Slashdot Mirror
Ask Slashdot: How Can Programmers Explain Their Work To Non-Programmers?
Slashdot reader Grady Martin writes:
I disrespect people who describe their work in highfalutin terms... However, describing my own work as "programming solutions to problems" is little more than codifying what just about anyone can perceive through intuition. Case in point: Home for the holidays, I was asked about recent accomplishments and attempted to explain the process of producing compact visualizations of branched undo/redo histories.
Responses ranged from, "Well, duh," to, "I can already do that in Word"...
It's the "duh" that I want to address, because of course an elegant solution seem obvious after the fact: Such is the nature of elegance itself. Does anyone have advice on making elegance sound impressive?
An anonymous Slashdot reader left this suggestion for explaining your work to non-programmers. "Don't. I get sick when I hear the bullshit artists spew crap out of their mouth when they have no idea wtf they're talking about. Especially managers..."
But how about the rest of you? How can programmers explain their work to non-programmers?
Responses ranged from, "Well, duh," to, "I can already do that in Word"...
It's the "duh" that I want to address, because of course an elegant solution seem obvious after the fact: Such is the nature of elegance itself. Does anyone have advice on making elegance sound impressive?
An anonymous Slashdot reader left this suggestion for explaining your work to non-programmers. "Don't. I get sick when I hear the bullshit artists spew crap out of their mouth when they have no idea wtf they're talking about. Especially managers..."
But how about the rest of you? How can programmers explain their work to non-programmers?
I tell them that my type of programming is almost all mathematics. Few programmers use this much math. I look at all kinds of problems and find solutions for them.
I use linear algebra all the time. You might have been exposed to matrix math, using 3x3 and 4x4 matrices, multiplying them together, finding inverses and transposes, working with vectors and dot products and cross products. That's the basic math of the 3D world, much like addition and subtraction are the basics of the math for processing a budget. I do linear algebra on paper and on calculators at my desk, plus I tell the computer to do that stuff. I read technical math research papers about many systems, often two or three per month, and I implement the algorithms they describe.
I look for errors in code and fix them. When people say "this isn't working right", I find the errors and fix them no matter who wrote the buggy code. Sometimes it is very difficult to understand other people's code, sometimes it is like a giant knot that needs to be untangled so it can be understood. When people say "this is running slow", I study the math that they use, I figure out what math can be faster, and I tell the computer to use that math instead.
I work with designers who say "I need a game system that does such-and-such", I figure out a way to make that happen. For example, on {project they would understand} I programmed the computer how to {action they would understand}.
Many times I am given very difficult problems that are studied by mathematicians and scientists, problems that are known to be impossible to solve at speeds games need. That's why the impossible problems, intractable problems, and fast-growing problems are important to study in college. In those cases my job is to figure out other ways to solve the problem that are fast enough to run in microseconds. Usually that means cheating and taking shortcuts. As one example, finding the perfect organization for objects inside a container is very difficult and slow, since the perfect organization requires testing every option. Instead we can cheat by taking the biggest items and stuffing them in the container first, getting smaller and smaller pieces until they are all present. Another is choosing the perfect path which takes a lot of processing, versus cheating and taking the first really good path. It often isn't the perfect solution, but it is good enough for what we need.
//TODO: Think of witty sig statement
"Well, you see, that form is actually an instance of a subclass that inherits from that object which can be stored into that templated array thanks to polymorphism", then no more question from the non-programmer.
if you're looking for a way to justify your existence (as opposed to *genuinely* explaining what it is that you do to an outsider) then i would suggest prefacing that with, "i'm going to start at a high level. i'm then going to go into detail. my ability *to* go into detail is precisely why you employ me rather than someone who can do stuff with a spreadsheet. please feel free to stop me at any time when you have heard enough"
followed by going into detail and not stopping until they tell you to. when they've had enough, you can finish up with, "so do you now appreciate that this is far beyond the skill set of a lay person, to cope with this level of excruciating detail? i deal with it so that you don't have to. it's extremely challenging and tedious in a mind-numbing but extremely rewarding way for me, but only in that it's a massive challenge well achieved. can we please, therefore, in future, keep our conversations to the high-level requirements, and more than that, when i tell you that i *don't know how long something will take* please be patient and trust me to work through it until i know more, okay?"
if on the other hand they *genuinely* wish to know about programming, my favourite way to explain that is as follows:
okay, the idea is, you're going to give me a series of written instructions - a recipe - which i can give to absolutely any person, for them to follow in order to get from one corner of a tiled room to the other, negotiating around obstacles. that person will be BLINDFOLDED so that they only have a sense of touch in any direction of distance equal to ONE tile.
you then give them the following example:
step 1: go forward one step
step 2: if you didn't bash into an obstacle, go to step 1
step 3: go right one step
step 4: if you didn't bash into an obstacle, go to step 3
step 5: are you in the far corner? if yes HURRAH
step 6: repeat from step 1
*we* know what the flaws are in that algorithm... but they won't. so, you LITERALLY get them to walk through it. as in, LITERALLY follow those instructions on a tiled kitchen floor. then you DELIBERATELY place obstacles so that they will get stuck, and ask them, "ok, so now how would you fix that?"
and when they go, "ahhh okaaay i get it. it's step-by-step stuff but you can get into trouble if you don't give the right instructions", then that really is the lightbulb moment for them in *truly* understanding the basics of programming. at *that* point you can explain to them that, unlike that very simple 6-step algorithm you write algorithms of TENS of THOUSANDS OF LINES every few months, and that the linux kernel is what... thirty MILLION lines or something insane, then they'll finally start to really and truly Get It.
if that's your boss they might even actually give you a payrise or at the very least treat you with a little more respect.
Case in point: Home for the holidays, I was asked about recent accomplishments and attempted to explain the process of producing compact visualizations of branched undo/redo histories.
You've gone into the wrong kind of detail. The useful answer generally has the form "X is the general problem that people have which you can relate to in some way from your personal experience. Y is the state of the art. I've improved upon the state of the art in Z."
Thus: "You know how sometimes in Word you type a paragraph, then want to undo it and start again, but you sometimes want to keep a sentence or two from the thing you typed even though you undid it? People usually use copy+paste, if they remember, but it gets hard to keep track and sometimes you accidentally mess things up so you can't redo back to your first draft. You're confused at this stage? -- exactly! :) Well, I've been working on a new way that avoids the pitfalls. It seems to be working, and users have been giving good reports so far. I'm not adding it in Word of course. But who knows? maybe my idea will catch on."
People aren't interested in the technical details of your solution. They're more interested in the general scope of human endeavor, and the conflicts and social dynamics in the research field. So if you meet a researcher or a PhD student, the second question you ask them (after "what's your field?") is "what are the main opposing ideas in the field?"
If you're not advancing the state of the art in any way, and if you're just implementing a solution that someone else has done, again don't talk about the technical details of the implementation. For instance you're doing a back-end database and you're copying some scaling algorithm/implementation from someone else, you can say "Imagine how Amazon must have to process like two hundred million order requests every day? My company also needs to process one hundred million widgets. We're not quite at the same scale as Amazon, but I've been copying some of their techniques too. It's fun. I've learned [incidental social fact about the human endeavor that is software development]".
My day job is doing technical implementation of language features inside a code editor (think autocomplete, signature-help, hover, ...). Even when I'm speaking with my MANAGERS and PEERS I don't talk about the technical side. The first and last thing to talk about is always what's my overall mission? and specifically, what user-facing problems/scenarios am I trying to solve? The technical details is always an afterthought. Successful software engineers are primarily good communicators.
Most people advise trying to dumb down the descriptions to make them more understandable the uninitiated. I don't typically do this, unless I'm dealing with children. When you simplify your descriptions, you are effectively simplifying (in their minds) what it is you do. You make it sound simple, and they suddenly think you're paid to do simple things.
Hit them with the complexity. You probably don't have to force it and go overboard with jargon (no need to activate peoples BS meters). Talk to them as if you were talking to a colleague. This will generally have one of two outcomes:
Yaz