Note: As in many of my recent posts, these ideas are new to me, and I'm just starting to test them out. This is part of a larger project of overhauling and refining my approach to scholarship. I'm excited to hear feedback and criticism. I anticipate this project will take years, and I don't know where it's heading next. Thank you for indulging me...

Interleaving Deliberate Practice

One criticism of using automated flashcard systems, like Anki, for conceptual topics, like math or physics, is that they don't offer you an easy way to practice solving problems. In theory, you might be able to use some flashcard platforms to automatically generate new problems to solve. However, this seems like a complicated, time-consuming, and ongoing challenge.

An alternative idea is to create flashcards that tell you what exercises to solve from a textbook. Such a flashcard might read:

"Solve 5 problems from Chapter 1.1 pg. 10-13 of Zill's First Course in Differential Equations textbook."

Assume that the spacing of flashcard reviews starts at about 4 days apart, and increases by about 2.5x each time. If so, you'd see that flashcard about 4 times in the first two months, 6 times in the first year, and 9 times over 15 years.

Imagine you wanted to spread your practice out over 15 years at increasing intervals, for a total of 9 reviews. Also, imagine there were 60 problems in Chapter 1.1. In that case, you'd want to assign yourself 60 ÷ 9 ≅ 7 problems every time that flashcard came up.

Since you'd most likely not want to do all the easy problems first and leave the challenge problems for 15 years later, you might want to refine this by requiring that you work on every 9th problem. Hence, on your first review, you'd do problems 1, 10, 19, 28, 37, 46, and 55 from Chapter 1.1. On the second review, you'd do problems 2, 11, 20, 29, 38, 47, and 56.

As you go, you'd need to keep a record of which problems you'd solved. You could do this in several ways:

1. Keep track in the flashcard itself, by editing it with the next set of "to-do" problems so that you'll know just what to do when it flashes up.
2. Mark off the problems in a physical book, or use a highlight annotation in an e-book.
3. Keep a physical document, such as a notebook, where you work the exercises for a specific subject so that you can look up what you did last time you worked on Chapter 1.1.
4. Keep a separate document, like a text file, that notates just what you have to work on next.

Of all these options, I like #1 the best. This whole strategy presumes you'll keep track of your flashcards long-term, and this way you don't have to worry about losing any other document. It's built into your workflow.

You could also use this for non-textbook-based forms of learning. You might have lots of physical tasks you wanted to learn about. Look up a website that gives lists of applied projects, and factor the projects that look interesting into flashcards:

Microscope projects

Bioinformatics and programming

Computer programming projects

Linux system administrator projects

Graded piano pieces sorted by difficulty

Photography projects

Core strengthening exercises

Learning what you could learn

Problems With This Approach

There's often a premium on getting projects done quickly. It costs time to switch between tasks, and there's a reward for finishing. Certainly, this method is not appropriate for work-related tasks.

Flashcards may not be the best way to randomize projects. For example, you might want to interleave 50 different grade 1 piano pieces for your first year of learning the piano, but then abandon them when you move up to grade 2. To do that, you might want to have a special deck for "piano" where you insert one flashcard for every passage of a piece you've learned. Perhaps you have one flashcard for every 8-16 measures, for example.

If you practice one hour per day, you'd divide that hour by the number of flashcards you have to do, to allocate the amount of time per flashcard. Then you'd select "again/hard/good/easy" depending on how that passage was sounding at the end of those few minutes of practice.

More challenging is when a task doesn't lend itself to factoring into flashcards very easily. For example, solving a bioinformatics problem on Rosalind.info often requires a lot of programming and creative mathematical thinking. They also take long enough in many cases that you wouldn't want to solve the entire project in a single day. Yet you finish an individual project quickly enough that you wouldn't want to be returning to it 15 years later.

It might therefore be better to use a simple randomizer to pick a project category. Imagine you wanted to learn 50 diverse skills over several years. Each project used to practice those skills requires at least several hours, if not several days. Some of them "stack."

In that case, you might want to put all your possibilities on the rows of a spreadsheet. Then pick a row using a random number generator to select the next skill. On that row, you'd include a link to a list of projects for that skill. Then use the random number generator again to pick a project that's at your skill level.