aspect of CLT but an important one because it highlights how context and prior knowledge can drastically influence cognitive efficiency and learning outcomes. (it also explains why Liverpool are better at Anfield than away)

6. Explicit Intention to Learn Principle: This is new and represents the conduit theough which humans as intelligent systems are capable of conscious, effortful learning. This intention drives the shift from implicit to explicit processes, enabling cultural and technological evolution.

For me, the borrowing and reorganizing principle was always one of the strongest and most persuasive aspects of CLT just because of how we learn - The unique thing about humans is that we are able to share VAST amounts of information with each other and we do that through explicit instruction, modelling, stories, analogies, persuasion etc.

The idea is that humans acquire most of their knowledge by borrowing information from external sources—such as other individuals, written materials, or visual demonstrations—and then reorganizing it to fit within their existing knowledge structures. If we accept this, then explicit instruction (or 'teaching' as we used to call it) is the rocket fuel that has propelled humanity to where it is today.

I mean, if it were true that we learn best by discovering things for ourselves, we would still be travelling around in carts and worshipping goats.

But it doesn't just apply to reading/maths etc, this is evident in practices like apprenticeship, where modelling and guided practice allow novices to rapidly acquire domain-specific knowledge. I love seeing videos of people learning to make complex things through knowledge that's taken centuries to attain.

There's not an awful lot of stuff on motivation in the CLT literature (apart from Ansrew Martin's work) but this incorporates motivation with Geary's theory on biological knowledge.

Biologically primary knowledge (basic language, social skills etc) is acquired rapidly and intuitively because humans are evolutionarily predisposed to learn these skills in their natural environment. Motivation is intrinsic and automatic here; humans naturally seek these abilities through play, socialization, and exploration.

HOWEVER...

Biologically secondary knowledge is not inherently motivating. Skills like algebra, reading, or scientific reasoning require explicit, controlled, and effortful learning, supported by working memory and long-term cognitive processes.

So...engaging in biologically secondary tasks (reading/writing/Maths etc.) demands sustained cognitive effort, which is not automatically driven by evolution. For humans to overcome the natural resistance to expending effort, explicit motivation must be present.

In other words, we don't really want to learn hard stuff. Motivation, therefore, is not just a psychological factor but a sort of cognitive enabler for secondary learning, making it critical for instructional design and pedagogy.

Lastly there is an interesting thought on the emergence of AI in relation to CLT, in the sense that AI systems are designed to borrow and reorganize information in ways that mirror human cognitive processes. For instance:Machine learning models train on vast datasets (borrowing knowledge) and refine their internal representations to improve performance (reorganizing knowledge).

BUT unlike humans, AI can process immense amounts of information without cognitive limitations like working memory capacity, enabling rapid reorganization and decision-making. Kalyuga raises questions about how humans can remain competitive in the technological phase of cultural evolution.

Also an over-reliance on AI for knowledge acquisition could diminish the need for humans to engage in explicit, effortful learning, potentially weakening their cognitive skills over time. In other words, we may well get a lot more stupid.

However AI might contribute to advancing CLT by analyzing vast datasets on learner behaviour, cognitive load levels, and instructional effectiveness, enabling more personalized and efficient learning designs. We'll see... 🧵

You can read it here: preprints.org/manuscript/202…