What is a Memory System?
A spaced repetition system or better called a memory system is meant to test whether you can reconstruct an idea (what you want to understand deeper and remember) from a specific piece of information (a prompt) using your internal models (current understanding). A memory system is not a process for “putting information into a bucket.”
- What is the goal of a memory system?
To test whether you can reconstruct an idea. - In a memory system, what serves as the cue for reconstruction?
A specific piece of information: the prompt. - In a memory system, reconstruction happens from what?
Your internal models.
Good prompts help you reconstruct from:
- A generative explanation
A causal story: why the claim is true, what problem it solves, what makes alternatives fail. - Problem-situations and triggers
The kinds of situations where the idea should come to mind. - Counterfactual structure
What would be different if the idea were false; what would follow if it were true. - Error boundaries
Common misapplications, confusions, and nearby false friends. - Procedures / decisions
What you do differently because you know it, what choice you make, what question you ask next, what you stop doing.
Ideally, that information is not merely a fact, but an explanation. That is the key question when you design a prompt: What material would reliably let me reconstruct the explanation I want? When the prompt works, repeated reconstruction helps that explanation take hold in the mind’s ecosystem, where ideas survive or fade, influence behavior or do not, and interact with one another.
- There are 5 handles good prompts let you reconstruct from. Which one says you "have" the idea?
The generative explanation. - Why is the handle, Problem-situations/Triggers, valuable to include as an internal model?
It lets you know when to use the idea. - When should you try to create the 5 memory system handles, during prompt creation or when attempting to understand the idea?
When attempting to understand the idea.
Properties of effective reconstruction prompts (based on Andy Matuschak's Good Prompts)
- Reconstruction prompts should be focused.
- Reconstruction prompts test one relation at a time.
- Reconstruction prompts should be precise about what they’re asking for.
- Reconstruction prompts should produce consistent answers each time you perform the task.
- Reconstruction prompts should be tractable.
- Reconstruction prompts should be effortful.
- State one thing that could be wrong with the following prompt?
Good reconstruction prompts test one r_ at a time, should be f_, p_, produce c_ answers, t_, and e_.
It tests many relationsips / is not focused enough / could be intractable. - There are several properties of good prompts but what is the general idea?
Make the prompt tight, clear, and worth the mental work.
We prefer explanations because they are often the kinds of ideas that can influence what we do rationally. But no general rule can tell us in advance what material will let a particular person reconstruct a particular idea. The only method is to guess, test, and revise for each individual.
Why reconstruction? Because you cannot reconstruct what you have never constructed in the first place. You must first understand something at least partially. Reconstruction then reinforces that understanding and may deepen it.
See also: Andy Matuschak's Good Prompts (I replace "retrieval" with "reconstruction" based on my understanding of the mind)