three people have taken this bait, but you are the first to respond and you link the original tweet for context, so here's the original answer i gave on curiouscat (back in 2016!!!!) which was actually in response to a person asking me why three-act structure is taught:
my answer on this point can only be partially correct, because there's probably a lot of historical politicking and received knowledge at work here, but if i hazarded a guess i'd say three-act structure is taught for basically the same reason the five paragraph essay is taught: it's very easy to grade.
i don't mean to say that elementary and high school educators are lazy. what i mean is, truly evaluating writing in the serious way a professional writer or academic or scholar does is incredibly time-consuming. as we all know, there are general rules to written communication except when breaking them allows you to make a point more profoundly. then the rules are out the window. a lot of intense textual work boils down to looking at written communication and asking "what here makes sense? what doesn't, and if i assume the stuff that DOESN'T make sense is a roundabout way of unexpectedly making sense, what happens?"
but when we're teaching kids just starting out how to write stories, or think about stories, or the same with essays, we're faced with the fact that teachers need to be able to teach something that can then be reflected in work students produce, more or less in line with standards set by various education bodies, from the school administration on up. so we have two issues here: we need to first of all establish things that students intentionally enact, and second, which teachers can recognize them enacting. that way we don't end up with a teacher writing on a report card "little suzy is excellent at catachresis" when what is actually happening is that suzy is really bad at using consistently sensible metaphors.
this is especially important when it comes to standardized tests at the state and national level. students need to be able to perform certain tasks in line with widely available and comprehensible metrics. having a student write a story that has a clearly delineated beginning, conflict, and resolution (or an intro, three points of evidence, and a conclusion) is much better for their confidence both as writers in the moment of the test environment, as well as in terms of evaluators' ability to recognize how they enact those specific ideas.
IN OTHER WORDS: three-act structure and the five graf essay are, in some sense, the training wheels of written discourse. they're intended to be things you put away once you've reached a certain level of mastery and can begin experimenting, but statistically most people probably don't have much reason to do that. as fundamental skills they're still very important and often you'll see them repeated at different levels, mostly because when we learn this stuff as kids it doesn't stick the landing in quite the pressing way it would in (for instance) an elective creative writing class as an undergraduate.
to this i add, just to further drill down on the essay part: the five graf format usually goes
- intro with thesis/claim
- first evidence point
- second evidence point
- third evidence point
- conclusion that returns to the thesis to demonstrate how the above points satisfy the claim
again, this is all about giving students small micromovements that can be clearly seen by an instructor in short order. it should be notable, also, that both of these things (three-act structure and five paragraph format) are the default mode of any lengthy response you get from something like chatGPT when you ask it to write an essay or a story. partly because there's such a rich dataset of introductory material like this, and partly it's because the extremely clear but not at all complicated movements are very easy for a word probability algorithm to replicate!
that's where what i said in 2016 about breaking rules comes into play: writing to a model is a perfectly fine way to learn something, but it's when you start realizing what breaking with a model affords you (and deciding whether or not you can implement it well, and how that might be accomplished) can really gas you up. LLMs like ChatGPT never break the model intentionally, only accidentally, which is why the results are always so fucking weird when it happens
