Skip to content

Chunking, Practice, and Recall

These notes are my reinterpretation of the material of week 2 (of 4 weeks) of the online course "Learning How to Learn" on Coursera by Dr. Barabra Oakley.

TODOs

There are a few in the document.

Important TODO: "Maintaining a Library of Chunks

Topics

  • What is a chunk?
  • How to form a chunk?
  • Pitfalls
  • Extending the analogy
  • A "Library of Chunks"
  • Additional context

What is a chunk?

Human brains have 4 slots of working memory, you can think of your reasoning as 4 armed octopus able to touch only 4 remembered items at a time.

A chunk is a collection of memories that form a concept. For example, the snick of a pop bottle opening connects to the word "pop" and the letters "p" "o" "p". One is a qualia, the other is a logical construction through written phoenetic glyphs. You can also sketch a can of pop as a visual qualia for this chunk.

A chunk can be very large or small. The course says a chunk is like compression but this anaology is flawed, because a chunk itself is more like a nexus of memories and reasonings.

A chunk is like a summary with hyperlinks to related chunks that might be larger or smaller in size or relationship. A chunk might be tightly related to some chunks and loosely related to other chunks.

"Conceptual chunk": Mental leap that unites scattered bits through memory.

Trent: There is a concept of scattered bits of memory that are not worth detailing as their own chunk in your context of study. It may or may not be useful to focus in on any of these scattered bits as chunks themselves. For example, in the "pop" chunk, the letter "p" is a scattered bit, but it has its own chunk with a lot of context, like the vocalization of the letter, it's various forms, "P", 'p", its history through the history of language. But, you probably do not wish to develop that chunk when focusing on a different topic, e.g. the topic at hand, "learning what a chunk is".

Every discipline is a little different for memories. Imagine: walking, doing a math problem, and telling a story.

Synthesizing a collection of chunks and creatively integrating two different chunks in a new context are related to the idea of "expertise".

Transfer: A chunk in one discipline might share commonalities with a chunk in a different discipline.

Trent's Thought Experiment about connecting unrelated chunks

This is kind of like mad libs.

Thought experiment: How are my grandmother's favorite shoes related to household accounting practices? I know my grandmother is good at accounting, and this starts to fire a lot of memories about how she learned these skills since she did not have much education. This road of obtuse reasoning is long. Thought experiments about unrelated chunks can yield interesting connections.

Expertise and Creative Integration

Expertise is chunking and performing creative interpretations as you gain fluency.

How to form a chunk?

Perform these steps when practicing or studying material.

1. "Undivided Attention"

  • Turn off all notifications
  • Any single interruption "uproots" some neural connections being made (weaker foundation)

2. "Understand"

  • Get the "main idea" or "gist": "top down learning"
  • "Superglue" the pieces together
  • "Broad traces" and "link to other traces you have"
    • Finding links to other knowledge will stimulate dopamine reward
  • Do a picture walk through a chapter before reading it

3. "Gain Context" - "deliberate practice", "recall", "review", "when to use the chunk"

  • Perform: "deliberate practice":
    • Process: work on primarily difficult problems
      • TODO: How do we identify "difficult problems" in each of the below contexts?
    • problem sets (math, sciences, computer science, etc.)
    • practical experience (medicine, carpentry)
    • physical techniques (judo, dancing)
    • songs
    • foreign language conversations
    • ... (what else?)
  • Perform: "recall"
    • Process:
    • Remember everything you can with only your brain: simply use your mind, or use a blank docoument
    • Review your notes against the material, identifying gaps
    • Mediums:
    • Mind: Simply recall what you can, then recall what you can about each item
      • Not easy to validate against the notes later, maybe has other benefits?
    • Physical: Outline, Diagrams, Notes, Sketches, Paragraphs (paper or digital)
      • Easy to validate your progress/success
      • Can create a more organized permanent reference
    • Recall in a different place to decouple your access to a chunk from where you learned it
    • Generation Effect: "The generation effect is a phenomenon where information is better remembered if it is generated from one's own mind rather than simply read"
    • On the first recall, you will prime the generation effect, creating "neural hooks"
    • On the second recall, you will take advantage of the generation effect.
    • Wikipedia: Generation Effect
    • Trent: probably can do this 1x before needing to dip back into "diffuse mode" (workout, sleep, socialize)
  • Perform: "review"
    • Perform more "deliberate practice" and "recall" with previous related chunks.
    • New relationships with the old material and the new material will be stronger
    • Review the material both "forwards" and "backwards"
  • Learn WHEN to use the "chunk"
    • Analogy: when to use a hammer vs a mallet
  • "Do a picture walk through a chapter before reading it"
    • Know how the "pieces of the puzzle" fit together (knowledge inside the chunk)

4. "Repair the Flaws"

  • Process: Identify your misunderstandings and repeat "deliberate proactice" with the right ideas
  • This is part of the same study methods above
  • It's important enough here to call out as a step

5. "Interleaving during Practice"

  • Mix up your learning within a subject or even between multiple subjects
  • Find other contexts that have this chunk or a similar chunk
    • In a textbook, read through nearby problem sets to find other applications that contain the problem
  • Trent: Time intensive? How to get a library of interleaved problems quickly?

Example: Learning a guitar song

I like this analogy because it is easy to "test" if the song sounds right.

  1. "Understand"
  2. Listen to the song
  3. "Practice"
  4. Grasp bits of the song
  5. Review chords and notes and be able to play them
  6. Join passages until you know all the parts
  7. Try the whole song
  8. keep improving

Pitfalls

  • Working memory doesn't work as well under stressors: low sleep, stress, sickness, etc.
  • Rereading material once you have context
  • Instead use "recall" then compare the material to your recall results
  • Research supports that this is a comparatively better method for learning
    • Compared to some other methods: concept maps, notes, rereading
    • Probably still want to take notes, that's done during studying and can be used for recall comparison
  • Overlearning
  • Repeating problems you already learned
  • Usually a waste of time, instead try to "cover a lot of ground"
    • Instead: Once you get more context with more chunks, then repeat your old problems
  • It is useful if you need "the perfect tennis serve"
  • Can be used positively to combat "text anxiety" but will cost time (tradeoff)
  • Overtraining for perfect public speaking: "A 20 minute TED Talk can take 70 hours of practice"
  • Einstellung
  • Stuck in a way of "focused mode" thinking you already have
  • Try taking a break, go into "diffuse mode" and reapproach the problem fresh later

A "Library of Chunks"

  • To develop expertise in a discipline, build two things:
  • "a bigger chunk library"
  • "a more well-practiced chunk library".

Extending the analogy

Consider a chunk as a ribbon of learning. The ribbon will grow wider and longer as you improve at making chunks. You might listen to the "whispers of the diffuse mode" that will forge a new connection between two distance places on the ribbon.

Problem: Too much ground to cover!

There are so many chunks in a discipline, how will you get to them all?

"Law of Serendipity": Lady luck favors the one who tries.

Solution: As you form a chunk, nearby chunks will be easier to form. A bigger "library of chunks" will naturally grow.

But you must practice each chunk to make it "wider" which will make later practice cover more ground. Later practice will not get easier, but it will cover more ground at the same difficulty.

Practicing related chunks will help you begin to classify "types of chunks" and "classes of concepts". I think these are probably also considred chunks. If you are well studied, you are probably already familiar with a library of "types of problems".

TODO: Maintaining your "Library of Chunks"

How can I stay in practice? I might have a group of chunks that I want to have front-of-mind.

One shortcut would be to carefully maintain the library of problems I used to develop that chunk and periodically repeat solving those problems.

Other parts of the founding material of chunks: primary readings, primary practices, related problems, related readings.

How would a graph representation work for managing the inputs to developing a group of related chunks representing a discipline?

Additional context

  • "Sequential Reasoning" are focused mode reasonings
  • "Holistic Intution" are diffuse mode ideas
  • You MUST carefully verify "holistic intuition" with "sequential reasoning"

Trent: Even modern physics has 2 incompatible models, I have enocuntered many people that try to use deductive logic on their mental models. Mental models are generally not sufficient to form a closed system. Deductive reasoning needs to be backed up with inductive facts if possible, otherwise it is generally too leaky and insufficient, especially in highly technical black box systems.

Neurochemistry

  • acetylcholine
  • form new long term memory, synaptic plasticity, "projects widely"
  • TODO Question: Does this mean this chemical helps form broader connections between topics?
  • dopamine
  • decision making, which sensory input your brain values (perception),
  • released after unexpected reward (unexpected connection between two topics (chunk linking)
  • serotonin
  • risk taking, depression, loneliness, low perceived social rank causes high risk taking
  • prozac is a prescription to increase serotonin

Neurochemistry: how to control it for "higher quality learning"

  • basic shared controls:
  • sleep
  • nutrition
  • exercise
  • acetylcholine
  • TODO ??
  • dopamine
  • "understand" connections to existing knowledge to release dopamine
  • Celebrate: reward yourself after learning - releases dopamine
    • mental pat on the back
    • vocalized verbal celebration "Yeah!"
  • serotonin
  • maintain a healthy social structure that fits your needs
  • focus on mental wellness and mental health
    • therapy, medicine, sleep, exercise, nutrition