On August 24th, 2026, all students in kindergarten through 12th grade are eligible for free rides on Chicago Transit Authority (CTA), Metra, and Pace.
That is to say, today is the first day of school for University of Illinois Chicago (UIC), as well as the Chicago Public School system (CPS). That means the active 9-month portion of my annual contract is now active, so it is time to do my job. And what is my job, exactly?
A coworker of mine has mentioned this George Pólya quote a few times, and I am rather fond of it:
"Our job as teachers is to provide opportunities for discovery."
A brief translation: when someone is just told information, that is rented. It's not really knowledge. But when someone discovers knowledge, they own that knowledge.
If you want to learn something, if you want to really know something, you must teach it to yourself. A "teacher" can only do so much.
This semester I am "teaching" an intro to data science course as well as two sections of the ethics course for computer science majors. Basically every Monday and Wednesday until December, I will be in a classroom with students, helping them to teach themselves about data science and about ethical issues in computing.
As I lock tf in for the next three months or so, my focus will be on doing my job effectively: providing students opportunities to follow their curiosity and discover the knowledge that these fields have to offer.
I used to think that "teaching effectively" just had to do with preparation, public speaking, and being smart. All of that can help! But you can also be very smart, very prepared, and very good at public speaking, yet still fail to provide opportunities for discovery.
My main question for self-evaluation for these next three months, when it comes to doing my job, is essentially this: "am I providing opportunities for discovery?"
For me, this often means slowing down, at multiple scales: month-to-month, week-to-week, day-to-day, moment-to-moment. For one, I want to slow down the general pace of the classes to make sure there is enough space for students to discover (and absorb their discoveries) at a reasonable pace. My data science syllabus is currently pretty ambitious, and I will probably need to slow down or even remove some topics over the next few months.
There is also a slowing down to be done at the moment-to-moment scale. This is something I continue to work on, and continue to struggle with, because I often still feel nervous when standing and speaking in front of people. And when I get nervous, I often rush, which tends to be detrimental for my attunement with the audience and the general mood/vibe of the learning environment.
In my view, it is easier to speak quickly and go through a script of sorts in class. Pausing, looking around, giving space for questions, listening and pondering those questions - that is more of an improvisation. It is much more difficult. It is also what gives students real, meaningful opportunities for discovery.
All this to say, by "it's time to lock tf in" I actually mean it's time to slooooooooow dowwwwwwwwnnn...
It's tough to mention learning and pacing and accelerationism and doing more and more and more, without AI coming to mind. Sorry. If you're as tired of reading/thinking about AI as I am, feel free to skip. It's a postscript.
But this seems like a good place to break the seal on a metaphor I've been considering this summer. Of course, there are many helpful metaphors for making sense of AI, and I would have to say that Shanon Valor's is still my favorite (AI is like a mirror, a Cloud Gate-style mirror).
I have added another metaphor through which to see AI, particularly in educational contexts, which is the general notion of debt. Here it is in brief: AI is like a credit card. Each prompt you give to an LLM is like using a credit card to "buy" something, some output. For me, as I mostly use LLM-based coding tools, I am thinking of prompts I use to "make" something.
According to a search on Google and DuckDuckGo, there is only one other page on the internet (as of right now) that says "AI is like a credit card." That post basically says, "AI is like a credit card: it makes spending a breeze, you don't feel the cost while you're 'swiping,' and then the bill arrives" (according to Google's overview).
Being a LinkedIn post, of course, it seemed focused on business debt: shillings, tokens, shekels, USD, bonds, etc. The numbers in the spreadsheet will change and maybe even turn red.
This metaphor is about a different type of coin, a different type of debt, that is more about institutional knowledge and wisdom. The kind of debt that accumulates in organizations from anti-social behavior, or dysfunctional social processes.
Knowledge debt is extremely difficult to pay back, especially compared to dollar debt. It is relatively easy to print more money. Someone basically just puts more 0s in a spreadsheet somewhere. In this sense we can easily "pay" for the tokens/hardware/data centers/etc. Or just declare bankruptcy. Or print more Chuck E. Cheese tickets. You get the point.
But knowledge debt is the kind of debt that someone must REALLY pay for. This is the kind of debt that is created even if I am running a local model on my own local machine for free (which I often do now, with Ollama, woohoo!).
Here is an example of knowledge debt, based on a dream I had recently. Suppose a teacher assigns a problem which the teacher has not actually made. It may be from someone else, or from ChatGPT, or from Claude, etc., but they did not go through the process of writing the problem and/or solving it.
That problem becomes debt. The teacher might "pay off" the debt by going through the problem-solving process and/or adjusting it for their specific class. Sort of like making responsible credit card payments.
But suppose the teacher is asked about the problem in class, and they have not yet gone through the process of solving it. Now, the debt might compound (i.e. if they fail to solve the problem, or solve either incorrectly, etc.).
Like many other issues related to AI, this one echoes earlier challenges. A teacher might have assigned homework problems from a textbook, without doing it themselves, and unready to teach it. Teachers certainly showed crappy code and problems they didn't understand before ChatGPT came on the scene. Give us some credit!
But the prevalence of AI certainly exacerbates the issue, increases the temptation, reduces the friction, etc. So I think this debt framing is worthwhile. When using AI/LLMs to make something that will be maintained, modified, and accessed in the future, someone in the future must pay that debt.
Of course, another corollary/parallel with saying "AI is like a credit card" is that there is a massive company somewhere that is really happy any time you swipe that credit card or prompt that LLM. There are full-time jobs for people to come up with ways to get you to use it more and more and more. "Spend more money, and spend more often!" Credit companies and AI companies spend millions of dollars on campaigning and strategic messaging toward this end. Why?
Edward Tufte once observed that "only two industries describe their customers as users: illegal drugs and the computer industry" (and maybe credit card companies). We could stop any time! No literally, we could stop referring to customers/audiences/people as "users" any time. Maybe we should?
So that is a very meandering explanation of why AI is like a credit card. Again, this is a blog post, not an essay - thanks for reading, fellow nerd!
And perhaps it's worth mentioning, this post was hand-typed with 🩵 in Chicago, IL - as my friend Nick Hagar wrote recently "the process is the whole reason to write something, and it’s what gives the audience something worth reading."