Artificially Intelligent - Part I

AI is not the solution the EdTechBros think it is right now, but could be if we got back to actually teaching kids to know stuff and do things.

My students keep telling me our literature class is mind-blowing. But it’s not down to me; I’m just the messenger. It’s not me who makes them gasp and sometimes teary-eyed. Yesterday, it was Douglas Malloch and Rudyard Kipling. It’s been Claude McKay and Paul Lawrence Dunbar and Robert Hayden. For weeks, Harper Lee’s deep magic held them spellbound, made their hearts pound, broke them, then sealed them back up: scarred but resolute.

There’s a loud contingent of people that think AI is the future, so it should also be the immediate now. Kids should, yes, learn to read, but they should spend their time learning to engineer prompts. They can always look up everything else.

Sounds great, like magic. But somehow, I don’t think this is going to work out the way you guys, with your preexisting wealth of knowledge, think it will.

For you parents, teachers, and the TechBros (and I love you, TechBros, don’t get me wrong — I’ll explain why later) today I’m channeling Dorothy Sayers to explain why your kids need to learn all the things they can without trying to cheat the process of learning with AI and why you should be highly skeptical of any teacher or school system that embraces its widespread use among students.

Because, as we all know: you don’t skip leg day.

---

Last year, at a classical charter, one of my colleagues confided that she’d been using AI to help her prepare her lessons to save planning time. This immediately set off warning bells. My child was in this woman’s class and it was quickly apparent to him that she was teaching a subject she knew almost nothing about. Since she was teaching middle school, it probably wouldn’t have taken her too much effort to get out ahead of them by reading the material and doing the work of breaking it down in more detail for her students.

This used to be understood as part of a teacher’s job. If you were assigned a subject where your knowledge was weak, you’d have to do the heavy lifting for a year or two to develop deeper ontent expertise.

But my colleague didn’t. Instead, she decided that 3:30 P.M. was her cutoff for work and that she’d outsource to ChatGPT the heavy lifting of explaining things she didn’t understand, as well as organize her presentation of the material and generate the review questions so students could demonstrate their understanding.

Okay. So AI solved one lady’s class planning problem, but it left her in the lurch. See, she had to grade all those answers. Chat GPT had given her an answer key, of course, but because she didn’t know the material, she had no clue how to accurately assess why the students were getting things wrong. She couldn’t identify the prior knowledge they were missing because she didn’t have it herself. She didn’t know what to ask AI to get the knowledge she needed in the time she had allotted. In other words, she effectively turned herself into a substitute teacher while patting herself on the back for her forward thinking.

Good thing she got promoted to administration shortly thereafter. Now ask me how many of the teachers there use AI in the same ham-handed way she did.

At any rate, ChatGPT solved her work-life balance problem (in the short-run, anyway). But what hasn’t been solved, and what teachers who rely on Large Language Models can’t solve, is the learning problem.

Moreover, AI can’t solve it for students either because the school system has created structures and policies antithetical to human learning. AI doubles down on them.

---

In a landmark 1946 speech nearly every homeschool parent has at least heard of if not actively sought out and studied, Dorothy Sayers laid out the argument in favor of the medieval Trivium in 20 pages titled ”The Lost Tools of Learning”.

In it, she breaks down the the three stages of learning that roughly correspond to human neurological development. These are commonly known in classical education circles as the Grammar, Logic, and Rhetoric stages.

Here’s how Sayers describes each.

Grammar, which Sayers terms the “Poll-Parrot stage”:

The Poll-parrot stage is the one in which learning by heart is easy and, on the whole, pleasurable; whereas reasoning is difficult and, on the whole, little relished. At this age one readily memorises shapes and appearances of things; one likes to recite the number-plates of cars; one rejoices in the chanting of rhymes and the rumble and thunder of unintelligible polysyllables; one enjoys the mere accumulation of things.

Sayers refers to the Logic stage as “The Pert Age” and has some fun opinions about that one:

it is characterised by contradicting, answering-back, liking to “catch people out” (especially one’s elders), and the propounding of conundrums (especially the kind with a nasty verbal catch in them). Its nuisance-value is extremely high. It usually sets in about the Lower Fourth.

Finally, she refers to the Rhetoric stage as “The Poetic Age” which

is popularly known as the “difficult” age. It is self-centered; it yearns to express itself; it rather specializes in being misunderstood; it is restless and tries to achieve independence; and, with good luck and good guidance, it should show the beginnings of creativeness, a reaching-out towards a synthesis of what it already knows, and a deliberate eagerness to know and do some one thing in preference to all others.

Let’s deal with all three of these in a little more depth so you don’t have to read all of Sayers speech (though I heartily recommend it if you have time; it’s linked above).

The Grammar stage more or less corresponds to the elementary school grades from K-5, though Sayers allows that some children are ready for promotion to the logic stage before 5th grade. To the modern reader, “grammar” means the study of parts of speech and maybe, if you’re old enough, the skill of diagramming sentences. That’s not what good ol’ Dot’s on about at all though. Her use of the word “Grammar” indicates the structure and vocabulary of multiple disciplines, most of which get short, if any, shrift in the modern public elementary school. To Sayers, Grammar means literally instilling in children the terms, names, and facts pertinent to at least eight different subjects.

Latin is first on her list. The serious study of Latin, specifically, helps students understand the grammar of a mother language, setting a child up for success not only in her primary tongue, but in other Romance languages as well, which Sayers believes students should also be taught when very young.

As for story, she emphasizes the reading of and questioning around English verse and prose, heavy on recitation of songs and poems, as well as myths, fables and fairy tales from the Mediterranean civilizations as well as the Norse and British.

Sayers posits that Poll-Parrots should know the dates, events, personalities, and anecdotes pertinent to History. In Geography, she takes it as given that students will be helped to an understanding of not only maps and geographic landmarks of their community, but the flora and fauna endemic to each region all the way down to the ability to identify locally-occurring mushrooms safe for human consumption vs. those most decidedly not. In addition, she believes children should also be taught the anthropology of each region, down to costumes and customs of its different peoples.

In science, she prescribes for the young child not only the accurate scientific classification of animals and insects, but the identification and memorization of the constellations so that they can be easily located in the night sky. In math she states (with what to the modern reader may seem a surreal calm) that young students should easily be able to memorize and quickly access their multiplication tables to do math in their heads on the fly.

In other words, Sayers says the goal of the teacher from ages 5 to 11 or 12 is the dreaded and derided “rOtE mEmOriZaTiOn”. And here’s why Sayers was right in graphic form:

The helpful model of learning above comes courtesy of Efrat Furst, a woman who has made it her mission to help teachers make the most of neuroscience to maximize student learning. I’ll be referring to her models throughout this piece, but you should mosey on over to Furst’s website, especially if you’re employed in a classroom or are a homeschooling parent and want to know how to get the best outcomes for your child. (For practical application at home, you may find some of my work helpful which started during COVID.)

So what does that model mean? In short, that there are very few ways to learn something new. In fact, I’d say there are only two: either repeated exposure to information (this IS memorization) or what Furst’s model illustrates above: layering new knowledge on top of existing knowledge.

You know how your baby learned to talk. She heard you say the same word over and over and over and, eventually, her brain connected the idea of “Mama” to you. Eventually, she learned to call you Mama, just like she learned to call the dog by her name, and her brothers and sisters by their names. Repetition over time, with space in between to forget (forgetting is crucial to learning), then the desire to recall the information when she wanted it took that little light gray triangle above and turned it dark gray. Obvious and clearly defined, the names of people and things became easy to recall.

New knowledge (Furst’s orange triangle) is added to the grayed chunks of learning. Mom and Dad and Sister and dog and Grandma and Grandpa become the larger concept “Family” in the bottom left diagram above. Once a child has a strong foundation, new information can be added to the old so the child can build out an understanding of larger, trickier ideas, which are represented in the model in the second picture of the bottom row, things like neighborhood, community, and city. In the third picture, you can go higher into the more ephemeral realms of human understanding where a kid has the foundation, finally, to understand concepts like “heritage” or “posterity” or “legacy” and all the building blocks that go into it.

As poorly as our schools have been performing, they more or less operated on this principle. Kids learned facts about science and history and read the mythology and fairy tales of Western civilization. They practiced math. In great schools, they also started to hear the important names and concepts in the arts. These things tended to be repeated over time. As the child rose through grade levels, teachers would pull up that old information and add additional nuance, asking students to put together more information to get a more complete picture of their culture, gearing them to be productive members of society.

And then things started to go very, very wrong. Many teachers will chalk it up to testing mandates started with No Child Left Behind and then added to with Common Core then intensified with the Every Student Succeeds Act. This put the onus on teachers to ensure students scored as high as possible on standardized tests focused on English and mathematics.

But something funny happened with that two-subject focus. As teachers began to teach reading and math to the exclusion of the additional subjects that Sayers mentions in her essay, scores went down. And they keep dropping.

That’s because grammar, as Sayers defines it, is the knowledge that a student needs to layer additional knowledge on top, to build the model of knowledge Dr. Furst illustrated for us. Knowledge is the cornerstone of literacy and it’s reading where humans get the greatest intellectual growth. In addition to the knowledge you acquire from reading specific texts, reading engages multiple brain regions, particularly those involved in language processing and memory, leading to better recall and overall brain function. The act of reading not only improves brain function now, multiple studies demonstrate their benefit to long-term brain health, including the prevention of dementia.

All literacy is, at base, is knowledge. In order to read a book well, you have to bring certain knowledge to the topic. I just started reading The Making of the Atomic Bomb with my science-first, 11-year-old nerd middle kid. As we read it, I have to ask him things because he knows more about subatomic structure than I do, but I know enough to hook in to his explanations. The book is easier for me to grok because I’m knowledgeable about the early 20th century run-up to World War II, but my 11 year-old gets most of it. The tome is accessible to him because of the foindational knowledge in science he brings to it. And now the DK book on World War II is helping him layer additional knowledge about the war, the personalities of the war, European and American politics and science, and the development of nuclear science. On top of that, he’s learning about Jews in Europe and connecting that to his pre-existing knowledge of the Holocaust and beginning to understand and ask questions about the ethical questions around government’s role in science, human rights, and death itself.

Okay, DT, stop bragging about your nerd kid and tell me what this has to do with AI.

Everything, but first I have to take you back to my high school teaching career, which I’ll do next week.