AI and the future of higher education is as urgent question as its impact on security, health and other highly important domains. We should be thinking of AI as not simply a new tool to add to the syllabus.
It seems to me that it touches the two things universities have always claimed as their core purpose:
a) transmitting knowledge and
b) shaping human judgment
Short video of summarizing this post:
Here might be a useful way with the framework pictured above.
The first pair, proposed by Paul Le Blanc in his new book is epistemological vs ontological, describes what education is asking about.
I) The epistemological question is about "what" should a person know.
II) The ontological question is about what does it "mean to be human" once a machine can think alongside you.
I have been thinking about a a second pair of words alongside Le Blanc. btw, this came up because I was in conversation with students during my recent visit to tmy undergraduate campus and had great fun interacting with faculty and students on this subject.
The first is cognitive offloading and the second is cognitive leverage,
III) "Offloading" happens when the machine does the thinking and the person's own capacity never gets built.
IV) "Leverage" happens when the machine makes the person's own thinking sharper and stronger than it would have been alone.
The old management consultant in me cannot resist the temptation to build a 2X2 matrix out of this! (Hence the picture above)
Epistemological + offloading:
Ontological + offloading:
This is subtler and more worrying. Here a student who lets AI settle bigger questions of meaning and value, and feels satisfied without ever doing the work of forming a view.
Epistemological + leverage:
The student usesAI as a demanding partner, pushing a weak argument until it holds up, sharpening real skill rather than replacing it.
Ontological + leverage
This would be the rarest and most valuable case (which we should be all aspiring to)
In this last instance, AI used as a mirror that helps a person see more clearly what judgment, values, and meaning should never be handed over to a machine at all.
I am contemplating that any serious plan for AI in education could be judged by which of these four boxes it is actually building toward.
Some early thoughts on implications of this for Higher Education:
There is a tempting analogy here, that "intelligence" should become as near-free and near-universal as water and electricity.
However, we should remember that water and electricity became cheap and accessible for everyone only after decades of public investment and, often, government mandate. Left alone, markets do not naturally hand out their best product universally.
To stretch the analogy further there is clean water and dirty water. There is salt water and fresh water. There is renewable energy and extractive energy.
So without real investment and real public good intent, we could end up with K shaped outcome of haves and have nots.
In light of this, the questions we could be contemplating for Higher Education include as proposed by Le Blanc:
1. What happens to the future of work, and how should that reshape what universities offer
2. Whether education should shift from the epistemological to the ontological: that is, from what learners need to know, to why universities need to exist at all, and what society wants out of them
3. What it means to live alongside a intelligence that's smarter than us and, for many, more appealing than other people
4. Whether personalized, high-quality education can be made available to everyone for free or near it
5. What happens to the "current education system and establishment" itself whether it's needed or wanted going forward (especially in light of the VERY HIGH cost of education).
Ontological Leverage : that is what I see as the highest pursuit in the era of AI & Higher Education and playing aggressive defense against Epistemological Offloading.
---
AUDIO:
For a deeper interpretative discussion, please find a audio discussion here: audio link
7 comments:
Great ruminations
Loved your quadrant
A great perspective !
Great questions at the end
We should have a conversation some time on this
I also loved your water analogy and especially the parity (purity as parity) question ❗️
But I have a contrasting reflection on the initial point❗️
Yes, for the (piped) water to become available to everyone took decades if not centuries. But I think there is a case why AI reach may not follow that analogy.
Building distribution infra for water was more of question of resource availability and logistics than the willingness of the authorities… it naturally took years to build the pipelines and the pumping stations not to mention the reservoirs…
In the case of AI reach, connectivity part already exists, largely. In other words, the pipes are already there (the bandwidth). The nodes too (the pumping stations). It is just the core infra that needs to be augmented. Largely that is…. In terms of Data Centre build out and the processing power. Of course there is a resource and capex issue but it is much less of a logistical or accessibility challenge as compared to the water (of the past), isn’t it?
Would you agree?
Let’s take water first
Then we’ll apply it to Ai
Water
This is craaaazy
Top 5 countries hold 40% of fresh water reserves
Bottom 50 countries hold 0.2%
USA rank 4
India rank 10
So these two countries are fine
Think of the bottom 50 countries 😳
Another - food
Cheap industrial food is abundant but drives diabetes
Healthy food is expensive!
Now AI
As Chitra points out there is good high quality AI and low quality AI
Haves - high quality tools
Have nots - low grade biased manipulative data!
That might be the implication?!
I have no dispute regarding how the “haves” and “havenots” will have a divide.
The chasm will be greater than the one with something like water.
My contention was more about the implication of the analogy in terms of infra buildout that caused the divide in case of water or electricity. When you lived in the city you got electricity whether you were poor or rich. When one lived in the countryside, he had to put up with electricity unavailability even if one was a rich landlord (until a few decades ago).
That was my only point.
The more sophisticated the technology gets (even biotechnology) the chasm will be bigger in terms of haves and have nots. For sure.
The people with “power” have a great incentive to retain it within their grasp.
But I still think AI may not follow this pattern.
My whole hypothesis would rest on at the juncture where “AI” would break free from being doled out to where it doles out. That juncture is a conjecture of course
Say more with some examples?
I am afraid no real examples one can furnish when it is a hypothesis.
Arguments for both sides rely on past examples of scientific development and advancements of life. Extractions and deductions therefore.
I am concerned with how we treat AI as a development or advancement which is of our making and therefore under human agency. The powers of distribution or its control will be therefore vested with authority(ies) which is (are) human.
AI at some stage will be “the authority” not only because its power to think far exceeds humans but because its goals will supersede that of human goals.
I can see why you may not be comfortable with that hypothesis, because you are more rooted on rational foundation than an abstract strata. Though I disagree it is with full deference to your advocacy
Story in the NY Times - September 27, 2026
Dr. Rachael Bedard, a geriatrician, confesses she uses AI daily and it makes her a better doctor, but worries it makes trainees worse.
She uses OpenEvidence, an AI tool trained only on peer reviewed evidence with citations, to choose antibiotics, interpret labs, consider migraines for dizziness, and discuss lung cancer care. She also uses it for family questions. As a primary care doctor, she faces unknown syndromes monthly and the chatbot reassures her she will always have access to facts. Her mastery is in history-taking and reasoning, not memorizing evolving facts.
Her students also use it, but seem passive and memorize less. She worries they skip productive struggle.
She started med school 20 years ago carrying Pocket Medicine, which required her to know what to look for. Phones, then UpToDate, now OpenEvidence allow iterative, case-specific conversations.
Harvard's Adam Rodman calls the risk never-skilling: in experienced hands AI causes little de-skilling, in novices it prevents skill from forming. Trainees still need to learn to hear problems as stories and develop judgment - sick vs not sick, prioritization.
Solution is not to ban AI, since doctoring is a service that should use best resources. Instead, create AI-free reasoning sessions and AI literacy training. Current trainees may be most affected, learning before we know how to teach well in this paradigm.
Unfortunately, its effects are already seen. My friend Greg tells me that students now ask ChatGPT for answers and cut and paste them as their answers. He has to change tactics every time.
Post a Comment