Why mathematics makes you better at programming (and so does everything else)

There’s been a bunch of discussion on whether mathematics is programming, whether you need to learn mathematics to program, etc. on Twitter recently. Some of it has been quite heated. So naturally I thought I’d wade right into the hornet’s nest because I’ve got no sense of self-preservation.

There are essentially 4 levels of question going on here:

  1. Is programming basically mathematics?
  2. Do you need mathematics to be a good programmer?
  3. Is there a useful mathematics of programming?
  4. Can learning mathematics make you a better programmer?

The answers to these questions are respectively no, no, yes but you probably don’t care most of the time and yes.

Cool, well, that was easy. Are we done?

Not just yet. I’m going to spend some time unpacking the last of these because it’s the most interesting: Does learning mathematics make you a better programmer?

Well, yes. Of course it does.

So does learning to write. So does learning to cook. So does learning about philosophy of ethics, or about feminism. Learning another human language almost certainly helps you to be a better programmer. Your programming will probably improve if you learn Judo, or knitting, or tap dancing.

I’d go as far as to say that it is unusual to find a pair of skills where learning one will not in some way improve your ability in the other.

Part of this is just that whenever you learn a thing, you’re also learning about learning. You learn what works for you and what doesn’t, and the next time you need to learn something you’ll be that much better at it.

But there’s more to it than that, right? Surely mathematics teaches you more about how to program than just about how to learn things.

Well, yes, but it’s not because by learning mathematics you’re necessarily learning anything directly applicable to programming. Sure, you might be, and you might even be doing the sort of programming to which that maths is directly applicable. But if even if you’re not, learning maths will still make you a better programmer than say, reading Derrida (probably. I haven’t read Derrida so I might be wrong) or learning to cook (Almost certainly. I have learned to cook, and mathematics was way more useful. Your brain might work differently than mine).

Why?

Well, in much the same way that there is a sort of generalised skill of learning that you can improve just by learning things, there are all sorts of other generalised skills. For want of a better word (if you have a better word, do tell) I call them “microskills”. They’re mostly not things that people think of as skills – it’s things like breaking a problem down into parts, abstract thinking, learning to ask the right questions, perseverance when stuck, learning when not to persevere when stuck, and thousands upon thousands of others.

People tend not to think of these as skills because they think of them as innate abilities. Stuff you’re just good at or bad at and you’ve got to cope with that. And I mean, it’s certainly possible that you can be innately good or bad at these (and I may be foolish enough to touch this debate but I’m not touching that one with a barge pole), but regardless of innate ability they all share a characteristic that makes them skills: they get better if you practice them.

Almost every skill you learn will depend on some of these microskills. Which ones is not at all fixed: People compensate by trading off one for the other all the time. You might be less good at working memory, so you practice your organisation skills and keep meticulous notes. You might be easily distractable and lack the ability to methodically push through a problem but make up for it by creative flashes of insight (yes you can learn to have creative flashes of insight, no creativity is not a fixed thing that you can never get any better at).

But it is at least biased. Different skills will encourage you to draw on different microskills. Mathematics will draw heavily on analytical ones – abstract problem solving, hypothesis generation, breaking down a problem into smaller parts, etc. All of these are things that are super useful in programming. So by learning mathematics you’ve developed all these microskills that will later come in super handy while programming.

But, of course, you can develop the same microskills when programming, right?

Wellll….

OK, no. I’m going to go with yes. You will absolutely develop all of the relevant microskills in programming. Learning to be good at programming by doing programming is 100% a viable strategy and lots of people successfully follow it.

The thing is, you won’t necessarily develop them to the same degree. Some you will develop more, some you will develop less. And this isn’t necessarily because the ones you would develop less in programming wouldn’t be useful to be better at.

There is essentially one key thing required for learning to get better at something (HORRIBLE OVERSIMPLIFICATION ALERT): Feedback

Feedback is how clearly and strongly you discover whether you did well or badly at a thing. If you can just try something and immediately find out if it worked or failed, that’s great feedback. If you try something and 3 months later you get a positive result that might be the result of that something (or it might be the result of something else that happened in the intervening 3 months), well… not so much.

Feedback is essentially how you get better because it lets you know that you are getting better, and that’s how you direct your learning. As you get better at a thing, the amount of feedback you tend to get degrades – and this isn’t necessarily because the benefit you’re getting from improving is decreasing (though it often does) it’s often because the benefit is moving further down the line, or into things that are harder to quantify (for example, the benefits of “a sense of good taste” in programming beyond a basic avoidance of complete spaghetti code are often not immediately apparent until you’ve sat atop a mountain of technical debt).

And this is where learning other skills can be super useful, because they provide different feedback loops. Sometimes they even let you develop microskills that appeared completely irrelevant and then once you had them turned out to be super useful (silly example: Strength training to develop core strength. Totally inapplicable to programming, right? Well, yeah, except that it turns out that actually not being distracted by lower back pain while you program is quite handy), but most of the time it’s just that the feedback cycle on them is different and the incentive structure for developing that skill is different – mathematics will lean more on some microskills than programming does and less on others. Similarly it will provide better feedback for some, worse for others.

I could provide some specific examples of which things I think mathematics will help you develop better, but I won’t. It’s very hard to judge these things, and people can argue endlessly about which is better for what. But the specifics don’t matter: Basically all forms of abstract thinking are useful for programming, and the only way that we could reasonably believe that mathematics and programming could provide exactly the same set of feedback loops and incentive structures as each other would be if we believed that they were the same thing.

And that would be silly.

This entry was posted in Numbers are hard, programming on by .