Comment by travisjungroth | Hacker News Reader

Comment by travisjungroth | Hacker News Reader

I figured out something very similar in flight school. When I did my instrument rating, I realized my instructor, Sean, wasn’t perfect. No pilot is. But, I believed he could pass an instrument check ride. I could not.

I decided that rather than figure out if everything he did was the best way or “worked for me”, I would just do my best Sean impression. I still had the opportunity to get even better by learning from other instructors or coming up with my own ideas. That could wait until after I got my rating.

He actually noticed. At the beginning of an early lesson he showed me his check routine during taxi. He flowed through all of the instruments and controls. The next flight, I did my best impression. It wasn’t perfect, but it was close. When I was done he told me no student had ever gone through the whole thing the next flight. Most don’t even try. Even the better students just grab a few things and end up with a routine close to his by the end of training.

It all ended up working. I passed my check ride with just about the legal minimum amount of training time.

Writing this all out and talking about what a great student I was feels pretty egotistical. Ironically, that’s the opposite of what this method is about. Just copy the closest person next to you who is better than you. They don’t even have to be that great. Then, find someone better, and copy them. Save breaking new ground for later in your journey.

Hmm. I think I have a similar story. I was a competitive ballroom dancer, and at one point, a fellow dancer was trying to teach me to have better frame. (Frame is some mix of good posture, being rigid but not stiff; the follower in a ballroom dance uses it to know where to go, and to know what the leader of the dance wants to do. Of course.) I don't think I really got it until at some point he led me (to demonstrate his frame, and how it should feel) and that was what really made it click for me. In an instant I was a better dancer, but I don't think words alone would have gotten us there. Imitating him made me better. (And he had to show me how to lead by making me follow him, and I definitely still have no real idea how to follow, so it was doubly impressive that he got the point across.)

And frame in general is one of those things that when instructors teach, is taught not only through verbal description, but through actual interaction with a person body. (E.g., we pair dancers off, and have them "move" each other using their frame.) And I don't think it could be adequately conveyed with only words, or, it's definitely much easier to demonstrate by physical contact. (And it's one of those things that instructors will critique from the very beginning until … well, if there is a point where one's frame is perfect, I've not hit it.)

The article also reminds me of when I was a TA for Computer Science. Some students need help, in the sense that they need some concept explained better, or they don't know what function or type to apply or don't know a particular data structure. But some students need help, in that they seem to lack something extremely fundamental; the closest I've come to it is that they lack the ability to establish a mental model of a program and how a computer might execute it that is coherent and consistent. (That is, they'll have some rules, but they're more like patches and bits of understanding and are sometimes even self-contradictory, because they the student are closer to grasping at straws than having some idea of what is actually going on.) Getting the student past that hurdle is something I struggled with as a TA. (I myself either picked it up so long ago I take it for granted, or it came naturally. But neither does any good when trying to teach it.) I feel like at some point I read a research paper that seemed to establish that this divide between CS students was the lack of a good coherent mental model, but I've since lost any reference to it.

You're referring to this: https://blog.codinghorror.com/separating-programming-sheep-f...

I was also a TA for CS, for three years, and when you're in that situation and encounter so many students in such rapid succession, the existence of a double-hump becomes really obvious. It's just that because these courses are self-selected, the second hump has a small number of people.

Reading through the linked retraction at the top, it appears it was only retracted for the controversy and claims outside the original findings, not because what it found was false. So it's really frustrating that people use this as a claim there's no such thing.

I can definitely imagine the cause being something like tacit knowledge that we just don't know how to teach, though.

I think you might be right, and interesting, I did not know it had been retracted. Is the retraction trying to retract some claim that people who lack a consistent model are incapable of programming, or something more? (As my original read of the paper was more that the lack of a consistent model was the cause behind poor performance, but not an unsolvable problem, and more of a great hint as to how we could or should approach teaching CS. However, while TA'ing, I did occasionally notice such a behavior, and tried to point it out to the student, and that in and of itself turns out to be harder and trickier than one thinks to communicate, like trying to describe the color blue to someone who hasn't seen it. And, of course, TA'ing is hemmed in by the real-world constraints of too many students, too little time.

> It's just that because these courses are self-selected, the second hump has a small number of people.

By second hump, you mean the weaker group (the "inconsistent" group, from the papaer)? At my college, the early CS classes — which was where I spent most of my time as a TA — were also taken by some students of non-CS majors, as some majors required two semesters of CS. (These were primarily major whereby one might need to say, write a computer model to help validate/invalidate something more related to one's field.) So, they weren't all their completely voluntarily, I suppose?

But we definitely had some self-selected students, those who had clearly been self-teaching themselves computers and programming from early in their teenage years. However, our college also permitted one to voluntarily skip one or both of the first two semesters, if one felt one was strong enough to do so. (There was a grace period at the start of the semester whereby you could still transition between the classes, if you found you had bit off more than you could chew.) So, those clearly strong students who had already been independently studying in the field prior to university could skip over some of the courses.

I like your choice of phrasing on “impression”. I think it captures what is meant by embodiment, that the knowledge is in the body, so you can get at it by imitating the whole of a person.

Thanks. I think it’s much more effective to copy too much rather than too little.

Some people who learn to speak foreign languages fast seem to do this. They copy a native speaker exaggeratedly, almost to the point of parody. And of course young children learn in the same way, by imitating everything rather than trying to filter for the "relevant" bits and imitating those. Most of us are too self-conscious to do this properly though.

lol this is me with Spanish. I say common phrases with all the details right, so people think I know more than I do. I actually speak better than I understand.

Ah, interesting. Back in my presentation days, I found that, if I pretended to be a fire and brimstone preacher when I practiced speeches, and dialed it back when I actually gave the speech, those speeches went a lot better, and were better received.

I'm really good at accents because I imitate exactly. People think I'm a native when I say hello then I can't understand anything after that because I'm really bad at remembering vocabulary.

It's kind of scaring me off moving to another country to learn the language because I struggle to remember words in English. At the same time I know I would master the words and phrases I do learn plus I've always wanted to try being immersed in another culture. Terrified of not being understood I guess.

A big problem is that novices don’t know which parts need to be copied and sometimes don’t even notice what they are missing, so without either (a) getting a decent amount of expert feedback or (b) extensively reasoning things out from first principles and then doing a lot of trial and error, it’s easy to copy extraneous elements or skip essential ones.

A lot of methods and devices that we end up stuck with today in various contexts are just watered down copies of things that were once great, because all of the superficial features were slavishly copied by multiple generations of people who didn’t understand the original constraints or think about which ones have remained constant and which have changed.

If you go from non-functional to functional with extraneous crap that’s a success. Unless you’re working at the forefront of research or at the absolute limits of your capabilities an explicit mental model of what you’re doing is of surprisingly little use. Even that exaggerates how useful it is. The huge majority of improvement is by trial and error, with shoddy and incomplete mental models.

Your last paragraph is just what happens when there’s no force demanding quality. If there’s not strong pressure to maintain or improve quality decline is inevitable.

There’s a really wonderful book called, “The Inner Game of Tennis”, and both your story and commoncog’s essays on pattern matching remind me of it.

This is how small children learn. They copy first, understand later, sometimes very much later.

Not only that - they prefer to copy slightly older kids (closer to them in ability) rather than adults (who are doing things they have no change of reproducing).

You never become exceptional this way. And if you run into situations you’ve never seen before, it’s harder to react competently. With coding, if you only ever use a style you copy, then everything looks like it could be solved in that way. This works and may help you up the corporate ladder. But it will never produce massively positive results

If you copy enough styles then you will eventually see the connections between the styles and come up with your own. This is exactly how you become exceptional, by copying the best over and over again in as many styles as possible until you 'get it'. You have to think about it and understand it too but that comes naturally to anyone who is capable of becoming exceptional, that's the seeing the connections part.

As a description of the limits of this strategy, your assessment is probably accurate. This is a reasonable strategy for developing common competence. Whether that means you should adopt it probably depends on where on the relevant competency curves you are -- or how uncommon the competence around you is.

That may or may not be the case, but note that there are many (most?) domains where the positive payoff is bounded, but the negative payoff is not. This method seems particularly useful for those scenarios -- if you have "adequate" performers to copy from.

Instrument flying is actually a great example of this. There are no instrument flying competitions. Well, I guess there’s one every flight. First place, you make it there. Second place, you diverted. Third place, you crash and probably die.

I wonder if the OP post has more to do with nerves/confidence than learning. By imagining he was Sean, he felt confident and the naturally performed better at the job at hand.

To answer your point, there are times in the "learning curve" you need to get used to doing something, and then times to become exceptional. Compare learning to drive to formula 1. And commongcog has an article for that: https://commoncog.com/blog/get-numb-get-good/

Yeah, I think there are issues that fall away when you stop trying to be good and just embrace pretending to be good. Confidence is one of those. I also think a lot about evaluative mindsets versus learning mindsets. Copying puts me in learning mode.

Counterpoint: the better you get at copying, the faster you accelerate your ascent towards being exceptional. Copying is generally the beginning, not the end. By this anti-copying logic, practicing scales on an instrument shouldn't have any bearing on playing a complicated piece of music, only playing that piece, right? And yet, practicing scales until they're perfect and mastered pays dividends because the closer they get to second nature the more firmly they can serve as a foundation for second order skills.

Learning how to copy things is an art in and of itself. A lot of engineers (myself included) got into the craft by taking things apart and putting them back together again. It's true that you cannot become a good engineer /only/ by doing this, as at some point, you must learn how to create things from scratch that you've never taken apart and never put together before. It's true that teaches you a lot, but I think it's a little bit of a stretch to imply that it's the only thing that can teach you.

This method won’t get you exceptional on its own. That’s for sure.

I’m not saying to pick one person and blindly follow them forever. It’s just a really fast way to get competent. I used the same technique to learn programming, scuba diving and sales. After you get the super basics down, find more teachers to copy and start even thinking for yourself a bit more.

It will take you surprisingly far, though. One thing I didn’t mention is it’s good for getting the most out of people who are master practitioners but not great teachers. Rather than ask them to take their skills and package them up for you, you just look over and make a copy for yourself.

Reply on news.ycombinator.com