Beeminder is kind of notorious for its jargon. We’ve got beemergencies, an akrasia horizon, maxflux, and, most recently, tare tags. Some of these seem to us perfectly self-explanatory — derail, pledge, ratchet, commitment dial, post-derail respite — but of course they’re not. From a newbee perspective, the site is full of chicken clucking nonsense. [1] We have a whole Beeminder glossary, for the love of Beejeezus.
The downsides of that are obvious enough. Jargon can be pretentious and obfuscating. The upsides are pretty obvious too. Jargon is efficient and precise. At its best, a piece of jargon puts a concept handle on a whole suitcase-ful of hard-won knowledge and insights.
The Worst of All Worlds
If you can say something in fewer words with more precision using plain English, the Loving Beejeezus commands you to do so. That’s obviously the best of all worlds. We might’ve achieved something close to that when we replaced the cute but opaque “yellow brick road” with “bright red line”. (Apologies to those still wistful about the yellow brick road. I guess we can’t call that one the best of all worlds.)
What I claim is the worst of all worlds is pseudovernacular jargon: using plain English as jargon, rather than instead of jargon. There’s a huge irony here. Plain English can be too understandable. It’s like the quote tenuously attributed to George Bernard Shaw: “The single biggest problem in communication is the illusion that it has taken place.”
If you say, for example, as we used to do, “max safety buffer” then users, understanding each of those words, will jump to the conclusion that they understand the phrase. But “autoratchet” (as we now say instead) means something very specific in Beeminderland and it’s less confusing for users to hear “autoratchet” and know that they don’t know what it means than to hear “max safety buffer” and think they know but be wrong. As Mark Twain supposedly said, it ain’t what you don’t know that gets you into trouble. It’s what you know that ain’t so.
Mathematicians are the worst about this: real, complex, imaginary, rational, natural, normal, field, ring, group, smooth, set, function, … They even do things like make technical definitions for “almost everywhere” and “almost surely”. Some of those work surprisingly well, some work once you’re sufficiently inculcated, and others are kind of a disaster.
If you’re a mathematician and you’ve invented a nifty new complete ordered set (don’t get me started on “complete” and “set” — but “ordered” I like) with binary operators that follow the distributive law, as a way to axiomatize the rational (🤨) and real (🤦) numbers, then, before deciding to call that set-plus-operators thing a “field”, at least consider whether it could cause confusion with the thing gravity and magnets do in physics. Not to mention the dozen other definitions of “field” [2].
I don’t deny that natural language is — as the “field” example illustrates — impressively robust to such ambiguity, but first of all, there are limits. I think the worst example I’ve encountered is the technical term “common knowledge”. [3] And the illusion of transparency means it’s dangerously easy to fail to notice these ambiguities.
But my real gripe is how people imagine they’re being less jargon-y and more layperson-friendly by sticking to simple English. That only works if you’re actually translating the technical stuff to simple English. (Which of course is tedious and inefficient among experts — the whole reason we have jargon.) If you’re just co-opting simple English words as your jargon, that’s what I’m calling the worst of all worlds.
If you overhear a pair of doctors talking about gastric emesis, it might be super pretentious that they’re not just saying “vomiting”. They’ve essentially encrypted their communication. But at least it’s obvious it’s encrypted. What’s worse is steganographic encryption — when you encrypt communication and embed it in other communication to conceal the very existence of the encrypted communication. That’s what mathematicians talking about how “in the field of the reals, almost all numbers are irrational” are like [4]. I mean, I exaggerate. It doesn’t take long to realize that two chatting mathematicians are on a different planet from you, but terms like “real number”, “normal distribution”, and “almost surely” [5], if used in an even slightly normie-adjacent context, are ambiguity atom bombs.
Check out the example of Grant Sanderson (of 3Blue1Brown fame) talking to Neil deGrasse Tyson about how in statistics “people will often assume something’s normal when it’s not,” and Neil has to interrupt to clarify for listeners that Grant means a normal distribution, making a bell shape with his hands. Or if a math newbie hears you talking about “real numbers” and presumes you just mean “actual numbers” as opposed to hypothetical ones. I’ve had this happen myself, not for lack of knowing about the reals, but by accidentally saying “real number” to mean “not a variable”, before noticing the confusion I’d unleashed.
It would, ironically, be so much clearer if the math term for the reals were some fancy Latin thing like “continuata” or “fluxorum”. You see that, immediately know you don’t know what it means, look it up, et voilà; now you know. With “real” you have to see it enough times to catch on that it’s being used in a technical sense and then construct a disambiguating search query like “real number math”. (Of course LLMs have made this easier.)
Again, I get that when fields (not that kind of “field”) do this it does mostly work and you can almost (not that kind of “almost”) always distinguish, say, “real” and “real” in context. But there really are so many examples of this causing problems and I want to hammer on them because so many people I argue about this with are determined to defend pseudovernacular jargon, as I’m calling it.
But I don’t want to pick on mathematicians too much. Using common words as technical terms is better than, as John D Cook points out, how business does the complete opposite and coins technical terms to refer to common ideas. I’m not sure how fair that is but it feels true. I think fear of gratuitous jargon is the biggest reason people do things like use the word “group” to refer to one of the most deep and powerful concepts in math. John D Cook cites the term “group” as the candidate technical term likely to be most confusing for his readers. If you don’t know better it sounds like you mean “collection” which, Cook says, “is just close enough to correct to be misleading.” [6] But at least it’s a noun (and a verb, I guess). I think the worst of all (close behind “almost”) might be “onto” to refer to a function that covers its whole range. [7]
Suitcases of knowledge
Turning back to software engineering, and Beeminder specifically, the pseudovernacular “tests” bothered me enough that we now call them quals. That was the topic of the previous Beeminder blog post:
Your software’s meta-software for quality assurance and regression-prevention infrastructure is not “just some tests” […] More pragmatically, the word “test” is just too common and overloaded.
There are many other examples of Beeminder jargon that’s a result of us attempting to thread this needle. Before we shipped the tare tags feature we debated whether the magic string should be “#TARE” or “#RESET”. Ultimately I vetoed “#RESET” because of how overloaded that word is — did the graph reset somehow, like derailing? a fresh start? resetting the commitment in some way? So many conclusions a user might jump to, thinking they understand the word “reset” perfectly well. Which they do, which is the problem.
We’ve given ourselves a lot of experience trying to find the sweet spot between all these competing constraints — precision, concision, evocativity, approachability, greppability, … Hopefully I’ve convinced you that you’ve got to be careful around plain English, and you’ll want to come back around for part 2 where I’ll give my advice on what good alternatives to pseudovernacular jargon look like. It doesn’t (just) mean translating to Latin.
Footnotes
[1] Fun fact: the word “jargon” comes to us from Old English via French “jargoun”, or “gargoun”. It’s an onomatopoeic word to describe the chattering of birds, thence applied to incomprehensible language. Also related: gargling!
[2] Field a question, field of view, battlefield, field of biology, in the field as opposed to the lab/office/classroom, field of candidates, field in a database, field a team, …
[3] “Common knowledge”, ironically, is such a thorn in my side in trying to communicate unambiguously. Unless you have common knowledge [technical sense] that everyone in the conversation is a game theorist, you’re constantly having to disambiguate it. Because it’s common knowledge [colloquial sense] that “common knowledge” is merely something commonly known.
[4] The math fact that “almost no reals are rational” is super cool. It means that, even though there are an infinite number of rational numbers (rational is “ratio”-nal — meaning fractions; this term I’m kind of a fan of) they’re embedded in a literally bigger infinity of irrational numbers (like pi and the square root of two, numbers that can’t be written as the ratio of integers) such that the percentage of real numbers that are rational is exactly zero! But, again, there are still lots of rationals, hence “almost” none are rational. It’s 0% of them and yet not none of them. Infinity makes things weird.
[5] What if instead of “almost certainly” and “almost always” we used “pen certum” and “pen universum” or whatever the proper Latin would be? It needs some more workshopping but the point is that terms like “almost always” are especially hostile to newbies who can be deceived into thinking they’ve understood something when they haven’t. Would it have occurred to you to seek out the “almost surely” Wikipedia page if I hadn’t already linked to it? It almost surely would’ve if we used something in Greek or Latin. Also you’d be able to grep for it in a paper or textbook.
If you’re still skeptical that this is actually a problem in practice, have a Hacker News thread full of people confused by “almost always” because of mathematicians ruining it.
Or consider an example from @Sean in the Beeminder Discord who was very confused about large deviations theory specifically because he didn’t know that he didn’t know what “large deviation” meant.
[6] Here’s a silly idea to get the best of both worlds: Take the ostensibly normie-friendly naming to an even further extreme and use cute phonetic spellings. Reel numbers, groop theory, etc. Now readers can tell again that you don’t mean the words it sounds like you’re saying. Which is good, because you don’t.
[7] Here are my arguments that the jargon-y sounding terms “injection”, “surjection” are superior terms to the pseudovernacular “one-to-one” and “onto”:
- You can grep (aka control-f) for “surjection” but you can’t grep for “onto”.
- When a newbie sees one of these -jection words they know that they don’t know what it means and can look it up. That’s less confusing than “since f is onto, blah blah blah” which looks like a typo or the author had some kind of stroke midsentence.
- After a small upfront cost, the Latin-based terms are more evocative and memorable. For one, they’re themed, by all ending in -jection. (There’s also “bijection” which means both an injection and a surjection.)
- “Injection” you can visualize as the elements of X mapping to a subset of Y — the function injects X into Y. “One-to-one” always makes me think it means a bijection — that everything in X has a thing it corresponds to in Y and vice versa. That’s the usual nontechnical meaning of “one-to-one”.
- (Which is my more general point: turning everyday words into technical terms is a recipe for confusion.)
- “Surjection” is easy to remember once you’re used to it. The prefix sur- means “over/above/onto” — like surname is the name everyone in the family has; surface is the outer face; surmount means overcome; survey means look over. So a surjection maps over the whole codomain, covering all of it.
- We definitely need a word for “bijection” and the plain English equivalents of these terms have got nothing here other than “one-to-one and onto”.