What is worth remembering?
Or, You Should Probably Do More Flashcards, Here's Some Ideas
There’s a grand tradition in intellectual work of posing a hard question, and then instead of answering that hard question, answering an easier one. In that spirit…
What is worth remembering? What flashcards should I add to my Anki deck?
Prelude
“I binged The Consciousness of the Litigator one night and generated a glossary of flashcards around law that I have yet to review the vast majority of. On one hand, I’m interested in how the U.S. legal system works. On the other hand, I’m interested in everything, and I only have so much time.” -a very relatable Parker Conley
A little secret about me is that I have been a regular user of spaced repetition software (SRS), mostly Anki, for the better part of two decades. The key function of SRS is to efficiently schedule reviews of flashcards to move the information into long-term memory with as little work as possible. And it works.
Ever since I developed a somewhat obsessive fascination with languages circa 2007, SRS has been a core part of how I have studied them. I’ve also used SRS for learning and remembering many other things, such as material from my two years of graduate classes at Georgetown which I had to recall for a comprehensive exam at the end of the program.
It’s not quite exaggeration to say that I love Anki. To me, it is a key tool that makes pursuing and sustaining certain kinds of meaningful personal goals convenient and efficient.1
However, I’ve long suspected there was even more value to be had from SRS than what I was getting. That’s one reason I was excited to attend Memoria recently,2 a cozy little conference focused on “memory systems” — largely SRS, but also other instances of this category such as SuperMemo, a system designed to support “incremental reading” (no, I don’t really understand what this means either).
Memoria was a rollicking good time filled with equal parts giddy nerdery and subtle humanistic reflection. Such a thoughtful group cannot approach the topic of memory systems without being drawn to the fundamental questions — how does memory enable us? how does it shape us? constrain us? what should we care to remember? what do we care to remember, whether we “ought” to or not? what, ultimately, is worth remembering? — questions that I will sidestep completely. Instead, I offer a listicle.
Things You Haven’t Considered Making Flashcards Of, But Maybe Should
If you’ve ever used flashcards, it has probably been for learning vocab in a foreign language, dates in American history, or whatever else was on a test that you needed to pass. You may not have any tests to pass these days, but I am willing to bet that there are more things it would behoove you to memorize as a Substack-reading denizen of the modern world than you realize. I hope the following suggestions can help make your life easier, more successful, and richer in fascination, meaning and joy through the magic of Knowing Things.
(Note: The below is a mix of anecdote, intuition and speculation. I have not necessarily extensively, much less rigorously, experimented with all the below categories, but I believe them to be worth considering. I provide this is in the hopes of promoting more tinkering and exploration. I suspect my own thoughts on this topic will evolve over time as well.)
Everyday practicalities
One of the best ways to improve your day-to-day experience, in my opinion, is to find opportunities to reduce friction. One significant source of friction is not having information that you need to make quick, daily decisions readily available.
You grab a loaf of bread at the grocery store — but wait, was your new girlfriend allergic to pumpkin seeds or sunflower seeds?
You get out of a train — wait, which exit is closest to your destination again?
Some bureaucrat asks you for a copy of your ZF-5 — which form was that again? … and where did I store it?
These kinds of moments are “not a big deal.” But modern life is chock full of them, and life feels much smoother if you can dispatch them without breaking your stride. Querying your long-term memory at <200ms latency will save you time relative to breaking out your phone, waiting for a text back, searching Google Maps, rifling through your files. But more importantly, it will make you feel more competent and in control of your life.
One way to think of this is as knolling the database of your mind. Your frequent-use information tools were certainly probably within reach somewhere amid the cluttered counter already, but there’s something comforting and conducive to high performance about totally clear, reliable, frictionless access to them.
Examples:
Important information about loved ones (birthdays, allergies, preferences…)
Details about yourself and your life (where you keep important things, your passport number)
Every major street in your neighborhood, or your whole city
All the keyboard shortcuts in every piece of software you use regularly (this one I strongly recommend)
Things you’re somehow inexplicably incompetent at and always forget (every adult has at least some)
As one example, recently I’ve been learning the major streets in San Francisco. It’s silly that I’ve crossed S Van Ness or taken the Muni up Church so many times and couldn’t already have told you the major street between Mission and Folsom or what street I was on if standing outside Beit Rima next to the Safeway off Market, but such was my situation.3 I’ve been surprised the extent to which it feels empowering and grounding to fully knit together all my knowledge of the city, as if beforehand I was almost always just a little bit lost. With a little bit of effort, Anki can help you go from chronically lost, dependent, and frazzled to collected, comfortable, and flexibly prepared for life.
Things that provide structure
Often the highest-value information we learn in any given academic field are the abstract frameworks and mental models that help us make sense of and work with specific cases. In cyber operations, there’s the Diamond Model; in project management, there’s MOCHA; in China studies, there’s “fragmented authoritarianism”; in narrative studies, there’s the hero’s journey. A primary benefit of these frameworks is accelerating our ability to solve problems in the relevant domain. As with mundane practicalities, being able to do this faster and smoother directly translates to greater mastery.
These structures are particularly high-leverage for learning because they help us organize and create hooks for details. If you know the hero’s journey model like the back of your hand, it’s a lot easier to remember the details of 10 novel plots that roughly map onto it. If you’re familiar with the fragmented authoritarianism model, it’s a lot easier to remember the details of some story of PRC bureaucratic infighting as a manifestation of the dynamics of the model.
Keeping these frameworks explicitly in mind, and focusing on expanding your repertoire, is also helpful for reminding yourself that any one such structure is only one lens on a certain type of question. Often by default, we get so comfortable with one framework that we may fail to notice when the world starts to diverge from the assumptions that underpin it. We want to use our frameworks, but not fuse with them. Anki can help ensure you are regularly reminded of all the tools in your toolbox, not just the favorite hammer you always reach for first.
There’s another kind of structure-giving information that can be very useful to remember: points of reference. For certain kinds of information, a fact on its own can be of very little use without more information about where it sits in relation to other facts. This is often particularly the case for quantitative information. If I tell you I have 100 widgets, your first question is probably: is that a lot? Points of reference provide the trellis that allows you to quickly contextualize particular details.
Knowing a lot of such point of reference facts can give you uncanny Fermi estimation abilities. This in itself is a fun party trick (if your friends are nerds, at least), but can also be very useful in practical circumstances. You cannot literally play the semiconductor manufacturing ecosystem like a violin, but if you are as fluent with all the facts of its operation as a professional violinist is with their fingerboard, you can identify facts and contemplate counterfactuals with the virtuosic, seemingly effortless flexibility of Itzhak Perlman.
Examples:
Points of reference across different orders of magnitude, eg the amount of energy generated and consumed by various entities — a car engine, a portable generator, a nuclear power plant; a toaster, a household, the United States
Meaningful points in a continuous variable, eg biographical birth and death dates of scientists as a latticework for the timeline of the history of science & technology4
Canonical examples of a genre in detail, eg several sonnets that illustrate common patterns for understanding other sonnets, or academic papers using a methodology relevant to you
Well-known songs that provide an example of each musical interval (eg the Jaws theme for the ascending minor 2nd)
Things you will enjoy remembering
From the above, you might be thinking of SRS as a tool for serious endeavors that are useful but not particularly fun. But there is no reason you can’t use it — even exclusively! — for purely pleasant purposes. You don’t even necessarily need to use a card to prompt you to recall the thing, ensuring you memorize it (although it could be rewarding to try to remember a favorite painting in increasingly great detail). You can simply have a card that serves to expose you to the thing regularly.
In this case you probably want to orient towards the scheduling algorithm slightly untraditionally. Marking a card as “hard” will bring it back earlier; marking “easy” puts it further out in the future.
Examples:
Paintings, poems, anything else you find beautiful
Prompts that elicit a feeling you once had
Pictures of places you’ve traveled5
Vibes that you want to integrate
Sometimes some information doesn’t sink in emotionally and intuitively as much as it should. Maybe some contradictory information lodged itself in your brain, or maybe it’s just not the kind of thing that sticks with you. (For example, for many people negative experiences adhere more vividly in their memory than positive experiences) Repetition of the information can provide the opportunity to integrate information and give it the subconscious weight you consciously believe it deserves.
Examples:
Positive (or constructive) feedback you’ve gotten from people
Life lessons you’ve learned (but might forget by default)
Things you’re liable to unwisely let slide, if you’re the kind of person to unwisely let things slide
Habits
Flashcards are not particularly well suited for helping you achieve habits like “be like Christ” or “be a high-agency person.” However, they can be helpful for achieving much more granular, concrete habits that can serve as building blocks towards those broader value-driven ambitions. SRS is a tool for shaping automatic responses, so formulating habits as automatic responses — when X, do Y — makes them amenable to the SRS format.
This happens to also be a widely-recommended way to build habits. In rationalist jargon, they are called Trigger-Action Plans, ie, plans in which when you observe a certain Trigger, you take a certain predetermined Action.6 The basic idea of the wildly popular Atomic Habits is creating these kind of automatic-response habits and stitching them together into a life where you make good decisions on autopilot.
Risks and opportunities in the tails
Some information you might expect to never have occasion to use, but on the off chance that you do, it’s extremely valuable that you are able to, either to mitigate some risk or to take advantage of an opportunity. Particularly if it’s unpredictable when you might need to know this information, SRS is an excellent solution to ensuring you always have it at the ready.
Examples:
First aid/emergency medicine — might actually literally save a life
Emergency preparedness information — might actually literally save several lives!
People you meet (eg, at a conference) and what they work on/know about/are interested in
A shared interest that someone you look up to mentioned on a podcast — a good conversation topic in case you run into them
Dangerous snakes that live in your region
Absolutely everything possible in your domain of expertise
Being an expert means being expected to have an all-things-considered view, and being able to marshal all the requisite details to substantiate that view. In some sense, it’s kind of crazy that we don’t expect experts to continuously maintain comprehensive knowledge of the core of their domain as a standard matter of standard professional practice. Fortunately, it generally becomes easier to learn and remember new information as you develop increasingly densely architected lattices of related facts and frameworks. So why not just go big?
One of the great values of an expert is being able to draw together obscure facts about a particular domain to shed light and create insights that a non-expert would not be able to. Often those obscure facts are surfaced by the expert just happening to read something at an opportune moment. But how many things might they already have read which could provide the same value, if only they remembered them or were prompted by a card?
SRS can also be useful as an expert in the crucial endeavor of keeping yourself epistemically humble. Over long years of study and activity in some domain, by default the information that sticks with us is more likely to be that which accords with our biases (especially whatever anchored them down in the first place). SRS provides an opportunity to ensure you are reminded of counterexamples to your prevailing notions.7 It also provides an opportunity to keep your central cases and the evidence for them concrete and grounded, rather than fading into indistinct, unassailable mythic truism with the rosy glow of long-held memory.
The challenge here is identifying what your “domain of expertise” is, or should be. For generalists or the flighty, there may not be such a thing, even among knowledge workers. But if there is a naturally well-scoped area in which you plan to engage intellectually for years and decades, it might be worth considering remembering basically absolutely everything.
(Note: I feel less certain about this item overall. I’d be interested in hearing from people who have made serious efforts to do this over a long time period. In any case, do not attempt without considering some caveats below under “However…”)
Low-hanging fruit
Some things are just right there in front of you. We’ve all had the experience of feeling silly that we can’t remember something very obvious and commonplace, or which we learned long ago, that’s just not exactly a part of your day-to-day life. Some things would just be really easy to know a little more deeply. Having some diversity in your decks generally feels good, so throw in some easy stuff too to give yourself a little break and some cheap dopamine sometimes.
Examples:
Acronyms — if you already know the acronym, you basically already have a mnemonic for the full name. Why not learn it? It can come in handy, particularly when talking to non-specialists. Often full names contain information that helps you remember what the thing is more precisely. Of course, it’s also just mildly embarrassing to not know, or mess up, the name of something ostensibly within your domain of expertise.8
Mnemonics you learned in school
However…
Before you go hogwild with Ankifying your entire lifeworld, consider a few cases where you should not make flashcards:
“Orphan” cards — It’s easy to learn things that are densely connected to other things you know, and the converse is also true. These will generally be more of a drag than they’re worth.
Things that you really won’t enjoy — If Anki feels like a drag, rather than a piece of beautiful architecture supporting your flourishing, it will begin to cast a pallor across everything you’re choosing to remember. It may not be clear in advance what things will cause this effect; as you do your reviews, you should be ruthless in deleting or suspending cards that threaten to kill the vibe.
Ill-formatted information — In general, each card should be atomic, with just one piece of information, to be efficient. I recommend reviewing other best practices listed here.
Acknowledgements
Thanks are due to many Memoria attendees for conversations that fed into this post, but particularly to Saul Munn, both for his heroic work making Memoria happen (alongside Raj Thimmiah!) and for insightful discussion and enthusiasm.
It is also fast, free on desktop and reasonably priced with a one-time fee for the mobile app, and free from ads — things I wish were true of much more of the software I interact with.
Ok, at this point it was really not that recently. Alas…
With technology, it’s become incredibly easy to be a little bit incompetent in myriad ways. In my opinion, there is something to be said for insisting on personal, embodied competence at least a bit more than is the default, even when it doesn’t necessarily clearly pencil out.
Thanks to Gary Wolf for this example.
One of many wonderful ideas shared by Michael Nielsen at Memoria.
It’s also potentially relevant to what Yudkowsky refers to as skills on “the 5-second level.”
Keeping an eye on perplexing anomalous cases over the years might also eventually present them in a context that provides some new solution or insight under which the anomaly is resolved. Innovation is generally combinatorial; the spaced repetition algorithm is one way to add some random combinations of [thing you’re currently thinking about] with [thing you’ve previously thought about] into your life.
Although I don’t recommend thinking this way, it is also a great flex to know precisely when even most experts have forgotten what the alphabet soup stands for.