This is adapted from a presentation I gave at the Berlin Quantified Self Meetup in November 2012. There are two parts:
- What is spaced repetition software?
- Why don’t more people use it, and what can we do about that?
Part 2: Why Isn’t Spaced Rep Software More Popular?
In my previous post, I introduced the field of spaced repetition software; in this one, I’ll discuss why it hasn’t spread further and some new apps that might change that, including one that I’m working on myself.
First of all, I’m not the only person asking this question. Here’s a quote from Gary Wolf, co-founder of the Quantified Self movement, who wrote a great article on Piotr Wozniak and Supermemo:
Why isn’t this amazing technique more common? I explained some of the obvious reasons in my story. Still, I expected that, having launched the idea into an environment well suited to nourish it (many Wired readers are passionate learners, and many of them have software, design, and business skills), I would soon see some new implementations. And I was not disappointed. There are half a dozen versions of Supermemo in common use today. But they are used by very few people. Clearly, the problem remains unsolved.
And the Supermemo wiki has a long list of ideas, some more constructive than others. But they’re clearly aware of the problem.
I have some first-hand experience with this: when I was using Anki and Mnemosyne to study German, I recommended them to many of my classmates, and a few of them got as far as installing the software and loading up my German/English deck (which I would email them) but none of them really stuck with it. And these were smart, relatively computer-savvy people, often with better overall study habits than me, but this kind of software just didn’t quite “click” for them.
Why not? Here are some of those “obvious reasons”…
- It’s difficult / mentally taxing. No use beating around the bush. These apps do require a certain type of concentration that’s a little uncomfortable at first. But once you get over that small initial hurdle… well, it’s still not Angry Birds exactly, but I find it a lot less difficult (and more satisfying) than other study methods.
- None of the engagement methods of other educational software, like levels and achievements, social features, or “gamification.” I’m a skeptic on just how engaging (as opposed to distracting) some of these other features are, but there’s no question that some of them help, especially early on.
Not enough feedback, or not the right kind. This is a critical one. Notice that the graph I showed you — progress over time — does not come standard in Mnemosyne. I had to record that data every day in Excel and make it myself. You can get “snapshot” charts in the app, but not time-oriented ones, which are a better motivator.
- Algorithm is too “strict.” If you stop using these apps for a few days, you’ll pile up lots of “overdue” cards. In theory you can ignore them and just take it slow, but there’s still that counter in the corner of the screen reminding you that you’re “behind.” In my mind, a good study app should not try too hard to dictate the pattern of usage, but rather adapt to make the most of whatever time the user puts in.
There’s also the “one piece of info per card” rule, which I’ll come back to. And because of the moving average nature of the algorithm, it can be hard to convince the software that you really know a card, and you’ll often keep seeing it after you’ve rated it very highly. Finally, when I said earlier that you can define the ratings for yourself, I think some purists would object to that too.
Here are two other reasons that I haven’t seen elsewhere, although to me they’re equally obvious:
- Still mostly client-side apps, not cloud-based
- syncing among different devices is annoying
- risk of data loss
- compatibility and version issues
- Still based on user-generated content
entering your own cards is time consuming, especially on a phone or tablet
downloading other users’ decks isn’t always much better
formatted text and other rich content are difficult
The risk of data loss is very real. It’s easy to say “well, you should be backing up everything anyway,” but the reality is that most of us don’t, and it’s hard enough to get users to adopt this one new habit without making them change others.
And the focus on user-generated content is a real problem as well. Most people are just never going to enter their own cards in large numbers. Downloading other users’ decks introduces the risk of typos and grammar mistakes (a huge problem if you’re learning a new language, less so if you’re studying something else) and, more importantly, their content is unlikely to be an exact match for what you want to learn, and it may not be comprehensive or well-formatted. I’ve found that downloading other users’ decks is a good way to try out the software, or try out a new subject, but when I know what I want to learn (say, B1 German vocabulary), I spend as much time editing someone else’s deck as I would in creating my own. And it’s very time-consuming to go beyond a small amount of unformatted text (or maybe a single image) on each side of a card.
Enough complaining — what can we do about it?
Well, here are some encouraging steps forward:
- Mobile, tablet and web extensions for Mnemosyne and Anki
- Fully “cloud-based” implementations (Kleio)
- Shared editing “wiki” models for user-generated content (Memrise)
- Moving towards professionally-created content (Skritter, Chinese3D)
The mobile and web tools around Mnemosyne and (particularly) Anki are getting better all the time. There are also some people building fully cloud-based implementations, and I recommend Kleio if you want to try one. Memrise uses some other mnemonic methods in addition to a form of spaced repetition, but in any case they’re the first app I’m aware of that lets users edit the same shared decks, so their corrections and tweaks are cumulative. This also allows a lot of “meta” content, with more than one piece of information per card. I think these things have a lot to do with the fact that they’ve achieved a relatively large user base in a short period of time, and (arguably) wider name recognition than the other apps I’ve been talking about.
The alternative to the “wiki” approach, of course, is professional content, and I know of at least two apps that do that for Asian languages: Skritter and Chinese3D. In this case, fixed professional content allows much more intricate graphical interfaces and user interaction, including drawing the actual characters on the screen in Skritter. If you’re studying Chinese or Japanese, I recommend checking these out.
And finally, I’ve developed my own prototype for European languages, with each pair of languages built from the ground up by professional translators and tagged according to real-world metrics (like the A1-C2 European language standards, or particular official exams). I’ve also simplified the algorithm and rating scale, added lots of extra content and formatting to each card, and made various other attempts to address the issues above.
The first “deck” is German for English speakers, which is now up to several thousand cards. We’ve got the web app up and running, and we’re starting on native mobile apps now.
I won’t go into more detail about my project here, but if you’re learning German and want to be a beta tester, send me an email at firstname.lastname@example.org and I’ll get you set up.
In any case, if you’re new to the subject of spaced repetition, I hope I’ve gotten you interested enough to try at least one of the tools I’ve mentioned. If you’re a user or creator of a tool that I haven’t mentioned, I’d love to hear about it and would be happy to append it to this post. And if you have other questions or just want to discuss (or disagree with) anything I’ve written here, just drop me a line.