Anki's Not About Looking Stuff Up

Attention conservation notice: if you've read Michael Nielsen's stuff about Anki, this probably won't be new for you. Also, this is all very personal and YMMV.

In a number of discussions of Anki here and elsewhere, I've seen Anki's value measured in terms of time saved by not having to look stuff up. For example, Gwern's spaced repetition post includes a calculation of when it's worth it to Anki-ize threshold, although I would be surprised if Gwern hasn't already thought about the claim going to make.

While I occasionally use Anki to remember things that I would otherwise have to Google, e.g. statistics, I almost never Anki-ize things so that I can avoid Googling them in the future. And I don't think in terms of time saved when deciding what to Anki-ize.

Instead, (as Michael Nielsen discusses in his posts) I almost always Anki-ize with the goal of building a connected graph of knowledge atoms about an area in which I'm interested. As a result, I tend to evaluate what to Anki-ize based on two criteria:

  1. Will this help me think about this domain without paper or a computer better?
  2. In the Platonic graph of this domain's knowledge ontology, how central is this node? (Pedantic note: it's easier to visualize distance to the root of the tree, but this requires removing cycles from the graph.)

To make this more concrete, let's look at an example of a topic I've been Anki-izing recently, causal inference. I just started Anki-izing this topic a week ago, so it'll be easier for me to avoid idealizing the process. Looking at my cards so far, I have questions about and definitions of things like "d-separation", "sufficient/admissible sets", and "backdoor paths". Notably, for each of these, I don't just have a cloze card to recall the definition, I also have image cards that quiz me on examples and conceptual questions that clarify things I found confusing upon first encountering these concepts. I've found that making these cards has the effect of both forcing me to ensure I understand concepts (because writing cards requires breaking them down) and makes it easier to bootstrap my understanding over the course of multiple days. Furthermore, knowing that I'll remember at least the stuff I've Anki-ized has a surprisingly strong motivational impact on me on a gut level.

All that said, I suspect there are some people for whom Anki-izing wouldn't be helpful.

The first is people who have the time and a career in which they focus on a narrow enough set of topics such that they repeatedly see the same concepts and rarely go for long periods without revisiting them. I've experienced this myself for Python - I learned it well before starting to use Anki and used it every day for many years. So even if I forget some stuff, it's very easy for me to use the language fluently after time away from it.

The second is, for lack of a better term, actual geniuses. Like, if you're John Von Neumann and you legitimately have an approximation of a photographic memory (I'm really skeptical that he actually had an eidetic memory but regardless...) and can understand any concept incredibly quickly, you probably don't need Anki. Also, if you're the second coming if John Von Neumann and you're reading this, cool!

To give another example, Terry Tao is a genius who also has spent his entire life doing math. Probably doesn't need Anki (or advice from me in general in case it wasn't obvious).

Finally, I do think how to use Anki well is an under-explored topic given that there's on the order of 10 actual blog posts about it. Given this, I'm still figuring things out myself, in particular around how to Anki-ize stuff that's more procedural, e.g. "when you see a problem like this, consider these three strategies" or something. If you're also experimenting with Anki, I'd love to hear from you!

Showing 3 of 8 replies (Click to show all)
3NaiveTortoise1moSo... I just re-read your brain dump post and realized that you described an issue that I not only encountered but the exact example for which it happened! I indeed have a card for Newton's approximation but didn't remember this fact! That said, I don't know whether I would have noticed the connection had I tried to re-prove the chain rule, but I suspect not. The one other caveat is that I created cards very sparsely when I reviewed calculus so I'd like to think I might have avoided this with a bit more card-making.
3riceissa1moI want to highlight a potential ambiguity, which is that "Newton's approximation" is sometimes used to mean Newton's method [] for finding roots, but the "Newton's approximation" I had in mind is the one given in Tao's Analysis I, Proposition 10.1.7, which is a way of restating the definition of the derivative. (Here [] is the statement in Tao's notes in case you don't have access to the book.)

Ah that makes sense, thanks. I was in fact thinking of Newton's method (which is why I didn't see the connection).

NaiveTortoise's Short Form Feed

by NaiveTortoise 1 min read11th Aug 201885 comments

In light of reading Hazard's Shortform Feed -- which I really enjoy -- based on Raemon's Shortform feed, I'm making my own. There be thoughts here. Hopefully, this will also get me posting more.