Spaced repetition software is a flashcard memorization technology based on the spacing effect. Personally, I think of it as a way of engineering dispositions, a form of self-programming. More concretely, I find that spaced repetition is helpful for
  • internalizing knowledge 
  • compressing recent experiences 
  • conditioning specific future behaviors 
  • making analogies 
  • laying the groundwork for future insights 
  • confusion identification 
  • concept clarification 
  • reconciling models 
  • creating new representations 
  • creating examples 
Below I give some principles, tips, and examples that aim at helping you get the most out of SRS. This post is compact, and I think it will be helpful to re-read it periodically as you use SRS more.

  • SR strengthens connections between mental representations.

There a variety of ways this can happen, seeing as there is both a variety of mental representations and connections between them. For example, the two mental representations could be of a context and a behavior, and strengthening the connection would mean making the behavior more likely in that context. 

  • Mental representations precisely condition behavior.

The point of doing SR is to change behavior (whether mental or physical) and I think it helps to keep the chain of causality in mind. It gives me a framework to think about the different things SR does for me, and how those things are achieved.

  • SR establishes a personal language of thought.
Cards about situations or concepts give you a canonical handle/representation for thinking about them (like Eliezer's clever post titles). This can have positive effects (consistency of thought and building potential) as well as negative effects (potential ridigity of thought).

  • Be very specific when conditioning behaviors.

Make cards that follow the form "Do actionable_behavior_x in specific_situation_y under condition_z." This will limit confusion about whether you should execute the behavior and also make it more likely that you'll remember. For example, "If you're eating when not hungry, reflect on what you're suffering from."

  • Use one unit of meaning per card.

This is similar to Piotr Wozniak's minimum information principle, which is one of his 20 rules for formulating knowledge  (totally recommended). I almost entirely use cloze-deletion cards and I think it's easier to follow this advice with such cards. Cloze-deletion cards are made by deleting parts of a sentence (or an image). For example, starting with the sentence "A lost purpose is a subgoal that no longer serves its supergoal." I made the cards

"A [lost purpose] is a subgoal that no longer serves its supergoal."

"A lost purpose is a [subgoal] that no longer serves its supergoal."

"A lost purpose is a subgoal that [no longer serves] its supergoal."

"A lost purpose is a subgoal that no longer serves its [supergoal]."

where the part in brackets is blank on that card.

  • Discover your own SR style.
I find that a small number of cards that are carefully phrased and highly compressed work well for me. Others use larger numbers of cards with less information per card, and then internally organize and compress the information. Try making different kinds of cards and using other people's decks to figure out what works best for you. You can also vary how long you spend recalling and reflecting on each card during your session. I sometimes spend up to a minute reflecting on a card, during which I do some of the activities I listed above.

Examples - situational questions

Some of my cards are aimed at conditioning myself to ask questions in specific situations. Here's one card inspired by divia:

Front: When you become aware that you are making a social judgment what should you ask yourself?

Back: What need of mine does this reflect?

This card has greatly helped me identify my unfulfilled social needs and outstanding concerns about my own social behavior. At the same time it has helped me increase my ability to empathize with others, and capacity to meet their social needs. 

Here's another similar card:

Front: When you notice yourself pulling your hair what should you do?

Back: Reflect on what you were just thinking about.

Sometimes I run my fingers through my hair when I'm stressed out. I made this card in order to use this habit to become more aware of when I'm stressed, and what my sources of stress are. It has served that purpose fairly well.

New Comment
15 comments, sorted by Click to highlight new comments since:

Anyone making cards like these (as opposed to more clear-cut cases) should be aware of huge potential for scheduling loads of cards that are useless in the long run, and only seem like a good idea at the time. Since the cost of maintaining knowledge of these facts is both delayed and distributed, I expect it's not adequately perceived.

What about simply deleting items no longer useful?

If the cost is inadequately perceived, you'll delete less than you should, and you'd have to pay the cost of considering the question of whether to delete in any case.

From my personal experience, reviewing cards feels costly, and I am likely to delete them if I do not see that they need be in my current 'working set'. This is especially true if you timebox card review.

We'll be using Anki cards like these to boost retention of new habits and skills at mini-camp (Luke Grecki made the post partly at my request, as a resource for the mini-campers). So if anyone has any further suggestions for how to use Anki cards to change one's habits or ways of thinking, please share them now, so that they can affect our practice.

One thing I am not quite sure of is when to rely on simple lookup (I keep a lot of information in an text file) vs. when to commit to memory with SRS.

Very good. Additional factor in this that I'm somewhat concerned about is "cluttering your head" for a lack of a better term, though this might be a non-issue if you know how memory actually works (I don't). Would be interesting to see it as a part of you write-up.

I don't really know of anything like that (aside from fictional evidence like 'Funes the Memorious'); there is research on weakening memories, but that focuses on stuff like rape and post-traumatic stress disorder.

If you were wrong that knowing something would save you 5 minutes, you can always just delete those cards. (The brain keeps plasticity even into old age, so whatever limit there is seems to not be reached in a normal life.)

If I remember correctly the supermemo website gives advice about this, somewhere. Its recommended not to be too selective about what to enter into the SRS system. If you find yourself spending a lot of time with an item that's not very useful, simply delete it later on.

This seems to be a useful technique, thanks for introducing it.

I have a bit of criticism concerining the article: It needs more introduction. Specifically, I would guess I'm not the only one who doesn't know what SR is in the first place; a few sentences of explanation would surely help.

Previously discussed a fair amount on Less Wrong. I made a wiki article and linked some of the articles/comments.

Thank you for the link!

I think would fit well into the introduction. You (or rather Luke Grecki) could just split the "spacing effect" link into two.


This is really interesting. I think the idea can be extended by adding theory on planned behaviour

''An implementation intention (II) is a self-regulatory strategy in the form of an “if-then plan” that can lead to better goal attainment, as well as help in habit and behavior modification. It is subordinate to goal intentions as it specifies the when, where and how portions of goal-directed behavior. The concept of implementation intentions was introduced in 1999 by psychologist Peter Gollwitzer.[1] Studies conducted in 1997 and earlier showed that the use of implementation intentions can result in a higher probability of successful goal attainment, by predetermining a specific and desired goal-directed behavior in response to a particular future event or cue.[2] History

The concept of implementation intentions originated from research on goal striving throughout the late 1980s and 1990s. Developing research suggested that “the correlations between intentions and behavior are modest, in that intentions account for only 20% to 30% of the variance in behavior.” Strong intentions (“I strongly intend to do X”) were observed to be more often realized than weak intentions. Past behavior still tended to be a better predictor for a person’s future behavior when it was compared to goal intentions. The research also suggested that the weak intention-behavior relation is a result of people having good intentions, but failing to act on them.[3]''