Variability, not repetition, is the key to mastery
scotthyoung.com
scotthyoung.com
I learn something new, every day, but what I learn, and how it is implemented, is highly dependent upon the phase my project is in.
Right now, the project I'm working on is in the "home stretch." There's still a lot more work needs doing, but it's fairly predictable, well-practiced stuff.
Getting to this point, though, has involved two years of researching alternative approaches, strategic and tactical designs and decisions, pivots, backpedalling, and bug-fixing.
And every day, I have been writing Swift code. Most of the time, I'm using the same techniques as I was yesterday, but, every now and then, I try something different. Sometimes, that becomes a new habit.
In my personal opinion, there's really nothing like developing and releasing shipping applications to provide a rich experience in both repetition, and variance.
Years ago, I was in a tiny town trying to teach kids to ride horses. It was a volunteer deal with a poorly-organized group. I was a fairly solid rider who had learned horsemanship through a patchy mish-mash of getting handed horses that were easy on beginners, getting handed horses who were way out of my league, buying an animal that was way out of my league (We finally match very well now after owning her for 13 years), and a few dozen lessons spread out over 10 years. I didn't know anything about how to teach children to ride. I honestly still don't know much about kids in general. Thankfully, learning and how to learn is super interesting to me, so I spent many a long drive home trying to figure out what components go into learning horsemanship.
Obviously these kids aren't going to go through my same experiences. They probably won't have a family friend who invites them out to ride a reliable, but energetic, old horse. They won't blunder into a job where they're spending 30 hours a week on horseback like I did. They've got an hour or so once a week to build their skillset. Teaching a kid the cues and posture and how to ride correctly - there's a thousand books for that. Getting them comfortable and to the point where they can handle a weird situation and a horse acting out, there's not much for that. I finally realized that while riding correctly and being held to a high standard by your instructor is very important, jacking around on horseback is important for making it second nature. (On well-broken, very well-behaved horses) try to whack the other riders in the arena with a lead rope. Play pony tag. Make dares, try stupid crap unless I tell you you're going to get yourself dead.
When you're always focused on what you're doing, you miss out on getting that instinctual reaction. I've come to call it fluency. If this happens, then _ . You've got to be able to roll with the punches and the best thing I've worked out for learning that is play.
While I was never pro, I still ride more days than not, and what I've arrived at from it is posture is the effect other people observe, but to your excellent point, only after we have polished a free and fearless foundation that comes from that time in the saddle, where everything just becomes natural. If we try to affect the posture without that foundation, we get heavy german seats and hands and floppy dressage legs. That affected rigid looking seat is about submitting the horse to our aids, whereas a seat that is the effect of the free foundation you can learn more easily as a kid is about releasing the horse's willingness from our need. The posture that is the polished effect of that freedom has more grace, like a dancer or a matador instead of that of a performer or a soldier. The lightness that is the result of it is the effect of a rider's patient posture, which to an observer, looks straight, but they can't see the underlying suppleness of it unless they are at that level themseves. If you ever see any old videos of Nuno Oliviera (the last riding master of the 20th century), I would say that his posture is an expression of this patience, which comes from his first hand knowledge of what is possible, and the horse is responding willingly because it is being met with the confidence of his knowing, and so when it moves, it is following him, but following at his leg, and really, through his body it is reading the intent of his mind. In this sense, the key to mastery is learning patience, and knowing when to be impatient.
Anyway, way to trigger a dressage bro, not a lot of opportunities for that on this site. :)
This is off topic, but...
Six years ago I caught a cold that affected my throat. After four days, I recovered from every symptom except the phlegm in my throat, which has persisted continuously from then to now. It catches my voice and blocks me from singing normally. (And, unless I cough regularly, it builds up over time to the point that my airway can become severely restricted.)
This seems like it would be professionally relevant to a professional singer. Does your wife have any knowledge of such a problem, or how it might be addressed?
I found one mildly locally active antibacterial throat pill that was the only thing that would remove some persistent phlegm like that. I'm not even sure if it was a cure, or just a remedy.
Other alternative: check for allergies.
If you can rock it on a total garbage instrument, it is going to be beautiful on a fine instrument.
"If your collection combines knowledge pertaining to different subject domains, the stream of new ideas and unexpected associations coming to your mind may surprise you" https://help.supermemo.org/wiki/Incremental_learning
After extensive practice with such systems (talking years between physical flaschards, supermemo and anki), I found it leads to shallow understanding and bad linking between knowledge bits, even though I make sure I understand before committing something to spaced repetition. The valuable part looks like to be in the effort of understanding, thinking deeply and widely and summarizing. If you frequently read, practice and revise your knowledge and see how it interlinks, you will approximate spaced repetition at the very least.
All in all flashcard systems taught me how to study effectively by being forced to understand and "atomize" knowledge, so it wasn't useless.
The issue with flashcard style practice is that it doesn't offer much in "application"-learning, which may help explain the shallow understanding. Spaced repetition needs to be repetition of applying the concept, not reviewing it. I know that its not easy to do for all domains, but if you look at musicians (as another poster described), actors, artists, athletes, and martial artists (to borrow the article's reference), their spaced repetition is more about applying their craft to build muscle memory as well as create "a-ha" moments of insight.
I've just started to think of spaced repetition as "personalized daily quizzes" to sort of partially dismantle the mystique built up around the process. Less the self-programming method it's usually written about as than a scheduler, one that gently reminds me "do you still remember that thing you learned? If not, this is what it was."
> create "a-ha" moments of insight.
This happens all the time with spaced repetition, and when it happens, I can make another card. People who do spaced repetition are intentionally trying to trick themselves into doing this when we mix up cards of completely unrelated subjects and do them at the same time. Switching from system to system constantly is supposed to make you see accidental parallels and find lucky insights. That's not strictly spaced repetition as such, but it's something that people have added to it, which shows they're aware.
> The valuable part looks like to be in the effort of understanding, thinking deeply and widely and summarizing.
Which part of this is not conscious effort? You just made a stronger case for such automated systems, especially for people who don't have good memory or abnormal levels of discipline and motivation.
Maybe the paragraph isn't phrased in the clearest way, but I agree with this interpretation, if that is what they meant.
I also observed myself learning non-motor skills better when applying this method in some form. I think most people here know about the yearly Advent of Code challenge. Often people use it to try out a new language and have fun. I believe one of the key reasons for its success (besides the fun) are the frequently similar, but slightly different, problems. Forcing people to approach it from slightly different viewpoints and trying out small variations, thus resulting in a deep understanding and learning effect.
Genetic algorithm: https://en.wikipedia.org/wiki/Genetic_algorithm
Mutation: https://en.wikipedia.org/wiki/Mutation_(genetic_algorithm)
Crossover: https://en.wikipedia.org/wiki/Crossover_(genetic_algorithm)
Selection: https://en.wikipedia.org/wiki/Selection_(genetic_algorithm)
...
AlphaZero / MuZero: https://en.wikipedia.org/wiki/MuZero :
> MuZero was trained via self-play, with no access to rules, opening books, or endgame tablebases.
Self-play algorithms essentially mutate and select according to the game rules. For a generally-defined mastery objective, are there subjective and/or objective game rules, and is there a distance metric for ranking candidate solutions?
For example see [1]
[1] https://best-essays-service.org/essays/health/the-effect-of-...
It is like seeing a 3 dimensional object from multiple viewpoint. Some things change some stay the same. You understand and thus remember the thing better because you know it's "essence", what is the same about it always.
In a nutshell, we knock that CS1 courses have a high number of drop/fail/withdrawals due to several different factors. These include course difficulty, time management skills, and not feeling welcomed in the community (lack of representation, assuming CS students must love video games, etc). I don't focus much in inclusivity (though a number of faculty at NC State do), but I'm looking for methods to reduce difficulty. My thesis stems from methodologies I've used while teaching martial arts - specifically focusing on lower-level "build blocks" practice.
We can represent different learning activities by the amount of "creation" a student must engage in - Passive (watching lectures) require no creation, Active is repeating solution steps, Construction is critiquing steps, and Interactive is a co-creation process between the student and another student, instructor, or system. These form the ICAP framework by Mickie Chi. My arguments are that if a student is struggling with an Interactive exercise, like a traditional coding problem, then they should "downgrade" their practice to something lower level like self-explanation or debugging. However, if THOSE activities are still difficult, we can downgrade to an even lower-level of having them simply repeat (the mindless 'copy the dictionary' activities modern education hates).
My rationale again looks at how technical skills, like martial arts, are learned. Typically we start with warm ups, a brief demonstration on a few movements from the instructor, and then students are asked to pair up and drill the moves. Then the instructor builds on the moves with another demonstration, and then students repeat their process. Eventually, students spar (or apply the technique in a live problem). I argue that traditional coding problems are analogous to live sparring because students need to incorporate writing code, problem solving, debugging, code tracing, etc. If they struggle with some of those skills, they struggle in the activity. Rather than having them learn all the skills "in situ" while coding, maybe lower level deliberate practice targeting a single skill can help strengthen their overall foundation.
This has been my experience with teaching math and CS.
The biggest causes of failure is not identifying and teaching prerequisite skills.
The key point is prerequisite. Many times the skills needed to pass a math or CS class are not part of that class's curriculum.
Students who have encountered the building blocks already tend to do well, students who have not, tend to fail - no matter how much the class goes over the material in this class - because the problem lies before that material.
Calculus is mostly algebra. If you can’t reliably solve algebraic equations, derivatives and integrals will probably be a step too far.
Algebra is mostly arithmetic. If you can’t reliably add and multiply rational numbers, solving for unknowns is going to be patchy at best.
The key part as a student is learning to recognise when I needed to re-tread the primitive operations. Having to go back and practice the mental arithmetic I first learned in primary school was a hit to the ego, but at least I could recognise and take that step. When I was in my late teens / early 20s I was much less able to swallow my pride and back-track.
This is, in my opinion, one of the issues with adults attempting to pick up new skills, in addition to that mythical "free time". Many of us have been conditioned that if we aren't masters in something immediately, its a lost cause. If someone suggests that we do "what kids do", then we think we're too "grown up for that". A good example would be tracing drawings. Many kids learn this way, but to suggest an adult do it and provide positive reinforcement in doing it is almost non-existent at the adult level.
Honestly, an adult version of Sesame Street (not "adult" in sex, drugs, violence, but the simplistic dialogue with repetition) may help adults learn a new language, but its competing with other factors like time, general attention, thinking this is "worthy" of our time, etc.
I disagree; both programming and martial arts require a level of technical motor control. Its easier to define and connect in a physical skill since that is where motor control is studied. However, programming as a technical skill involves familiarity with syntax, familiarity with the tools (literally keyboard and mouse), and translating plans into valid syntax. Likewise, sparring in martial art is as much a problem solving skill as it is in programming - you need to evaluate the problem at hand and do problem decomposition to create subgoals until you produce an actionable item.
> There's no reason for them to be connected
The connection to martial arts works for me because that's something I've done for over a decade. It doesn't immediate click with everyone if they don't have that same history with it. However, many people have seen parallels when I describe this through learning music, dance, chess, or second language acquisition. They recognize the skill was a combination of lower level technical knowledge and higher level problem-solving/creativity. My base thesis is "if you're struggling, why not focus on refining the lower level skill you seem to be having trouble with"?
I saw a documentary, long ago, about primary school students in China with abacuses. They used the abacuses for mathematics, and they were blindingly fast with them.
And then they stopped using the physical abacuses and just sort of mimed the actions with their hands. And then, later, they stopped using their hands.
So, as far as I could see, they were training themselves to do maths via their muscle memory, and it seemed terrifyingly effective.
At the very least, on the second one I was a little bit older :)
To me It’s not variability, it’s repetition plus attention (which is much easier to get if you enter The Zone)
I'd argue that it's the same for non-physical things as well.
Yes, Bruce Lee must be wrong, because obviously it has to be one end of the spectrum or the other /s. How about the man who has practiced 100 kicks 100 times each? Maybe he’s the one to fear.
If he practiced it 10,000 times in the exact same spot on a punching bag he wouldn't have the same understanding of it that he would if he practiced it high, low, with the other leg, in a tournament bout, on the sand, eyes closed, after a punch, etc.
Of all people, Lee would never say to practice 10,000 times in exactly the same way ... and definitely wouldn't suggest doing so 10,000 times on a punching bag, seeing as Bags Don't Hit Back [0].
Lee obviously meant "the man who has practiced one kick 10,000 times" in varying conditions, with varying timing, energy, and motion ... not in a single dead pattern.
An easy comparison is take a driver who's had three months of experience driving versus somebody with three years, they'll be a huge gap in proficiency. Now take the three-year driver and compare them against someone who's been driving for three decades, you're fine surprisingly there's very little difference in ability.
Our brains are ruthlessly efficient and the moment that they can optimize away learning, that's when you're no longer acquiring skill.
IMO this explains a great deal about current society stuck on the idea re-training is a waste, the habit of re-electing politicians for decades being one outcome of living life “on the career escalator.”
Such an inner monologue becomes a default state of being.
Accepting simple memes like “will work for money” become the norm and “will work to acquire knowledge” becomes vulgar language.