How to Interconnect Ideas as an Anki learner (3 Ways)


“Knowledge in bubbles.”

From high school all the way up to my senior year of college, I’ve ALWAYS felt like all my knowledge live in bubbles.

I learned them through cramming, and then after the exam, the bubble would “pop” and BAM…

…I’m a blank slate again. Tabula rasa action.

As an Anki learner, you would guess that this wouldn’t happen because of the power of spaced repetition, but really, you just get a better problem:

As long as you review, your bubbles don’t “burst”, but the knowledge is confined within your decks. (Ehem, why do you think I recommend as few decks as possible, anyway?)

There’s a ton of knowledge inside each bubble but you may feel like they should be updated over time and linked to each other.

At the very least…

My guess is that you may heard of the Zettelkasten Method or Tiago Forte’s concept of “second brain” where you interconnect everything you know and then thought about how you can apply it into Anki.

But how the heck do you manage that when you already have hundreds and hundreds of cards?

What if there are different subjects in different decks, but only some of the cards are connected?

What if there are different authors who use different terminology to refer to the same concepts?

And surely, you still are making “atomic” flashcards, but that’s yet another problem: with hundreds of cards, how the heck are you supposed to interconnect everything?

You can borrow from the principles of effective Zettelkasten practice…

You see, when you learn about the Zettelkasten Method for the first time, what you commonly see online and what other people teach is “linking notes.”

But in reality, especially when it comes to note-taking…

People ALWAYS miss the mental processes.

Here, specifically, you are not “linking notes” but rather “connecting ideas” and then linking notes. The linked note is merely something that you leave in a note for your future self to act on.

It’s just like how you take effective notes using the encoding paradigm as I say in the study system course:

Your notes should mainly consist of the ideas you have processedrather than the raw information.

This applies to the connection as well, since a connectionis a note.

Put simply…

The connection happens in your memory, and you are just making it explicit by giving it a physical form either by:

  1. Leaving a link and link context in a note; or
  2. Making a separate note about the connection

The answer is the similar when you apply it to Anki — with a bit of modification upon implementation:

You don’t focus on interconnecting the cards; you focus on interconnecting ideas.

You make the connection in your memory, and you optionally make it explicit either by:

  1. Writing about the connection in the card that contains your target idea; or
  2. By creating a separate card that asks you about the connection

Optionally, because the thing about memory is that when you know the individual ideas and how they are connected (i.e. you have achieved enough coherence) then reviewing the ideas are usually sufficient to remember the connection.

And when you think of it…

“Interconnecting ideas” instead of “linking cards together” highlights the crucial importance of understanding:

You CANNOT connect ideas conceptually at a deeper level if you don’t understand it. The connections would be limited to terms or similarity, but the meaningful connections won’t be there.

Truly understanding an idea peels the layer of “terminology” and bares it down to the fundamental thought.

So what happens is that even if you have different authors referring to the same concept in a different way, you still are able to manage them.

Let me show you what I mean.

Example from my own learning

As you may know, I’ve been trying to apply a Hidden Markov Model in my thesis prototype. Of course, I had to learn the theory before I could apply the HMM tools available online.

I first learned about the three fundamental problems in HMM:

  1. The “evaluation” problem. How probable a sequence of observations O = O1O2O3OT is, given an HMM λ = (A,B,π) = P(O|λ).
  2. The “decoding” problem. Choosing the best statesequence q = q1q2q3qT that best corresponds to an observation sequence O = O1O2O3OT.
  3. The “learning” problem. Adjusting model parameters λ = (A,B,π) to maximize P(O|λ).

To give you perspective on how important understanding is, know that in this example, the “evaluation” problem is simply just some arbitrary term.

The first time I learned it, I called it the “optimality” problem because the goal is to find the optimal model that explains the sequence of observations.

In plain terms, I was not confused by terminology even if it was referred as something else because I understood the thing conceptually rather than perceptually(i.e. in words/terms).

After learning the three problems, I would then learn about Forward Procedure in Hidden Markov Models.

This took a LOT of time and I had to find other sources to help me understand this.

And apparently, the motivating idea behind the Forward Procedure is that solving the evaluation problem is computationally inefficient, and we need to use a “forward variable” to “store” the computations we made.

This is a connection that I made in my memory, with the help of the source. (Good sources are important!)

Now, like implied earlier, I have three options on “what to do with this damn connection”:

  1. Write down the connection in either the Forward Procedure card or the “Three fundamental problems” card (or in both)
  2. Create another card “What is the most computationally efficient solution to the evaluation problem in HMM?”
  3. Do nothing, because I understood this connection anyway hence the “connection” is stored in my memory

I chose Option 3 because I know I can trust my memory to remember this. After all, I have understood the entire thing and I was able to derive the formula myself.

All I needed to do was to create cards for the ingredients (card for Forward procedure, card/s for the 3 problems of hmm) that will help me “reassemble” this connection when needed.

Think “ideas,” not “flashcards” – You cannot automate understanding

So again, if you want to interconnect what’s in your Anki decks, don’t think in terms of “cards” but in terms of “ideas.”

After all, you can only truly connect them after understanding the ideas. This principle applies whatever tool you’re using for learning.

Linking notes in Obsidian or in Anki won’t do crap if you don’t understand or process anything. It happens in your memory and requires the complex processes of encoding, storage, and retrieval.

Let’s keep things simple.

All that being said, for “managing your cards” i.e. updating them, you should be using tags or hierarchical tags at the very least. You don’t need to create a note ID as in the Zettelkasten Method, but you can for the sake of indexing.

Though, I would bet on the possibility that you’ll find it high-friction to maintain such a thing.