Planning, Precision, and Patience

I’ve started and restarted this post a dozen times as I’ve gotten ready for hip replacement surgery this summer (which finally happened last Wednesday!).1 Experiencing the detailed care of my surgeon, nurses, and physician assistants only reinforced how much their protocols run on technology. Visits are often recorded, images documented, and almost every stage of the process digitized. When I asked whether any of this data was integrated with AI, the short answer was yes - with the privacy and security caveats you would expect. The hospital was the latest reminder that AI is being folded into every industry we encounter, whether we know it or not.

Everyone, including teachers and students, needs to work in an AI world. Almost every story about AI and education invokes the risks of cognitive debt, cognitive offloading, and a loss of cognitive patience. As we enter the 5th academic year where students have had access to powerful AI, most of the news out of schools is negative. More telling, even the most charitable observers would agree that few of us know what we’re doing when it comes to AI. But I’m hopeful that will change this fall.

I've read the reports and documented my AI fears when it comes to unsupervised student use. What I'm trying to square is that my own experience with AI has been mostly positive. Working with AI over the past three and a half years has required a learning curve, but it has made me a better planner, a more precise writer, and it has forced me to slow down and review my work much more intentionally. With billions of people now using AI models, it's clear I'm not the only one benefiting.

So much of the angst around students' AI usage presumes that they will continue to use it the same way they have over the past three years. We have to remember it is still early days. Getting real benefit from AI takes a classic set of skills (or habits or dispositions if you prefer), and every one of them is something we have always taught, long before 2022. Effective AI use requires planning, precision, and patience. They apply whether you let a chatbot into your classroom or not.

One major problem for anyone advocating student use of AI - whether in the classroom or on their own - is that there is very little evidence it helps students learn. The further I get from November 2022, the more it shocks me that this technology was unleashed on an unsuspecting public, especially teachers and students, with so little warning. No other “arrival technology” in my 30 year career has disrupted the teacher-student contract as much as generative AI.

I’m aware that my usage rests on skills I formed over decades. Yet, I’m confident we will ultimately find ways to absorb AI into education by leaning on the same meta-cognitive habits we have always valued. Planning ahead, communicating with precision, and having the patience to see a project to completion are all things any teacher in any classroom already models.

The critical question over the next several years is when and where students will continue to build these skills - alongside AI assistance or away from it. Right now, most teachers lean towards the latter, and that may hold for a while. What I don't see, however, is how we are going to have a choice regarding student usage outside of school. My hope for this fall is more evidence of educators successfully modeling these skills, regardless of whether they bring AI directly into the room with their students.2

I often share with students the reality of work. Years ago I read that college-bound high school juniors are as busy as CEOs. There is almost no other time in your life when you will be expected to excel at five entirely different subjects. Most of us will eventually get the opportunity to specialize in an area where we are strongest, but that is usually not the case in high school. That fact alone explains a great deal about how students use AI. A kid managing five courses plus outside school commitments has no time to plan, no incentive to be precise, and little patience left for a second draft.

I've explained to students that almost all work is best approached by breaking it into smaller parts. To do that, you need to check your anxiety, step back, and survey what actually needs doing.

AI raises the stakes on that crucial habit. We have all gone down a rabbit hole, and AI rabbit holes are among the worst, partly because the machine flatters you at every turn and will happily lead you somewhere useless or overly complicated if you're not careful. The discipline any project requires has to be in place before you ever go near a chatbot. If you know what you're doing, though, AI can assist with the planning by challenging assumptions and expanding on strategies. It can also help you decide if the task is worth using AI on at all.

My hunch is that a startling number of people still use their chatbot as a string of disconnected one-offs. If you use AI infrequently, you probably turn to it for a few questions, close the tab, and move on. Which means you have dozens of orphaned chats piling up in a sidebar, unnamed and unrecoverable. If that is still how you work with AI, you are missing what's been happening over the past two years.

I have no interest in writing an AI tutorial - there are dozens available online. But the models have rapidly added organizational tools and none are difficult to learn. Keeping related chats inside a dedicated Project folder means the work stays together and the model can draw on related material. Adding a set of instructions and reference documents within each folder considerably improves the output. These are simply AI versions of separate notebooks and divider tabs, a habit older than most teachers. Gemini, ChatGPT, and Claude offer almost all of these for free, and the lowest-tier paid versions provide access to even more sophisticated capabilities.

We constantly assign students work to do, but in the age of AI, we have to remember that planning out a multi-step activity - whether it is a research paper, a science project, or a presentation - is a meta-cognitive process. Is the goal to produce a document or ask for feedback? Are you creating, analyzing, or organizing material? Is this for you, or for someone else? Nine times out of ten, simply asking yourself what it is you are trying to do will force a decision about whether or not to use AI in the first place.

None of this is new to teachers. We have explained and modeled executive functioning skills for years, through graphic organizers, outlines, and day planners, well before AI was available. I am a longtime evangelist for David Allen's Getting Things Done, and I wish someone had handed it to me before college. Planning ahead does not come naturally to the teenage brain. It never has. That is precisely why we teach it. We need to prioritize the planning before the AI, now more than ever. Refusing to recognize that AI can help with the planning itself, however, is also a mistake. As adults, we must constantly model judgment.

Using AI has made me a better writer for one simple reason. It’s made the cost of vagueness instantly painful.

I've written previously about growing up in a household filled with lawyers who were maddeningly fixated on using precise language. It created an ideal training ground for writing clearly. Law school provided three more years of practice. Writing with precision and clarity was not an abstract concept in my household - I saw it up close in every brief my father ever showed me.

Iterating a conversation with a chatbot is an exercise in specificity, context, and requirements. If you want to convert notes from a meeting into a bullet-pointed to-do list with each entry no more than five to seven words, then you need to describe those specifications. Or if you want to design a research prompt that will focus on retrieving publicly available academic, government, and other vetted documents only, you need to ask.

When you do not precisely identify what you want, when you do not supply the context of the request, when you are not even sure yourself what you're after, you should not be surprised when the output is subpar. The more detailed and clear your request, the better the results. Whenever I write a prompt, I am constantly revising and scrutinizing my work to ensure there can be no misunderstandings. Most of the disappointing output traces back to an unstated assumption in my request, or some instruction I didn't even realize mattered until I saw how the AI responded.

We do the same thing with students. Using AI has reminded me how much depends on clear instructions. We cannot blame the student for misunderstanding an assignment when we don't take the time to be specific.

This is also true from the student side, and one reason that modeling precision is the province of every writing teacher. Anyone who grades student work has written some version of I don't know what you mean here in the margin - writers often forget that the reader is not inside their heads, and younger writers are especially prone to it. Using AI delivers feedback on fuzzy requests immediately and without mercy. The truth is that a teacher who never allows AI in class is still teaching the skill that matters most if a student opens a chatbot on their own.

Most kids just want to be done with school, period. They want the current assignment off their plate and the very thought they need to “do it over again” induces horror. Anyone who works around teenagers knows this.

If you put the same kid on the field, however, something surprising happens. They will run the same drill over and over until their coach stops riding them. It’s not that they don’t understand what patience looks like. It just doesn’t map onto their experience with reading, writing, or arithmetic.

I am a firm believer that writing is rewriting, and what is rewriting if not patience? When a student wants to hand in a first draft, the question I ask is whether it is their “best work.” Most of the time they hesitate which tells us all we need to know. Our first drafts are never our best work - adults included.

None of this has anything to do with AI. Patience is what we model every time we tell a class that something polished - a poem, a painting, a song - took longer than you might imagine. Extraordinary results require extraordinary patience.

This is why the way AI often gets demonstrated to the public bothers me.

Ethan Mollick recently dropped another one of his “state of the union” posts on the current abilities of the top models. His work is the most up-to-date and accessible I have found, and I recommend it to anyone who wants to keep track of AI developments.

One area where I think he does his readers a disservice, though, is his demonstrations of impressive and "cool" applications built from a single prompt. In his most recent post he compared one line about a procedural brutalist building creator across two models, with the newer version significantly "better," if better means it looks cooler than the first one. There are a lot of things AI can do that are "cool" - but is that what makes it useful?

I understand his larger point, and the mere fact that these models can do such things, considering where we were even a few months ago, is mind-boggling. But no one doing serious work with AI does so based on a single sentence prompt.

The way AI is going to change work is through the people who figure out how to tease a project apart, separating the human review from the repetitive labor AI excels at. When those people begin to pull ahead of the people who either do not use AI or use it ineffectively, then we will see its impact. That takes the patience to iterate a project from the beginning and see it through, rather than issuing a one-shot ten word prompt.3 Perhaps we will get to a stage where that becomes the norm, but we are not there yet.

Two years ago I was fixated on simply being AI-Aware as we were all learning how to handle the emerging AI crisis in schools. Last year my strategy was AI-Proofing, which was defensive by design, as I, along with many others, attempted to recreate assignments that could not be easily outsourced. The evolution makes sense, but it's not where I think we will end up.

My mindset this year turns to instruction - not specific to AI, but a re-commitment to the durable habits of mind that improve learning no matter the tool. By reminding students of the importance of planning ahead, communicating clearly, and exercising patience in everything they do, we will have a successful year -with or without AI.

I have seen all three of these skills converge this summer as I have been building out full websites for my courses.

The planning came first, as I had to think carefully about what I actually wanted these sites to do, what information they needed to contain, and how they were set up. Precision was present throughout, as I needed to specify exactly what I wanted each page to do, where and how I wanted the links to operate, and what kind of layout and formatting would make it most useful for my students.

Patience has been ongoing, as I am still iterating and fine-tuning before the school year starts. A one shot prompt might have created something "cool" but would have produced garbage as far as my actual classes were concerned.

The web design project required skills applicable to any task - this one just happened to incorporate AI.

Hospitals, marketing firms, technology startups, and a whole host of other industries have leaped feet first into AI. No one wants to be left behind. Results have been unsurprisingly mixed even though massive amounts of capital have been invested. How this all shakes out is anyone’s guess but we won’t know for awhile. Amara’s law hovers in the background - “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”4

Education is still behind. Few teachers are taking advantage of the most powerful ways AI can be used to augment their work. Most of them are likely keeping it to themselves or, occasionally writing about it publicly. Many teachers (unlike their students) are still deciding whether AI should be allowed on campuses at all.

That argument, I think, is the wrong one to be having. Teachers who want to use AI with students and teachers who want nothing to do with it can model the same habits, and both can point out that those habits are what separate using these tools well from using them badly. In both cases, student walk away with something that works. Technology changes. Good teaching does not.

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.