Insight

[liv]

I was totally going to post about something entirely different, and then I saw an essay linked from Making Light which blew my mind, so I'm talking about that instead.

When asked for advice a while back, siderea came up with:

Never identify with your virtue or lack thereof. That way lies compulsive rigidity and painful blindspots that lead one to walk off cliffs. Should you think of yourself as a "bad person", the problem is not in failing to think of yourself as a "good person", its in trying to categorize people as good or bad.

Which I thought was pretty sound and elegantly phrased.

This oldish LJ post by celandine13 points out a similar fallacy in thinking about competence, as well as virtue. Delightfully titled: Errors vs. Bugs and the End of Stupidity, I do recommend it. I definitely like the idea of a debugging metaphor for learning a skill, because it gets away from the talent vs practice frame. I mean, people can argue endlessly about whether something is a matter of innate talent or whether it's a craft that you have to work at, but both of those risk invisibly assuming that "good at" a skill or not is something that you are, part of your identity.

I like that celandine13 provides concrete examples from music practice, because part of the problem is that simply determining to work at something isn't very useful unless you have a concrete idea of what work consists of. And yes, a debugging mindset may well be more productive than just doing the same wrong thing repeatedly. For me personally, learning to play the piano moderately badly was one of the most useful parts of my education, because it taught me how to learn, how to build a skill that didn't at all come naturally to me. Which was extremely valuable when I started trying to do science at a level where it's actually hard, even when I'm generally successful at absorbing and retaining complex information in an academic setting.

I also find her thoughts specifically about education and training highly relevant, and likely to be very useful to me as a teacher. It ties into the stuff about praise I learned when I was trying to take that pedagogy course: praising learners for being good at something doesn't help them improve and can even make them worse in the long term, because their sense of self is challenged every time they find something difficult or perform less than perfectly. Educationalists such as Dweck argue that you should praise learners for the effort they put into succeeding at something, so that their achievement becomes the thing they did rather than who they are.

But then the second half of the post, bam, it broadens out into talking pretty much about life. It's not a matter of being less stupid or less lazy, it's a matter of debugging the process that is causing you to make errors and not grasp a subject or excel at a skill. Circling back to siderea's good advice, I am very seduced by the idea that virtue is also a skill or a practice. (I got a very positive response when I mentioned casting happiness as work, too.) And as such, it makes sense to debug the underlying issues that cause you to sometimes act against your moral code, rather than just working very hard at being a better person.

Comments

[jack]

I love "Errors vs. Bugs and the End of Stupidity". It ties into a lot of thoughts I've had about linking software engineering principles to/from the real world (like the unfinished dome). In fact, I'm going to add it to my list of "essays I never get tired of linking to".

[liv]

Yay! I suspect I wouldn't entirely see eye-to-eye with celandine13 on all political issues, but this essay seems to be a really useful way of explaining something important.

[jack]

Oh? I don't know anything else about her. I followed her journal for a while after that essay because it was so awesome, but there was nothing else public.

[liv]

Oh, almost her entire friends list is Less Wrong people, and her interests list is a lot of code words people use for "I'm cleverer than everyone else", including Libertarianism. Could be wrong, it's superficial impressions because as you say she doesn't post much public.

[simont]

The errors vs bugs thing is a nice way of looking at it, and definitely in line with my own approach to learning. (Partly that's because debugging is a metaphor I'm very comfortable with because I do the thing itself so much, of course! But also it actually is how I learn a lot of things – I've always gravitated to things I can largely teach myself, and that process has a lot to do with how I do it.)

For my money the interesting question is what you do when that approach stops providing useful results. For example, some bugs have the property that you work out quite quickly what the cause of the failure is, but can't quite see what you can do to prevent it – e.g. because the part of the system causing the failure is not readily modifiable, or because every change you consider making turns out to break in some other situation (what I think of as 'wallpaper-bubble syndrome' – you can move the problem around but not get rid of it completely). And my experience of self-tuition is that some learning issues run up against similar barriers, and it would be nice to learn some (meta-)techniques for getting past those.

[jack]

Ah! That makes sense, I can see some less-wrong-y clues now you point it out, but I didn't notice because that's probably a lot closer to me than you.

[liv]

That seems a really useful question, and thanks for posing it. What do you do when you can't debug something? I suppose at that point you've narrowed it down and you have a pretty good idea why your bug can't be fixed, because it impacts something outside your control, or because fixing it causes wallpaper bubbles (lovely image!) So you're in a pretty good position to go and seek external advice; you may ask an expert, not just "help I can't do this" but "this problem is preventing me from progressing, has anyone ever solved it before?" (The expert could be a teacher or could be written resources, in the self-teaching paradigm.)

I mean, sometimes it might actually be an unfixable problem. I might be able to train myself to move my fingers more nimbly, or work out an alternative fingering that I can actually do, or even simplify a piece so that I can play something which gives the same general musical idea even if I can't do that twiddly bit. But I probably can't get myself to the level where people would pay money to hear me playing the piece not just correctly but well.

[simont]

*nods* A good point – 'find a teacher' is an obvious meta-technique to try when one's self-debugging seems to have hit a brick wall!

In fact, circling back to the original subject of piano-playing, I actually did that once as a child. Piano was never an instrument I formally or 'properly' learned (that was violin), but I had learned on my own to noodle around on electronic keyboards at home, largely playing stuff I'd composed myself. Then one day I wrote something so complicated that I wasn't able to play it personally, and only the computer could do it. But it didn't look quite impossible, in that my fingers were not actually too short to reach all the notes at once in principle. So I took it to my school's piano teacher and said 'Hi, I've never had a piano lesson but I can play simple stuff and I'd like to be able to play this, can you help?'. She was entirely unfazed by that unorthodox approach from a pupil she'd never interacted with before, and we had a one-off piano lesson in which we successfully worked out a fingering :-)

[naath]

I think some things are more amenable to the "just keep trying, you'll get better" than others. The muscle memory to play a piece perfectly *every time* once you've figure it out comes only from doing it *over and over again*; the fitness to run a marathon can only come if you *spend lots of time running*... of course you may also have bugs that need fixing like "has entirely wrong fingering" or "is trying to run without drinking enough water".

I also disagree with the idea that there is *no* innate component. For instance I play the violin extremely badly - why? because I do not have sufficiently good relative pitch to stay in tune and everyone expects violinists to be able to do that. I think it might be possible to play perfectly through muscle memory alone but that would have required spending huge amounts of time practicing with someone standing there saying "sharp" or "flat" every few seconds and that's a lot of work (for someone else as well as me).

I do agree that finding the bugs is a much better way to help students progress than simply saying "work harder" as if repeatedly being told "no, that's wrong" with no explanation is a helpful way of learning.

[doseybat]

Wow, convincingly argued essay by celandine13. Will link to it from faceborg - let me know if it is OK to tag you which I assume it is not. In fact I should probably not be posting this comment to avoid account connection: feel free to delete!

[siderea]

Ooh, this is some of what I clicked through to say.

I very much liked the essay. It's written from a perspective where "just trying harder next time" is the default attitude/behavior, and debugging the problems is rare. But/and I too come from a music background and a programming background, and I offer in the spirit of thesis-antithesis-synthesis that one of the powerful lessons I learned from studying music is that not all failures of performance are remedied by debugging.

I come from an intellectual tradition which treats every error as a problem to be solved. Don't get me wrong: I think this is a fabulous way to approach life and that most people in the world would be much improved by adopting it. However, among the very cerebral who are my people, the idea that there are problems that cannot be solved by comprehension, but only by repetition, is a very necessary and powerful one.

Years ago I had an apprentice. He was an undergrad at MIT; had been a CS major, and then switched to Philosophy. So I had him to take up the study of an instrument, and based on his musical interests, I recommended harp. It was powerfully edifying. No amount of understanding will build the muscles in your hands. No amount of understanding will give you the sense of timing necessary to move your hand in advance of needing to move your finger. No amount of understanding will increase your physical agility. No amount of understanding will give you the muscle memory to execute complicated passages at speed. These things you can only get by putting your hands on the instrument and doing.

Also, the experience of going from notated page to performance is humbling in what it teaches one about the gaps between what one has understood and how much there is to understand and be capable of, to perform. It shows us how we kid ourselves: we can look at the page and say, "Sure I get it," but then we play it and betray how far we are from rendering the piece wholly. "The difference," I reminded my apprentice, "between theory and practice, is that in theory, there is no difference, and in practice, there is."

I do agree that finding the bugs is a much better way to help students progress than simply saying "work harder" as if repeatedly being told "no, that's wrong" with no explanation is a helpful way of learning.

Indeed. And I find fascinating that apparently, in reality, when people say of themselves, "I'll try harder next time," they don't seem to mean anything other than "I'll feel more anxious about it next time." "Try harder" doesn't actually seem to have any meaningful referent. It doesn't actually refer to any specific behavioral or attitudinal difference.

In the early music ensemble I lead for many years, sometimes a player would be struggling with a passage, and we'd kick it around in rehearsal, and eventually they'd say, "At this point, I think I just need to go practice this." And we'd drop it. And they'd go home and practice it -- possibly with debugging -- and then bring it back ready to go.

It would be really neat if that's what people meant by "try harder": practice more. But it's not. Not usually. So not only are they not debugging, they're not practicing more, either. That's the worst of both worlds, and not going to lead to much progress.

For instance I play the violin extremely badly - why? because I do not have sufficiently good relative pitch to stay in tune and everyone expects violinists to be able to do that. I think it might be possible to play perfectly through muscle memory alone but that would have required spending huge amounts of time practicing with someone standing there saying "sharp" or "flat" every few seconds and that's a lot of work (for someone else as well as me).

(Psst. I understand they have machines to do that now, actually. Also -- speaking as someone who started with a profound pitch deficit (on the order of not being able to differentiate semitones) and is now a sight singer -- there are exercises which build tuning skills. One of my favorites, and the simplest, is to practice against a drone pitch. I found a constant pitch generator on the web, but back in the day, I'd make do vocalizing against any electromechanical device with a hum -- computers, the fridge, my bathroom fan, electric subway cars, you name it. You can practice matching pitches, matching specific intervals, improvising, or rehearsing specific pieces. Ping me if you want more info on this.)

[siderea]

For my money the interesting question is what you do when that approach stops providing useful results. For example, some bugs have the property that you work out quite quickly what the cause of the failure is, but can't quite see what you can do to prevent it – e.g. because the part of the system causing the failure is not readily modifiable, or because every change you consider making turns out to break in some other situation (what I think of as 'wallpaper-bubble syndrome' – you can move the problem around but not get rid of it completely). And my experience of self-tuition is that some learning issues run up against similar barriers, and it would be nice to learn some (meta-)techniques for getting past those.

The aforementioned get a teacher is one. So is consult with peers.

Another is, try explaining the problem to someone. I know someone who was locally famous at the Media Lab for being a good person to hash problems out with; story has it that one day, somebody came to his office door, knocked, said, "I have a problem I'd like to disc--" then got an intense look on their face, stood there with furrowed brow for some long minutes, and then announced, "Thank you! That was very helpful!" and wandered off without ever having actually shared what the problem he was having was. Simply anticipating what the problem-hasher-out would ask and running through the conversation in his mind was adequate to get him unstuck.

Then there's have a cup of tea, and its close friend shelve it and come back to it. Your subconscious may continue to gnaw on it, and throw out a solution when you're least expecting it. You may encounter an account of a solution. In solving other problems, you may learn more about solving that sort of problem so that you can return to it with more clues.

List your assumptions about the constraints. Test them all methodically. If every change you consider making turns out to break in some other situation, then maybe there's some change you haven't considered.

Break a working one apart to see how it works. A favorite of mine, in code and music. In learning the craft of composition, nothing beats analyzing your favorite works.

And, like I wrote below, Nothing substitutes for practice. Practice is for solving different problems than debugging, and they're not great substitutions for one another.

[jack]

I come from an intellectual tradition which treats every error as a problem to be solved ... this is a fabulous way to approach life ... However ... there are problems that cannot be solved by comprehension, but only by repetition

Yeah, this very much so. At school I hated sheets of identical problems because they were normally trivial, so it was an exercise in "can I do something boring twenty times without making a mistake" which I found hard to motivate myself for. And then by the time I got to calculus which I _didn't_ know in advance, I didn't realise that being able to do one problem, I still needed to practice to be able to do the problems *easily* which was necessary later on when those problems because just a small building block of larger problems.

But if we approach problems in a debugging mindset, I hope we won't necessarily focus on *just* that, we'll be open to the idea that problems may need a specific solution, not just "try more", and hope that once you're thinking like that, it'll be obvious when the answer is "practice" as much as when the answer isn't "practice".

[liv]

Part of me is a little sad whenever I see a comment like, this is good, I'll put it on FB. But anyway, yes, I'd really prefer you don't tag me. I don't think you need to delete your comment; I'm not trying to hide this journal from real people, just from algorithms.

[naath]

I think I may also have a deficit in the "being willing to be repeatedly told I'm Doing It Rong" department because the idea of practicing tone matching for hours just... kinda hurts. Oddly playing Guitar Hero for hours and hours badly does not set off this "argh I suck" reflex; maybe my violin teacher just really made me hate learning this thing wrt *violin playing* which is kinda sad but mostly I conclude that violin and me do not go well together.

[liv]

That's a really cool story, and yes, it really does fit into what we're discussing about meta-skills for learning.

[liv]

What an awesome list of suggestions, many thanks. I was trying to think of something along the lines of List your assumptions about the constraints. Test them all methodically to suggest to simont but couldn't think of a way of phrasing it that didn't sound like I was doubting his judgement about the impossibility of fixing something.

[liv]

Hmm, I didn't read celandine13 as arguing against repetitive practice. Indeed, repetition could be part of your debugging process; you might conclude that the error is occurring because you need to have a particular sequence of muscle movements down to the level where you don't need to think about it consciously, and the solution to that is to repeat the motions until you can do it as one chunk. Or that you need to build up muscle to be strong enough to do something, which requires repeatedly working against load.

The paradigm of bad instrument practice is that you just keep playing the piece over and over again badly, usually just reinforcing the same mistakes. You need to focus on a particular section and perhaps initially slow it down enough that you can get each note right, or train each hand separately. And as for running, I can speak from my own experience here; I've never run a marathon but I wanted to be able to run 5 km and when I started I really couldn't. I wasn't sure what all my problems were, but it was clear that not having the pure aerobic fitness was a big blocking issue. Simply making repeated attempts to run the distance would have got me nowhere (which is where I went wrong in the past), because I didn't start out even having the fitness to run for long enough to improve my fitness. So I needed experts to tell me that the way to improve your fitness is to start by running for very very short intervals and gradually build up.

I think celandine13 makes it clear (especially in the comment discussion) that she's being hyperbolic when she claims nobody's actually stupid. Yes, there most certainly are individual variations in... base ability, and probably in the maximum ability attainable with adequate practice and training. I think it's partly a framing issue rather than a disagreement of fact. If you say "I'm bad at playing the violin", then it becomes part of your identity that you're just bad at it, and you're unlikely to be motivated to practise; just think of all those people who say "I can't do maths". It's perfectly reasonable to conclude, as it looks like you have done, that your innate abilities mean that the amount of work it's going to take you to play violin to a tolerable level isn't worth it for you. But I do think it's dangerous to rush to that judgement, without unpacking what you mean by "bad at", if you see what I mean? Especially as a teacher, telling someone they're bad at the skill they're trying to learn is really demoralizing.

[naath]

I guess I mostly learned to run distance by... running as far as I could and trying to go further each time. So I suppose people vary a lot in what they need.

It is true that just saying "you are bad at this" doesn't help.

[doseybat]

Thank you. I do seem to have switched to what I would describe as "unsatisfying short attention span recreational web use optimised for immediate small rewards". Out of the five things I sort of read and post to (lj, dw, faceborg, private twitter, work twitter) it is faceborg and work twitter that provide me with the most interaction for the lest time invested. Perhaps if I continued thinking about your post instead of switching straight to the unanswered work email I would have more interesting thoughts to post here instead!

!!!

[liv]

Talking of debugging versus practice... I think you may actually be my fairy godmother. Basically my whole life I've been miserable about not being able to sing in tune. And I've done all the things that people suggest, not caring what people will think if my voice is audible, even taking singing lessons for some years, and I still can't reliably hold a tune. Here you are, an actual living example of someone who learned to sing having started with a genuine pitch problem, not a confidence or training / technique issue.

This makes total sense, it's possible to improve on that specific thing by practising correct pitch directly, and not trying (as I did with very limited success) to jump straight to singing actual songs. If you do have further info on how you managed to get past the pitch thing, I would really really appreciate it. If that's a hassle, well, even just this footnote to your comment gives me hope that I might one day get there by just repetitive practice of more atomic tuning skills.

[siderea]

I think I may also have a deficit in the "being willing to be repeatedly told I'm Doing It Rong" department because the idea of practicing tone matching for hours just... kinda hurts. Oddly playing Guitar Hero for hours and hours badly does not set off this "argh I suck" reflex

Sure. All else aside, there's a huge difference between a person telling one for hours one is Doin It Rong, and a computer giving value neutral feedback. Shame is an interpersonal feeling; it has to do with our feelings of relationship with others. One way to, sometimes, pull an end run around it when attempting something ego-threatening or ego-wearing is to do it without witnesses.

maybe my violin teacher just really made me hate learning this thing

Quite possibly. A sad thing that, but not like it's uncommon.

Re: !!!

[siderea]

Your and naath's comments are fascinating to me, because -- allowing as how the plural of anecdote is not data -- I've long wondered whether my path through music helped or hurt my development of pitch discrimination, and it suggests maybe it was very helpful after all.

I'm afraid that the way I learned to discriminate pitch and accurately sing was to first spend ten years studying piano, an instrument where "tuning" is something you pay a highly trained professional to do for you twice a year. I started piano at 6yo. Because I didn't have to discriminate or reproduce pitch, I could get on with all the other learning of music, with which I didn't have any problems. I built a very strong core of musical skills before I ever tackled pitch, which was my one area of enormous deficit.

And I've long wondered whether the deficit was in some sense the product of being trained on an instrument that didn't require the skill. Had I studied violin as a child, an instrument which required pitch skills, would I have been given useful instruction and expectations to meet that would have elicited the ability? Or would I have tanked, stumped by the most basic parts of the task? I have no idea, but I suspect it was the latter.

I started with piano at 6, added recorder at 8 or 9, fife at 10, guitar at 13, flute at 14, etc. I signed up for every singing opportunity that would have me, including a semester of voice class in my freshman year of high school (14-15). All of this exposure to music was informal ear training. When I got serious about doing that with my throat, I had lots of experience what it should sound like. (Back then, before the iPod, just when the Walkman was coming in, music wasn't ubiquitously available, the way it is now.) I had to learn to tune my guitar to a pitch pipe, and to match pitch in an ensemble with a recorder -- not that I did either well at that point, but it was a low-priority demand that exercised pitch discrimination a little.

Then, in my late teens and twenties, I decided to get serious about singing. I practiced every night with a tape recorder. For about ten years. (Man, GarageBand would have made that so much easier.) I took every singing class I could.

Here's what I know. Most pitch instruction I've encountered wasn't helpful because while it was in terms of relative pitch, it was in terms of relative pitch across time: moving in vectors, up or down from where you are a certain amount. My brain doesn't organize pitch that way, but across space: I move in absolute (if imprecise ;) degrees from a fixed referent. "Up a third" is conceptually useless to me; "sing the third degree of the tonic" makes perfect sense.

I'm the only sight-singer I've ever heard of who actually need to know the nominal key the piece is in to be able to read it (we can take it at any pitch you'd like, but I need to know what note of my score is the tonic to orient myself.)

So for me, trying to sing pitch more accurately monophonically was hopeless. I had to have a referent pitch. And starting out, that pitch couldn't be something I was imagining, because my head couldn't handle that yet. I needed a stable sound against which to make pitches in harmony. Accuracy was evident by the quality of sound made by the two pitches.

It is not impossible you are having the same neurological issue I did. In which case, it will be easier for you to learn to reproduce pitch accurately against a drone.

Vocal sound production is a profoundly sensuous experience. Not only are you learning what the tones sound like in your ears, you're learning what it physically feels like to produce those sounds in your body (the tension of the vocal apparatus, the friction of air, the activation of breathing muscles, etc), and what those sounds feel like resonating in your body.

Further, pitched sound makes complex sensory impressions in our minds. Even without full-on synesthesia, we experience different timbres and pitch combinations (chords, modes) as having qualities, to which we can only refer in metaphor, such as "bright" and "dark" and "moody" and "shimmery" and "crunchy" and "bell-like" and "horn-like" and so forth and so on.

In exploring vocalizing against a referent pitch, one will experience some of these things and they will begin to inform one.

Start with something simple, like singing with the drone letting your voice melt into it so that you can't tell where your voice leaves off and the sound outside your body begins, and then sliding the pitch up or down (away, any direction) slowly, and then undoing it. Feel how almost immediately, there's an aversive energy, which some people describe as resisting it, as a feeling of wrongness. That's "dissonance". If you keep going on through the dissonance, it will reduce in harshness, keep going far enough, and you'll find yourself going through a few fluctuations of harshness, and eventually finding another place where your voice can disappear into what you're singing with, even though the drone hasn't moved.

Try singing a series of a few pitches, steps of a major scale up from the drone note and back. Go very slowly, trying each note as if each were a step on an old rickety staircase you're not quite sure can hold your weight. Your first note should blend completely, your second degree feel uncomfortable and unstable and wrong, your third degree uncomfortable and unstable and surprisingly right. Wiggle your third degree around (up and down) and you may find a point where it has a particularly keen vibration in your throat, with the drone, a bright shimmering quality. That shimmering, carrying sound, two notes sounded with a relationship of a third, was famously England's gift to Western music, driving Continental composers mad with ecstasies for it. Strings of them, like beads, in sparkling garlands have been a favorite ornament of composers ever since. Keep going and feel more intervals, or walk back down the stairway, feeling what it is like to move into each relationship from above.

I'm running low on spoons and need to stop here. Does that give you somewhere to go?

Re: !!!

[jack]

Thank you, that was fascinating.

[jack]

there's a huge difference between a person telling one for hours one is Doin It Rong, and a computer giving value neutral feedback.

Yeah. When I'm learning something I'm completely unsure of, I often like to experiment in quiet and private somewhere, where no-one is going to interrupt and tell me to do it differently.

But you *can* replicate this with a computer. I found Wii Sport almost as annoying as an actual human, because lots of the skills you couldn't easily practice alone, you had to practice them as part of the appropriate game, and it would randomly keep stopping to say "you lost" when you were still learning how to swing the bat.