Spacey
What is Spacey?
Spacey is a ES6 library implementing a review prioritisation system. It is intended to allow developers to build systems that help users optimise the time they use to memorise facts. I originally started this project in Coffeescript, but lost interest pretty quickly after I finished ranting in this document. I'm coming back to it mainly because I want to write some code and I don't care what.
What is Spacey not?
In some ways it is easier to describe Spacey by contrasting it to what it is not.
-
Spacey is not flash card system. It can be used to build a flash card system, but it can also be used to in other areas where review is necessary. See the Future Steps section for an example of what I want to use Spacey for.
-
Spacey is not a Spaced Repetition System (SRS). Spacey will include an SRS system, but it will be used in conjunction with other techniques such as spacing (different than spaced repetition) and interleaving.
Using Spacey
Spacey is currently in development and doesn't do anything yet. Note that Spacey is a development library. It is only useful for developers writing applications.
Helping with Development
If you wish help with development, you will need the following:
- node.js: latest version (11.6) as of this this writing
To run the tests type: npm run test from the main directory.
Features/Anti-Features (Depending on your point of view)
ES6 code base
Mainly I wanted something that would make it easy to write quiz systems that would be easy to port and deploy on different systems. The precursor, JLDrill was impossible from that perspective. I also just enjoy writing ES6 code. Whether it is actually useful or not, for me that's the biggest issue.
Unique SRS algorithm
Spacey uses a unique SRS algorithm. I could have used something proven like Supermemo's algorithm (that's in use in systems like Anki) but I don't like the fact that it is ubiquitous. Also, quite a lot of the rationale behind Supermemo's algorithm seemed completely arbitrary to me. There are also a few advantages to my algorithm (which I will describe at another time).
What are the differences between Spaced Repetition, Spacing and Interleaving
Spaced Repetition is a technique to periodically review something so that the slope of the "forgetting curve" on that thing is lessened. The "forgetting curve" is the speed at which you forget something. It measures the probability that you will forget something over time. For example, if you don't review something for 5 minutes, you might have a 95% chance of remembering something. If you wait 1 day, maybe that drops to 60%. If you wait a week, perhaps it drops to 20%. By periodically reviewing something, the slope of the curve diminishes. After you have reviewed something several times you might still have a 95% chance of remembering it even if you never see it for a week.
The trick to spaced repetition is to wait until the best possible time to review something. If you forget something, you must relearn it, which takes time. Most spaced repetition systems try to time the review so that you remember it just before you are likely to forget it. This minimises the number of times you must review something, while avoiding relearning.
Spacing is a different technique than Spaced Repetition (with a similar name, unfortunately). The idea is that if you wait until you forget something and then relearn it, you will remember it better than if you had reviewed before you forgot it. In other words, the forgetting curve will have a shallower slope than if you used spaced repetition and continually remembered it.
The trick to spacing is to intentionally forget something and then relearn it. This appears to work best when initially learning something, so it may be that spacing is best for new items, while spaced reptition is best for older items. This is an area of active research, though, so solid evidence is sparse. Some have called the "spacing effect" the most powerful effect for learning that there is.
Finally, Interleaving. Interleaving is the idea of studying for a bit on one subject and then switching to another subject. You then study a bit for that subject and then swithch to yet another subject. Eventually you go back to the first subject.
So if you had 3 subjects (A,B,C) that each took 30 minutes to learn, with interleaving you might spend 10 minutes on A, 10 minutes on B, and 10 minutes on C, before cycling back to A. You would then cycle through all the subjects 2 more times. The theory is that you will remember the items better than if you had studied A for 30 minutes followed by B for 30 minutes, followed by C for 30 minutes. In other words the forgetting curves for the items will have shallower slopes.
Because interleaving causes spacing to happen (i.e., you may forget what you did in your first 10 minutes when you start the second 10 minutes), it is unknown whether interleaving is an effect in itself or simply an easy technique for causing spacing to happen. There is some speculation that changing topics makes it easier to forget what you have learned, which triggers the spacing effect faster. In other words, because you switched topics rather than just having a rest, you forgot the first part faster than you normally would. This allows the spacing effect to be used the second time around. I you had just rested, you might still remember the material, meaning that you can't use the spacing effect.
Most people who have used spaced repetition realise how massive a benefit it is for learning something. It has been said that spacing and interleaving eclipse spaced repetition in effectiveness by at least a binary order of magnitude. That is why I really want to try to find some ways of implementing it in Spacey.
Future Steps
While I have avoided talking about why I am writing Spacey, there are actually a few things that I have in mind.
First, when I originally wrote JLDrill, I chose Ruby and GTK+ because I was interested in learning Ruby and I was only concerned with deploying on Linux based systems. As development continued, I became increasingly frustrated at the difficulty in deploying GTK+ Ruby applications on Windows and Mac. While not impossible, it is quite difficult and it cut down on potential contributors. The set of people learning Japanese while using Linux based systems and secure enough with Ruby to install the app through a Gem is quite small.
Quite some time ago, someone suggested that I switch to making a web application. I initially threw out the idea as I did not want to maintain a website for others to use. However, the idea lodged itself in my brain and because deployment of JLDrill is so difficult, I may yet do it.
The choice of ES6 as a development language is strange. I admit this, but I enjoy writing code in this environment, so why not?
Some ideas I have for Spacey are more practical, though. First, I can see potential in utilising it together with browser plugins like Rikaichan. In fact, if Rikaichan had had a spaced repetition system built into it, I would never have written JLDrill. The whole idea of being able to browse the web searching for unknown words/sentences and then easily sticking them into a review system appeals to me greatly. I think writing Spacey is a much better idea that extending JLDrill so that it can browse the web.
Still with plugins I can see potential for other useful things. Human languages are not the only thing that can benefit from review systems. It occurs to me that we have a learning problem amongst programmers. We don't identify and study good code. A plugin that allowed one to browse something like Github and help the user review code that was good, so that it becomes second nature to them might be useful.
Even more ambitiously, I have an idea that such a learning system could be useful for training new users on a website. I once jokingly suggested that we use electric shocks in our application to ward users away from features that were buggy. The idea is to train them to use it "properly" (where "properly" means the way we tested it).
In seriousness, trying to build a system that helpfully reminds people how to do things at regular intervals might be beneficial. Making a system that determines what thing is more important to be reviewed at any given time could be useful. Although, I am aware of the risk of developing the next MS Bob.
In any case, at the moment, this is mostly a programming exercise for me, but in the long term I think there are some opportunities for Spacey. Only time will tell.