Partnering with Anki

Less experienced and less successful spaced-repetition users tend to treat their SRS as a magic box: put information in, and knowledge comes out. In one way, this is absolutely true: SRS fulfills this role much better than any other learning method can. But in another, it is an utterly wrong role to be filling in the first place: the magic-box approach leads to worse retention and more frustrating study and leaves many of the benefits of Anki on the table. For the best results, rather than treating Anki as a magic box, we want to treat it as a partner. And we need to build some skills and take some actions to hold up our end of the partnership.

This shift in attitude can quickly take your Anki skills, and the benefit you get out of it, to a whole new level, but from hearing about and trying out Anki, it’s not at all obvious that you should have this attitude, and it’s not widely discussed either. In the first two sections of this post, we’ll examine why Anki doesn’t solve the problem of learning by itself (and why people think it does). In the remaining sections, we’ll discuss what you need to do to complete the picture and use spaced repetition to actively and iteratively build your knowledge, rather than passively accepting piles of disconnected facts. And in so doing, we’ll put the final nail in the coffin of the notion that Anki isn’t valuable because rote learning isn’t valuable. Many spaced-repetition detractors, and a large number of erstwhile or even current Anki users, have never understood that, even if rote learning really is useless, Anki can be used for building a deep understanding of a topic – not just remembering facts.

The problem

Tools are not sovereign

Let’s take a step back. You may not be immediately convinced that spaced repetition shouldn’t be treated like a magic box. Even if you are, understanding this incomplete point of view will make it easier to evolve your Anki skills beyond it. So why isn’t spaced repetition at its best when it’s a magic box?

I ultimately trace the expectation that Anki should be a magic box to the culture of people who typically use Anki (programmers, scientists, optimizers, analyzers, data-driven people, life-hackers, extensive learners), and quite possibly today’s industrial civilization in general. People who are part of this culture often exhibit an odd mindset which we might term sovereignty of tools. Sovereign tools would be those that bring about useful outcomes all by themselves. I say “would be,” because it’s not at all obvious that such tools exist at all, much less that they’re common. Robots could potentially fall into the sovereign-tools category, but even the most advanced robots available today still need someone to purchase them, set them up, and give them instructions before they accomplish anything, and then they still need to be supplied with electricity and occasionally repaired. We could think further into the possible future and imagine a benevolent artificial general intelligence in humanoid form that runs around the world unbidden solving problems, but now that it has its own agency, it’s not clear that we can call it a tool at all.

Ordinary tools – non-sovereign ones – don’t do anything at all by themselves. They only do something useful when you use them, and the effects are primarily because of your actions, not because of the tool. Frequently, though, people look around and see practices that others have found success with, attribute the success to the tool most obviously associated with the practice, and then get frustrated or start badmouthing the practice because buying the tool and changing nothing else didn’t work well. The tool might be necessary to participate in the new practice, but it is not sufficient to get the benefits of that practice; to see the benefits, you have to study and understand that practice and figure out how you can integrate it into your own life or business process so it gets used. If you get a new tool and you don’t have a new or updated practice to go with it, buying and learning to use the tool was a waste of time and money.

Certain tools seem more prone to elicit this fallacy than others. Nobody, for instance, would buy a hammer at Home Depot, bring it home, put it in their toolbox, go make dinner, and then check in the living room before bed and complain that their pictures weren’t hung up yet. The more complicated the tool gets, the less obvious its internal workings, and the less connected to one’s current reality it is – in short, the less understandable it is to the new user’s mind – the more likely they are to fall victim. That probably makes sense; with something like spaced repetition, the gains (“remember things ten times better with less effort!”) are accessible and clearly stated, while the inner workings of the tool are hidden from view and its algorithm relies on properties of human memory which even experts have a limited understanding of.

A certain deceptive simplicity is also to be found in many tools, including Anki. When I introduce Anki to others, whether one-on-one or in a lecture setting, I hold off on explaining that it’s a flashcard app until I’ve explained spaced repetition in some detail, because I’ve found that people’s brains often shut off and they miss the power of the scheduling algorithm once they box it into the “flashcard app” category, something they’ve likely seen many times before. There are only 47 menu items and subitems in all of Anki; in comparison, just the File menu of LibreOffice has 47 items, and most people find word processors simple enough to be usable, even if they don’t use anywhere close to all the options. Yet I don’t think there are many more generally applicable skills I could gain in using a word processor, while I’m sure I’ll continue getting better at using Anki for the rest of my life.

In short, the complexity of Anki isn’t visible in its description or on screen; it comes in figuring out how to make your memory dance with the algorithm. It’s all too easy to pass over this step entirely, in which case you end up standing awkwardly on the dance floor while Anki pulls you around, which, needless to say, isn’t a recipe for effective learning.

The allure of technical details

Another common mistake made when working with tools is to confuse understanding of the tools with understanding of how to use the tools. Again, this becomes more troublesome the more complicated and less obvious your tools get. For a hammer, those types of understanding are more or less the same thing. When you get into something like Anki, however, people often get sidetracked trying to use all the features of Anki to their fullest. They’ll spend hours running through all of the community add-ons looking for ones that could conceivably make them more efficient, or building baroque note types with two dozen fields and page-long card templates to automatically generate a bunch of cards from one entry, or experimenting with all of the scheduling options. Meanwhile, the same people might be dumping imprecise cards into their collection, studying leeches over and over again without stopping to work out why they’re hard, and even learning material they don’t understand or don’t care about.

Understanding Anki itself consists of knowing what the menu options do, or having strong opinions on what options and add-ons to use. Understanding how to use Anki to improve your life consists of understanding what your goals are, how to choose material that fulfills those goals, how you and your memory work, how learning works, how information can be divided into units for study with spaced repetition, how to recognize when you’re struggling with something and take a different tack, how to know when to throw in the towel on studying something, and only at long last how to manipulate the controls of Anki to get it to store and schedule your material. Merely understanding Anki won’t get you much more than a smug feeling and brownie points on the user forums, while understanding how to use Anki will help you learn better.

Don’t get me wrong. I find a number of add-ons quite useful; for that matter, I’ve written several of my own that I’m quite partial to. Custom note types and card templates offer many new possibilities for improved formatting and consistency and reduced effort (though they aren’t without their flaws). But I guarantee that someone who has no add-ons and nothing more than the Basic note type but who understands perfectly how Anki should be used will get more value out of Anki than someone who has spent tens of hours optimizing their knowledge of the technical aspects of Anki but hasn’t the first idea of how Anki should be effectively used. The tools aren’t useful until you integrate them into your process, which means understanding how they should be used, not just how to use them.

The lesson here is simple: it’s worth learning about advanced Anki features if you get serious about Anki, because they will absolutely make your life easier, but you shouldn’t get caught up in them at the beginning, nor should you equate more advanced Anki features with more effective learning. The basic technical skills I demonstrate in Getting Started with Anki and the few recipes and procedures peppered throughout the other articles will get you a long way. If you need something specific like mathematical notation, audio recordings, or the ability to import data in a custom format, you can go look up how to do that thing. Otherwise, your Anki-learning process should begin with learning Anki-related skills and figuring out how Anki integrates into your life, rather than learning the nitty-gritty details of Anki itself.

The solution

Understand Anki’s role

If effective Anki use is a partnership between you and Anki, what is Anki’s responsibility and what is your responsibility? Let’s start with Anki’s side, because that turns out to be minimal and easily stated.

• Anki faithfully remembers material. You won’t be able to memorize something effectively unless you have a known-correct copy of it in front of you.
• Anki reliably presents material to you at scheduled times. Whether you choose the time or Anki calculates it, it knows when you want to see it again and shows it to you then.
• Anki effectively guesses when to show you reviewed material again, based on easy-to-provide feedback, by combining your difficulty rating and its notes on your past performance to predict your future performance.
• Anki seamlessly transfers data so you can study in multiple locations, share your material with others, and import material others have shared.

That’s basically it! All of Anki’s features can be traced back to these four roles. For instance, all note creation, formatting, templating, and marking functions serve only to help you store and retrieve the information on your cards and make edits to it later. All of the scheduling and rescheduling commands and options serve only to allow Anki to reliably present material in the future. And so on.

Equally important here is what Anki doesn’t do. In particular, Anki doesn’t help you do any of the following things:

• Decide what you should be learning.
• Determine how you should represent that information in your mind and on your flashcards.
• Understand why you’re finding difficult material difficult.

These things are all up to you. You can either choose to take on that responsibility or to not do these things at all, which will certainly make your studies less effective. (For the most part, not only does Anki not try to do these things, but being a computer program, it can’t; how is a computer supposed to determine for you what you want to learn, or your subjective reasons for forgetting something?)

Do not learn if you do not understand, redux

Let’s look at an important implication of two of the items I said were your responsibility in the previous section: deciding what you should be learning and deciding how it should be represented.

In the title of this section, I quote again from SuperMemo’s 20 rules (I’ve previously mentioned this particular rule in the posts on precision and leeches). I think there are several layers to the exhortation not to learn without understanding. It’s easy to get stuck at the surface layer and assume you’re doing great at this – of course you would never be so silly as to learn something you didn’t understand! Violating the surface layer of the rule would work like this: You tell me that “argle kerflooey blarg is 63”. I take this statement, put it into my SRS, and learn this “fact”. Naturally, this will be a more challenging task than learning something that makes sense to me, and in the end it’s a big waste of time since I don’t know what it means, assuming it means anything useful to me at all – ergo, I’ve violated the rule. Fine.

We ought, however, to recognize that understanding comes in several stages and on several levels. Someone once pointed out to me the literal, spatial meaning of the English word understand – to stand beneath, as if you are in among the thing you understand. To choose a word that hasn’t been spoiled by shallow everyday usage, we could say the deepest level of understanding is grokking something. Grok is a word that hasn’t quite made it out of the hacker–Silicon-Valley arena into public discourse yet, and it deserves to; it means exactly this, to know something so thoroughly that it changes your identity. For instance, I could say that I can use the source control system Git, which would mean that I can copy and paste Git commands from Google; I could say that I know Git, which would mean that most of the time I know what commands to use and I only have to visit Google for unusual circumstances; or I could say that I grok Git – a true statement about me – which means that I can dash off weird and unconventional operations that take advantage of esoteric details of Git to achieve my goal without thinking twice about it, and I struggle to follow the thought process of people who don’t completely understand Git’s data model, because Git’s model has become part of my worldview. Indeed, someone who groks Git has a markedly different mental model of not just Git but the whole process of developing software, or creating anything that goes through multiple versions, than someone who has never encountered it does. Even as they get better at using Git, Git has changed them so they can use Git more effectively.

We can put understanding on a continuum from “I am aware that argle kerflooey blarg is 63,” to “I comprehend the literal meaning of this statement,” to “I have some knowledge of the effects of this statement,” all the way through to “I grok the whole idea surrounding this statement.” Learning a topic is the process of moving your knowledge of that topic from the meaningless side of the continuum to the grokking side.

What does all this have to do with Anki? In my experience, it’s way too easy to turn learning with Anki into creating a heap of cards that target the “I comprehend the literal meaning of this statement” level of understanding, and then leave it at that. This is unfortunate, at least for those of us who want to do more with our knowledge than be a walking encyclopedia, because it means your SRS won’t capture many of the most fundamental aspects of the topic. You might understand the individual cards, but you’re studying the topic as a whole without understanding it; you don’t understand how the items relate to each other or why each item is important. It’s not only more useful but also more fun and engaging to study content that mixes basic facts with questions that ask you to apply your knowledge, demonstrate a deeper level of understanding, and work the knowledge into your personal identity. These kinds of questions create much denser webs of knowledge, and those connections between pieces of knowledge are what produce…well, knowledge, or memory. Memory is about getting a certain thing to come to mind when you notice a particular pattern or are in a particular situation or think about a particular other thing. A pile of disconnected facts produces a disconnected and anemic memory.

At the same time, even once you realize the value of creating cards that connect your knowledge, when you’re starting out with a new and complex topic, it’s rarely possible to create cards that reinforce all the important connections – if you knew all of those connections, you would already know the topic well! A good textbook will bring some of them up as they come along, but you often won’t have the luxury of a good textbook, and even with one, there will be some to figure out yourself. It would be silly to attempt to fully grok a subject prior to beginning to create Anki cards for it; by the time you’ve reached that point, you know the subject thoroughly enough that Anki cards won’t do anything besides help you remember small, rarely-needed details, which is useful but hardly revolutionary.

So how do we resolve this dilemma? We turn learning with Anki into an iterative process – one where you read some material, add some cards, learn the cards, read some more material, add more cards, learn those cards, and repeat until you know enough for your purposes at the moment (see also lazy-learning). On the first iteration, you might learn the basic terms and concepts without which it’s hard to understand what the material is talking about. On the second, you can read much more easily and you can start picking out more complex facts and implications. On the third, you can start making connections to other areas of knowledge and how you might use the information in your work. And so on and so forth. Iterative learning without Anki isn’t a bad approach – indeed, any good course or textbook will do something similar – but iterative learning with Anki is even better: because Anki will prevent you from forgetting what you’ve learned so far, you can stop when you’ve learned enough for the moment and pick up again anytime right where you left off, even years later.

Here’s the dance again. Anki prompts you to remember things as you go about your life, but it’s up to you to decide how deep you want to go and when it’s time to learn more.

Note: You might be wondering how exactly to create cards that “make connections to other areas of knowledge.” The card-design case studies may give you a start. In addition, I’m hoping to publish descriptions of some useful patterns for Anki cards in the future, which will hopefully make this much clearer.

Engage in metacognition

Before we finish up, let’s look at how you can take responsibility for the last item on my list of things Anki doesn’t do for you: understand why you’re finding difficult material difficult.

Metacognition is, put simply, thinking about thinking, or thinking about the way you think. When reviewing cards, it’s easy to get into the habit of uncritically flipping through the cards and rating them. If you do this, you’re missing out on a lot of opportunities for improvement! Your Anki cards are a simulacrum of your mind, and the information in your mind could surely use some improvement. Anki gives you a chance to improve that set of information to your specifications by adjusting the cards in your collection, but it won’t do that by itself. You need to figure out what’s suboptimal about them and tweak the cards.

When you’re reviewing, you might encounter a card that is insufficiently precise; fix it so you don’t keep failing the card for reasons unrelated to your understanding of the material. You might realize there’s a gap in your understanding of a topic; fill that gap (do not learn either a card or a topic without understanding it). You might realize you don’t care about this information anymore; delete that card. You might realize you’re confusing this item with another item; work out the difference so you stop doing that.

A little more on that last one. My mostly unscientific theory of forgetting states that almost anything you repeatedly struggle to remember can be placed into one of the following four categories:

1. You aren’t actively trying to remember it in the first place. (This is what’s happening when you forget where you put your glasses or parked your car – you were distracted, so you weren’t trying to remember where you put it.)
2. You’re trying to remember it without understanding it, connecting it with something else you know or with the context in which you need to remember it, or creating a mnemonic to make something inherently difficult to remember into something reasonably memorable.
3. You aren’t reviewing it frequently enough or at the right intervals.
4. You are confusing it with another fact or concept – memory interference. (This is SuperMemo Rule 11.)

Number one doesn’t apply to us (presumably you’re trying to remember anything you put into Anki). Number two we covered in the previous section. Number three we have covered because that’s Anki’s responsibility, as long as you keep up with your reviews. Number four is the new one.

Sometimes memory interference is clear enough. For instance, my English vocabulary deck has cards for tuff and tufa, both of which are types of rock (but quite different ones). You probably could have predicted that I would struggle to keep those straight. But interfering items can be much less obvious. Here are a couple of other word pairs I’ve been mixing up recently:

• japes and jackanapes: Japes are jokes (especially practical jokes), while a jackanapes is an impertinent person. The meanings are quite distinct, but the sounds are similar enough to cause confusion.
• loggia and parterre: A loggia is a room that is half open to the outdoors, while a parterre is a type of flowerbed arrangement in a garden. These have only the faintest connection in meaning and no connection in etymology or sound, and I have no idea why I kept mixing them up, but I was eventually able to recognize that I was, and I then stopped forgetting both of them repeatedly.

To catch these kinds of interfering cards, you really have to start paying attention to yourself and your thought process as you review; if you’re not aware that interference is a common cause of forgetting, you might well never realize you were struggling because you were confusing two items. But it’s well worth the attention – when you catch interference and add a quick card explaining the difference or tweak your other cards to make the difference more obvious, those cards will likely go from being frustrating and time-wasting to easy.

Conclusion

The mechanics of spaced repetition make up only one piece of the learning puzzle. To be sure, Anki offers a potentially life-changing amount of value, but to take advantage of all that value, or even most of it, you need to see Anki as a partner, build the skills necessary to hold up your end of the partnership, and keep exercising the skills when appropriate.

Perhaps I am in danger of adding too many metaphors to this post, but using Anki well is a skill rather like writing prose. Many valuable rules of grammar and style can be clearly articulated, and I try to do the equivalent for Anki in this series, but any set of prescriptive rules rigorously applied will produce dull and sometimes even comical results. For good results, you have to practice a lot until you develop an intuition for what “sounds good,” and then for each individual piece of writing, no matter how good you get, you have to work your way through each of the sentences multiple times, sometimes throwing out entire sections or going to research and explore new elements of your argument or story. I recommend doing the same with the creation of your Anki cards.

To complete the metaphor, we might liken technical Anki skills to active vocabulary: a few well-chosen, hundred-dollar words here and there will often improve your writing, and sometimes you might need to spend some time finding a specific word on which your argument hinges, but a good vocabulary doesn’t make a good writer. It’s not knowing words that makes someone a good writer, it’s knowing how to use them effectively. Start with learning to work with Anki in an iterative process to create useful cards with the basic tools, and only then worry about optimizing your Anki vocabulary.