Fast Learning Loops

And how they can change K-12 education

I replaced my car battery last night. I’d never done it before for this particular make and model, and it had been 20ish years since my last time performing the procedure, so I had some knowledge gaps. So I turned to YouTube and opened the hood. Over the next 15 minutes, I ran into obstacles, experimented through trial and error, skipped back to the relevant part of the tutorial video, tried different tools, and got pretty close to knocking on my neighbor’s door (who actually knows how to work on cars). Thankfully, it all turned out fine in the end—no injuries and a car that starts.

I thought about learning loops. This is a concept discussed a lot in entrepreneurship (e.g. Eric Ries Lean Startup methodology), but a quick google search on learning loops reveals that there is no consensus view on how this applies in general education.

That’s great, because we can apply this powerful concept to basic academics, without the encumbrance of convention. Here we go!

You can think of a barebones learning loop as having two basic parts:

1. Input. New knowledge or information entering the brain of the learner.

2. Output. Application of the new information in the form of an assessment or a demonstrated skill.

Hopefully, the learning loop leads to personal growth, greater capacity, or some other increase in the life of the learner. Even though its uncomfortable to tackle something new, humans do it all the time. We’ll talk about this loop as it shows up in education.

I’ll sort this (incomplete) list by timescale, from the longest loops to the shortest.

The Degree

Timescale: Years

You can think about a high school, undergraduate, or graduate degree as a multi-year learning loop. The premise is that the learner acquires a bundle of knowledge and skills that will elevate their productivity and life satisfaction. The input of this loops is thousands of hours sitting in class and fulfilling the requirements of the degree. The output is not visible until long after the degree is received. Does the learner get a good job? Are they able to function in the world? Do they feel like their life has meaning and purpose?

We of course hope the answers to these questions are yes. But a student debt crisis, wage stagnation, declining standard of living, and suicide and opioid epidemics are enough to raise doubts.

Signing up for a degree program takes faith because there is a very long time until you see the results. You’ve forgone many opportunities to do other things. If you don’t like the results, it’s a real problem.

The Course

Timescale: Weeks/Months

Most degrees require a student to complete courses. These require weeks or months of attending lectures, completing assignments, and passing tests. At the end, the learner has a letter grade and some credit hours as approximations for their understanding of the subject material and ability to demonstrate that knowledge.

Courses as learning loops introduce new challenges. What does the letter grade really mean? Professors and teachers have a lot of leeway; some look at course grades as a measure of mastery, while others calculate grades based on attending class, turning in homework assignments, having a good attitude, or pleading your case with conviction.

Even if the letter grade for a course was a consistent indicator of knowledge demonstrated on a test, will the student be able to apply their new knowledge for success in other courses, or in life? It’s still a long time before you find out whether the course was a wise use of the weeks or months you spent.

The Unit

Timescale: Weeks

I’ll move quickly through this one because the logic mirrors the discussion about courses. In time-based education, educators break up the material of a course into units that typically last 2-4 weeks. Students end up with some points toward their final grade, ideally representative of the knowledge and skills they acquired during the unit.

The Lesson

Timescale: Minutes/Hours

In contrast to the long timescales we’ve talked about so far, a lesson typically happens within a single day. It has the potential for learners to take multiple runs through the loop while the context stays fresh in the short-term storage (although that’s hard in a “stage on the stage” classroom). Lessons have a short enough timescale to allow learning to be faster and go deeper. Let’s take a look at learning loops for a few common academic subjects.

Writing

From forming letters on a page to constructing sentences to mastering grammar and punctuation rules to crafting narratives and essays, learning to write is a lifelong quest. Depending on who you ask, writing skills are either obsolete or more important than ever in the world of AI.

Writing lessons take a long time. In the first place, it’s hard to break up writing skills into bite-sized pieces. So instead of focusing effort on, for example, using commas in a list, students end up with a page full of red ink flagging every spelling error and every awkward participle. It can be overwhelming.

Another challenge with the timing of writing lessons is the need for an expert reviewer to provide feedback. If I turn in an essay on Wednesday, I might not get it back until the next week, long after I’ve forgotten the context and moved on psychologically. My teacher, meanwhile, is banging his head against the table as he marks up page after page. By the time the learning loop is closed (i.e. performance assessed and returned to the student), it’s unfocused and too late to be actionable. No wonder so many kids (and teachers?) are ready to delegate all writing to robots.

Social Studies

A lesson in history or current events can go one of two directions. The old way is to generate lists of facts, ask kids to memorize them, and then check for memorization on a multiple-choice test. If you have a scantron machine, or the modern equivalent, you could close the learning loop within the same hour of delivering the content. Teachers who want to encourage deeper thinking and drive higher engagement have learned that social studies is best explored through a dialogue, with open-ended questions. That’s better, but it requires essay questions and projects, which brings us back to the long timescale of writing lessons.

Neither of these is ideal, but one possible improvement is a Socratic discussion or other inquiry -based approach. In this mode, students can access the lesson content through many formats (e.g. reading an article, watching a video, attending a lecture, researching across multiple sources) and then come prepared to engage with their peers in an open-ended dialogue. The benefits of this approach are a higher level of engagement, higher efficiency of information transfer, and a faster feedback loop (i.e. evaluation from student participation in the discussion delivered immediately). The downside is logistical—it would take a tireless expert to facilitate and evaluate 25+ kids every day! Microschools and technology can help here, but that’s a post for another day.

Math

When I was in school, the learning loop for math was: listen to the teacher explain a topic, do a homework assignment (1-79 odds only), turn in the assignment, wait for the teacher to grade and return it, then (theoretically) take a look at missed problems.

It’s a good learning loop, but it takes too long. Interest wanes, the next lesson shows up, and the result is critical holes in math mastery, compounding over time and leading to a generation of “not math” people.

Computers have allowed learners to navigate this loop faster and drastically increase the number of useful reps. Take Khan Academy as an example. A student clicks on a lesson. They can watch a video or read about the topic (input), then answer questions about it (output). If they get all the questions correct, they have mastered the topic and move on. If they miss questions, they can see a detailed tutorial about that topic or chat about it with an AI tutor. And in a microschool setting, the student can ask a peer or an adult learning guide Using these technologies, the learning loop for math is much shorter. That means more iterations, which means faster and more efficient learning.

Reading

The learning loop in literacy may be the fastest of all. See some letters on a page, recognize the phonemes, blend them together and try to recognize a word. You know it worked if you can string together a few words and it makes sense. Simple, right?

Not exactly. For one thing, learning to read requires a starting point of knowing letters and letter combinations, memorizing their sounds. And English is tricky because these relationships are not one to one. The letter a makes at least nine sounds, and there are at least five ways to spell the “er” sound. Learning all this requires appropriate content (which was scarce for a long time thanks to the reading wars). It also takes a lot of practice.

Even with a working knowledge of letters and sounds, the process of blending and decoding requires an enormous cognitive effort. It’s a fast loop, but it’s a challenging one.

Which takes us to the complementary idea of human motivation. Fast learning loops can unlock incredible knowledge and skills, but only if the brain engages. As we all know, there are a million reasons not to push ourselves into uncomfortable learning.

We need faster learning loops in order to unlock the benefits of modern education for individuals and societies. We also need environments that encourage humans to care about learning, to lean into the hard parts, and to stick with it. More about that in a future post.