<aside> 👋🏽 Overview: Spaced-repetition and other cognitive science ideas are akin to building blocks: powerful, but to compose a system they require a theory of architecture. Practically, this means (a) being explicit about your theory of architecture or (b) reproducing a tacit architecture.

Unfortunately, cognitive science ideas are usually applied without rethinking dominant architectures for learning. I examine the limits of Quantum Country as such a system and offer ideas for how constructivist 'architectures' might offer better frameworks for pulling together cognitive science 'building blocks.'


Draft March 9, 2021. Minor edits Sep 29, 2021.

As more complex things are made, architecture dominates materials... what lies behind what we touch and think about are often wonderful things that are mostly organizational in nature.

— Alan Kay


Cognitive science ideas — think spaced practice, elaboration, and far transfer — are powerful building blocks [1]. But blocks alone don't make a building. They need to live within an architectural system: a set of ideas for how different parts might fit together in pursuit of certain goals.

Montessori and Reggio are examples of architectural systems [2]. They have (a) clear goals (b) strategies, practices, and technologies for pursuing those goals and (c) a theory for how the different pieces fit together. Components (e.g. practices or technologies) are designed with respect to their purpose within the broader system. Traditional education models (think teaching machines or a traditional school) offer a more common architecture.

Unfortunately, most learning projects focus solely on architecture (e.g. social constructivists who gag when someone mentions spaced-repetition) or on cognitive science ideas (e.g. those focusing on spaced-repetition systems without questioning the broader systems in which spaced-repetition is used). Sanjay Sarma, MIT's learning czar, examines this divide in Grasp [3].

In this post, I'll examine the limitations of only focusing on cognitive science ideas. I'll then argue how constructivist 'architecture' can advance cognitive science ideas.

The limits of a cognitive science approach: Quantum Country and spaced-repetition systems

There has been a resurgence of systems that help people learn through spaced-repetition: think flash cards, but with automated intervals for revisiting cards based on how you perform when you encounter them. Quantum Country is one of the more compelling examples.

Developed by Michael Nielsen and Andy Matuschak, the project introduces readers to quantum computing. It is effectively a digital book with two added wrinkles: (a) questions appear in-text at varying intervals, then disappear once you answer them so that you can continue reading and (b) review questions are emailed to readers at spaced intervals that vary in frequency depending on how you handle them, making it easier for a reader to practice spaced-repetition.

After reading some text, we encounter a few questions that ask us to recall what we just encountered.

After reading some text, we encounter a few questions that ask us to recall what we just encountered.

After answering the questions, the 'quiz' collapses. The text above then appears.

After answering the questions, the 'quiz' collapses. The text above then appears.

The interface is beautifully designed and content is both engaging and clear. However, Quantum Country and its siblings (Orbit, Anki) remain trapped within traditional learning architectures. [4] Why?

How you think about spaced-repetition is important. Practicing a particular surgery technique over spaced intervals is qualitatively different than recalling information based on spaced prompts. Surgery practice involves applying a skill in context.

Quantum Country's prompts focus on remembering content outside of any constructive context. While helpful in sparking one to remember, this doesn't necessarily encourage one to elaborate mental models, develop deeper relationships with ideas, or practice diverse epistemological styles.

This points to a deeper challenge in Quantum Country: Andy and Michael optimize for spaced-repetition within the form of a linear book that remains largely grounded in traditional ways of thinking about learning and knowledge: learning is about acquiring information, which can be broken down into atomic chunks, and assimilated by an individual. Spaced-repetition simply becomes another tool within this conventional architecture.

But if you believe that knowledge is constructed, this leads you down a different design path that requires reconsidering architecture.

Instead of first designing for spaced-repetition of concepts, you might design a tool that supports a person in reconstructing and tinkering with ideas, visualizations, and concepts — in building relationships with ideas and connecting it to their interests. You might also place a greater emphasis on connecting with a learner's interests and creating opportunities for playful engagement with ideas. What Papert might call 'hard fun.'

This might entail creating a microworld (what a Montessorian might call a prepared environment) for exploring quantum computing ideas and then using spaced-repetition prompts to encourage construction related to ideas one had not explored in a while. This latter project is much more difficult (building worlds where a person creates is much harder than building artifacts where a person largely consumes). But the payoff would be profound.

Some other constructivist ideas include [6]: