What makes programming so hard? Can non-programmers invent algorithms?(amazon.com)
amazon.com
What makes programming so hard? Can non-programmers invent algorithms?
http://www.amazon.com/gp/blog/post/PLNK2DYCU7ANMP5E6
7 comments
Teachers are operating "close to the limit of human capability"? Farmers? Policemen?
Some professions are just more natural for people - programming on the other hand is fairly unnatural.
Some professions are just more natural for people - programming on the other hand is fairly unnatural.
> Programming is a profession. Professions are hard. That's what makes them professions.
It could be a matter of tools, though; computing (ie. by hand) used to be a profession, but now we have calculators and spreadsheets and anyone can do a long series of sums.
It could be a matter of tools, though; computing (ie. by hand) used to be a profession, but now we have calculators and spreadsheets and anyone can do a long series of sums.
Its all about motivation. 20% of his students genuinely wanted to know how computers work. They were naturally curious. Curiosity is what drives an educational process. Good teachers try to get children curious, fascinated by something, and then very effectively explain how the phenomena works.
I hated physics in middle school, but in high school I got lucky and met the most amazing physics teacher the world had seen. That changed not only my grades, but my perspective on technology, world and life.
I hated physics in middle school, but in high school I got lucky and met the most amazing physics teacher the world had seen. That changed not only my grades, but my perspective on technology, world and life.
Part of the problem is that we usually teach in procedural style -- but this requires learning the MOST abstractions to do anything useful. Procedural style is based on variables, loops, conditionals, and side effects. Functional style is just based on function calls and conditionals; this makes it far easier to explain.
A larger part of the problem, I suspect, is that people are just not interested in the process of programming -- it's the product they want. The first thing you'll notice when you try to teach someone to program (in the normal case) is that they are focused to a fault on making the program work. This is the same problem that keeps people stuck to a single language: "well, it works." You need a genuine interest in the journey, not just the destination, to really be able to wrap your mind around the skill of programming. That's what gives someone the ability to intuitively understand how computers "work."
You wouldn't try to drive cross-continent if you're only interested in what's on the other side -- you'd just fly there, or take a train or bus or something. Likewise, if you're not interested in how a program does what it does, you should just pay someone who does care to do it.
A larger part of the problem, I suspect, is that people are just not interested in the process of programming -- it's the product they want. The first thing you'll notice when you try to teach someone to program (in the normal case) is that they are focused to a fault on making the program work. This is the same problem that keeps people stuck to a single language: "well, it works." You need a genuine interest in the journey, not just the destination, to really be able to wrap your mind around the skill of programming. That's what gives someone the ability to intuitively understand how computers "work."
You wouldn't try to drive cross-continent if you're only interested in what's on the other side -- you'd just fly there, or take a train or bus or something. Likewise, if you're not interested in how a program does what it does, you should just pay someone who does care to do it.
Who needs conditionals? ;) http://en.wikipedia.org/wiki/Church_boolean#Church_booleans
But really - I agree. Functional programming is easier to explain without resorting to all sorts of helpful lies. At the university I am helping freshmen learn C.
But really - I agree. Functional programming is easier to explain without resorting to all sorts of helpful lies. At the university I am helping freshmen learn C.
Learning to program means mastering lots of new abstractions, becoming comfortable with them, and being able to use them. And I know that if I encounter too many new abstractions at a time, my mind simply shuts down. This happens each time I try to learn Haskell for instance. I'm following a tutorial, and suddenly I encounter this:
Another thing is attitude. Everyone programmer encounters really hard or confusing problems all the time - the trick is being really stubborn, and never give up. That doesn't come naturally to most people.
Finally, most things we learn, we learn by imitating others. Our brains are hardwired to learn that way, but it's not very helpful for learning to program.
parseNumber = liftM (Number . read) $ many1 digitAnother thing is attitude. Everyone programmer encounters really hard or confusing problems all the time - the trick is being really stubborn, and never give up. That doesn't come naturally to most people.
Finally, most things we learn, we learn by imitating others. Our brains are hardwired to learn that way, but it's not very helpful for learning to program.
"Finally, most things we learn, we learn by imitating others. Our brains are hardwired to learn that way, but it's not very helpful for learning to program."
I would like to disagree (respectfully). Although in most CS programs, algorithms and PL classes tend to be heavily proof-based and end up being exercises in tedium. As a TA for a intro. programming class, I would like to present the latest problem-set to students as some kind of narrative, sketch out what makes the problem interesting and the natural flow of the program structure that follows the problem, that comes spontaneously to the mind. So in a way, people are learning by imitating/doing, and not grinding out the hard theory first.
Honestly, I don't think everyone here learned for the first time about loop's, as according to proving that loop variants hold on during the iterations; or convinced themselves the way that various data structures work, by way of category theory. You learned it by doing, and you know that it just intuitively works. While functional PL is harder than Java, I think it works the same way; you change the way you think and in turns, the way you program, by doing.
I would like to disagree (respectfully). Although in most CS programs, algorithms and PL classes tend to be heavily proof-based and end up being exercises in tedium. As a TA for a intro. programming class, I would like to present the latest problem-set to students as some kind of narrative, sketch out what makes the problem interesting and the natural flow of the program structure that follows the problem, that comes spontaneously to the mind. So in a way, people are learning by imitating/doing, and not grinding out the hard theory first.
Honestly, I don't think everyone here learned for the first time about loop's, as according to proving that loop variants hold on during the iterations; or convinced themselves the way that various data structures work, by way of category theory. You learned it by doing, and you know that it just intuitively works. While functional PL is harder than Java, I think it works the same way; you change the way you think and in turns, the way you program, by doing.
Maybe that is one thing that makes CS more difficult than other areas... that you must learn by doing. We really are hard-wired to learn by imitation (as the original commenter said). That is how we learn language, social norms, etc. CS, however, is definitely something that you need to learn by doing.
Interestingly, that may also explain the idea of "computer geeks". If CS selects for people who learn by doing instead of by imitation, you are more likely to see those types of individuals. If social norms and social interaction is learned by imitation, those are abilities that are harder to learn for someone who is more inclined to learn in another way.
Interestingly, that may also explain the idea of "computer geeks". If CS selects for people who learn by doing instead of by imitation, you are more likely to see those types of individuals. If social norms and social interaction is learned by imitation, those are abilities that are harder to learn for someone who is more inclined to learn in another way.
Some people are good at constructing mental models from a symbolic or linguistic description. Most need help constructing the mental model. That's why Montessori math education includes so many hands-on materials.
DrScheme has a stepper that helps visualize expression evaluation. I think that can go a long way toward helping "non-programmers" learn.
DrScheme has a stepper that helps visualize expression evaluation. I think that can go a long way toward helping "non-programmers" learn.
I like the mental model description. I think programming (like other skilled arts) requires constant evolution of programming mental models or memes.
Developing new languages is a great exercise in creating new tools to solve specific problems. Rails really kicked the status quo and made programming so much easier for a bunch of us. But is it possible to ever create a language that will make all the different kind of programming problems equally easy? It always seems like there is a tradeoff, programming abstractions vs performance. As another case, all of us know english, but how many of us are great writers? (not I! haha)
Great link though. Look forward to the day when game programming is as easy as brainstorming out loud to our computers.
Developing new languages is a great exercise in creating new tools to solve specific problems. Rails really kicked the status quo and made programming so much easier for a bunch of us. But is it possible to ever create a language that will make all the different kind of programming problems equally easy? It always seems like there is a tradeoff, programming abstractions vs performance. As another case, all of us know english, but how many of us are great writers? (not I! haha)
Great link though. Look forward to the day when game programming is as easy as brainstorming out loud to our computers.
Personally I think there might be something genetic. The same people that get computer science are good at a lot of other things that folks that don't understand programming don't seem to be great at. Look at Calculus or Physics. 20% of the students just get it, most of the class needs to memorize formulas and 20% of the people just don't get it at all.
I think it's not about Calculus or Physics, it's about Logic.
Exactly. Algorithms are simply ways of approaching and solving a problem, and the ability to devise clever ones is portable between disciplines. At my college, several professors in the CS department hadn't programmed anything since the '70s.
The balkanization of science and liberarl arts is probably the worst thing about university education today. The unwillingness of either to import valuable concepts from the other ends up hurting both.
The balkanization of science and liberarl arts is probably the worst thing about university education today. The unwillingness of either to import valuable concepts from the other ends up hurting both.
The best programmers are generally those who have learned it at a young age. Maybe that's the best time for the brain to pick it up.
Programming is a profession. Professions are hard. That's what makes them professions.
If it was easy, then people would just do it as part of other activities, and there would be no specialization worth of the title profession. We don't think of typing in URLs as a profession, or spell checking documents, but we do think of graphic design or architecture as professions.
Programmers are operating close to the limit of human capability, just as other professionals are. Only a few people can do it well because it's a talent which is unequally distributed, just like being able to draw.
As for why this particular activity pushes that limit? The hardest part of programming is building mental models of non-existent objects which have no natural analogs. It's dealing with the "virtual."